A sample blog post from the new Nothing is Mere, now with less mereness than ever before:

A Near-Term Policy for Not Getting Killed by AI

Hundreds of scientists, including three of the four most cited living AI scientists, have said that AI poses a very real chance of killing us all.

We’re in uncharted waters, which makes the risk level hard to assess; but a pretty normal estimate is Jan Leike’s “10–90%” of extinction-level outcomes. Leike heads Anthropic’s alignment research team, and previously headed OpenAI’s.

This actually seems pretty straightforward. There’s literally no reason for us to sleepwalk into disaster here. No normal engineering discipline, building a bridge or designing a house, would accept a 25% chance of killing a person; yet somehow AI’s engineering culture has gotten to the point that no one bats an eye when Anthropic’s CEO talks about a 25% chance of “doom” for the entire world.

A minority of leading labs are dismissive of the risk (mainly Meta), but even the fact that “will we kill everyone if we keep moving forward?” is hotly debated among researchers seems very obviously like more than enough grounds for governments to internationally halt the race to build superintelligent AI. Like, this would be beyond straightforward in any field other than AI.

Obvious question: How would that even work? Like, I get the argument in principle: “smarter-than-human AI is more dangerous than nukes, so we need to treat it similarly.” But with nukes, we have a detailed understanding of what’s required to build them, and it involves huge easily-detected infrastructure projects and rare materials.

Response: The latter points are also true for AI, as it’s built today. The most powerful AIs today rely on extremely specialized and costly hardware, cost hundreds of millions of dollars to build, and rely on massive data centers that are relatively easy to detect using satellite and drone imagery, including infrared imaging.

Q: But wouldn’t people just respond by building data centers in secret locations, like deep underground?

(Read more)

The adaptation is… OK?

There are a lot of fun things about it. And it really nails a few characters, and reinvents a few other characters in ways that make them more awesome and interesting than they were in the original comics. (Though overall I’d say there are more misses than hits on characterization, if we weight by importance.)

I don’t want to discourage people from watching the show (especially if this causes them to read the comics ), and I don’t want to say “your enjoyment was wrong!” to people who love the adaptation. I’ve been super delighted to hear people’s positive reviews of the show, and it always makes me happy to hear about which things people found fascinating or moving. I plan to continue watching it myself, and I’m excited to see what comes next.

But I do think there are several extremely core things about the comics that the Netflix adaptation misses on. So here’s my review of the Sandman TV series, and things I’d change:

(WARNING: The rest of this post will spoil important things about the comics that haven’t happened in the Netflix series yet. Don’t read this post unless you’ve read the Sandman comic series. You might also want to see the TV series before reading on?)

.

.

.

.

.

SPOILERS BELOW

.

.

.

.

.

I’d say there are three core problems with the Netflix series.

⠀

1. The series has no Morpheus.

To put it simply: The heart of the comic is Morpheus’ personality and character arc, and Tom Sturridge doesn’t feel like Morpheus, so the show fails.

In the source material, Morpheus’ typical vibe is quite alien, remote, closed off, cold, deliberate, robotic, elegant, precise. He’s a king, but he’s an emotionally dissociated, alien king, a low-key cosmic horror in his own right; and a lot of the comics’ key “glue” is built out of the tension between this aspect of Morpheus and his more human side.

In the Netflix series, Morpheus’ vibe reads to me more as “normal disaffected college-age person”.

Sturridge emotes the FUCK out of every scene. A typical scene will often have his eyes shimmering with tears. He’s a bit stiff, but he cries and laughs and looks shaken quite freely.

I think he sometimes tries to look distant and aloof, but the facial expression I read is somewhere between “vacant” and “trying-to-be-cool”. He looks… a bit like a Zoolander model doing duck-lips?

⠀

⠀

And this “make Morpheus more of an approachable everyman” is very obviously a conscious decision on the part of the showmakers! The story is systematically tweaked in ways that are aimed at humanizing Morpheus, giving him opportunities to express strong emotion (e.g., Jessamy), softening many of the points where he looks weird or evil (e.g., the curse on Alex Burgess), giving him a fast redemption arc in the cases where he does act like a dick (e.g., the events culminating in the season one finale). They want to make his emotional state more obvious and easier to sympathize with.

He feels like… some dude who runs the Dreaming, a guy you could easily be roommates with. Sure, he has magic powers, but we’re all used to normal relatable humans having superpowers in series like Doctor Who and The Avengers. Isn’t Sandman supposed to be a show like that?

(Spoiler warning for the rest of this review: It’s not supposed to be a show like that.)

The original Morpheus’ “resting state” is closed off and internal, with occasional breaks in the façade where deep feelings burst through. In contrast, Sturridge feels to me like he’s exerting active effort to hold himself upright and regal, like a young boy who’s just been granted an important responsibility. (Less alt-Morpheus, more alt-Daniel, maybe?)

Sturridge’s resting state is very human and emotional. To capture Morpheus’ essence, you want someone who can pull off “aloof” more naturally and convincingly, like a younger F Murray Abraham or Alan Rickman. Someone who can completely break down and shatter as a person when the plot calls for it, but who doesn’t parse as “your emo college roommate who has a bit of a chip on his shoulder”. Someone more dangerous, and more self-possessed, and with more depths.

In retrospect, I think the source character’s visual design (chalk-white, an explosion of wild black hair, hollow black eyes with glinting points of white) was actually doing a surprising amount of work in the comics. Especially the eyes.

⠀

To hit emotional beats at all similar to those in the comic, Morpheus needs to have that barrier to seeing how he’s feeling, a bit like trying to peer into the mind of a character wearing dark sunglasses. And there needs to be something very inhuman, uncanny, and otherworldly about Morpheus’ appearance. Missing that in the show, all on its own, takes away a lot of the distinctive personality of Morpheus.

If that specific look is hard to pull off visually in a live-action medium, then I’m totally fine with reimagining what Morpheus looks like (e.g., giving him dark sunglasses, or empty white eyes, or glowing blue eyes, or whatever). The point here isn’t source-fidelity for its own sake, IMO; it’s doing what’s necessary to tell a certain kind of story, regardless of how much that requires being unfaithful to superficial aspects of the comics.

⠀

2. Broadly speaking, the series opts for the same sort of vibes, tone, style, pacing, etc. as the Doctor Who revival.

(Especially obvious examples of this: Rose Walker’s new “Generic Determined Hopeful Person” personality, replacing her complicated old walled-off personality; everything about the Constantine episode, down to the special effects and cinematography.)

But most of the cool content and ideas of The Sandman don’t work in a world and story frame that feels like Doctor Who.

The first few issues of Sandman do have elements of “this is a comic about the adventures of Morpheus as he battles evil and fights crime” to them (though this is balanced by large doses of horror, surrealism, and artsiness that are mostly absent in the Netflix series).

But the rest of the comic series quickly moves away from this focus as Gaiman finds his feet and figures out the kind of story he wants to tell. The story instead evolves into some combination of “campfire stories that, in subtle or peripheral ways, advance the plot / develop the characters“, plus “grounded, unmagical, psychologically realistic slice-of-life storytelling that collides with magical-realism elements“.

Notably, neither of these aspects of The Sandman center Morpheus. Morpheus is neither the hero nor the person whose feelings/reactions/experiences are foregrounded. Rather, the heart of The Sandman is the human world, and the perspectives are human perspectives.

It’s a special treat to periodically run into Morpheus (often very briefly), and an even rarer treat to see past his mask to a true and deep display of emotion. Morpheus is the fascinating, mysterious alien whose puzzle-knot of a mind the comic gradually zeroes in on over thousands of pages; for Morpheus to be a normal sympathetic (albeit mildly aloof) MCU-style protagonist totally breaks the structure of the comics, which depend on him as a strange, complicated, mysterious, magnetic force whose hidden humanity is slowly revealed in stages.

Why is The Sandman compelling, given its really weird structure?

1. Because the execution is amazing (the comics are just real well-written).
2. Because the action is mostly centered on other characters who are easier to empathize with, keeping the whole comic from feeling as cold and rigid as Morpheus himself.
3. Because keeping the focus off of Morpheus allows for a lot of variety and surprises issue-to-issue, and makes the unraveling of the puzzle-knot a slow and satisfying burn.
4. And, crucially, because Morpheus doesn’t behave like an ordinary human being with standard emotions and Hollywood tropes of how-a-mostly-sympathetic-protagonist-is-supposed-to-behave.

If Morpheus does behave just like the other characters in the show, and just like the audience expects, then you lose the whole structure that makes the original story work.

⠀

3. The series doesn’t try to tell a totally new story, either.

In principle, 1 and 2 aren’t fatal, because there are cool new stories one might tell that are very loosely inspired by The Sandman, but don’t try to hit the same story beats and concepts that made the comic awesome. Instead, you can tell a story that’s awesome in very new ways.

But the TV show doesn’t actually try to do this. The series is just trying in a competent and workmanlike way to take a lot of cool dialogue and plot beats from the comic and turn them into a vaguely fun, watchable TV show. (Albeit with nonzero genuinely effective and clever surface-level ideas for adapting a fair number of specific plot points in new ways.)

But 1 and 2 make this a mostly doomed approach; the series either needed to nail the basic characterization and style that made the comic work, or it needed to set off in its own completely new direction.

(I guess there’s some hope that it might start boldly moving in a very new direction in future seasons, in ways I can’t currently anticipate, now that it’s had a successful first season? But I think I understand the values and mindset that produced the first season, and these don’t make me optimistic unless the creative team changes in some way.)

There are subtler issues with the series too, IMO.

⠀

4. Each scene feels very carefully crafted, optimized, almost focus-grouped.

(Gaiman has indeed mentioned that some scenes were cut because test audiences didn’t like them.)

Everything serves a purpose. Elements in the original that seem bizarre and “one damned thing after another” are replaced with elements that obviously and explicitly resonate with the show’s stated themes (dreams vs. reality, responsibility vs. freedom, etc.). Plot arcs always end on satisfying, uplifting notes, rather than deliberately hollow or dissonant ones. Everything has a very clear moral, and the moral always falls within the range of fairly standard morals you’ll see in any other popular mainstream TV show or movie.

The net effect is of a pretty standard, hand-holdy, predictable Doctor–Who-ish fantasy/adventure drama, with occasional unusually-quirky or unusually-violent moments sprinkled in.

Notably, the net effect is not one of peering through a pinhole into what feels like a cohesive, lawful world.

In an actual world, things don’t always happen for reasons. There is real danger, and real unpredictability, exactly because the events aren’t all carefully curated and optimized to achieve standard storytelling goals at every level of granularity.

The Sandman comic is a story, but a lot of what makes the story arresting is its willingness to let Naked Reality creep in — to populate the world with elements that are deliberately never explored or resolved (because in the real world, things exist even if they aren’t plot-necessary and aren’t foreshadowing anything), with elements that are ugly and unpleasant, with elements that are raw and honest, with elements that violate the pattern and rhythm of the previous issue. And then somehow, out of this strangely real-feeling din, the long arc of a deeply satisfying story rises up from the depths, woven together from a thousand dissonant, weird, un-story-ish pieces.

(Which is doubly interesting given that The Sandman is in large part a story about storytelling! Maybe Gaiman on some level recognized that this kind of story gains its power from the interplay between messy non-stories and tidy author-optimized stories; that this is a way for it to keep its soul and humanity, and keep it from becoming one-note in spite of placing an extremely strong focus on a small number of unusually-abstract themes.)

There are reasons that the standard Hollywood/TV tropes are as they are. You really will lose some viewers if you stray from the Path. You’ll confuse people, upset people, make missteps (not just superficial missteps, but genuine deep artistic missteps ⁠— good art is genuinely risky, and failures will happen).

But while each local change can be justified on the basis of best storytelling practices, wanting to make a cool story intelligible and grabby to a wider audience, etc., the net effect of thousands of artistic decisions like this is a show that feels relatively flat, soulless, and over-polished.

If it were just hopeless to tell a story like that in TV or film (for reasons of medium, or because of problems with the industry’s incentives), then I’d be more resigned about all of this. But in fact, I think the right genre, tools, techniques, aesthetic, etc. is quite common in the industry today: it’s more or less the approach you see in artsy HBO dramas (think Carnivàle or Six Feet Under). Sandman just needs an approach that’s more like that, and less like Doctor Who or the MCU.

⠀

Beyond “the series feels too much like a Hollywood movie and not enough like a world”, there’s also:

5. The show lacks a sense of the cosmic, mysterious, ancient, and archetypal.

This is a harder point than the others to put into words, and I think it’s easy to overlook how important this is for making a bunch of the other things work. A lot of the power and poetry of the original comic relies on its use of established myths and patterns that feel somehow timeless and true; “more true than truth itself”, is how I might put it if I were Gaiman.

If I were to try to grope at what makes “archetypalness” important to The Sandman, it would maybe be:

• Archetypes that feel old and “mysteriously true” create a point of tension with the real/modern/messy/human world. And this tension is shown and felt, not just told: both sides have serious emotional pull for the reader.
• Archetypes are an easy way to create beauty, because (a) they contain a lot of pre-existing emotional content the author can draw on, and (b) drawing on cached concepts allows more complex and ensemble-y storytelling that avoids feeling cluttered or inelegant/forced. It’s one way to make the world feel like a large, coherent world (rather than a focus-grouped movie script), without resorting to mere realism or giant exposition dumps.

The “tension between the human and supernatural worlds” thing also connects up to the show’s decision to gender- and/or race-swap a lot of the cast.

Gender- and race-swapping is a great way to avoid making characters a rote rehash of the comics, and to see what new character dynamics and vibes fall out when you reroll the dice. (Plus it can help make a show/comic that’s nominally about Intergalactic Cosmic Beings feel less like it’s confined to a few neighborhoods in the US and UK.)

But the specific way the show does this is pretty weird. If the show-runners want the show to feel less white and male on the margin, it’s pretty easy to think of ways to do that (e.g., gender-swap Matthew, make the Endless all-West-African or all-Persian, etc.) that work well with the show’s themes, or even strengthen those themes. But many of the actual flips instead hollow out or undermine the themes.

On the whole, I don’t think this aspect of casting matters a ton. But it makes for an interesting case study in ‘losing sight of the basic tensions that the comic is built around’.

A central theme in the Sandman comic is that the Dreaming, the Endless, the gods, the Kindly Ones, etc. are all (with occasional exceptions, noteworthy exactly because they’re exceptions) forces of conservatism and non-humanism. Indeed, Morpheus’ arc (and the arc of many of the characters) is about overcoming this (or tragically failing to do so). The universe is a hostile place, governed by old Rules and Patterns that often echo humanity’s concepts, but rarely echo or push-in-favor-of humanity’s highest values and hopes for a better future.

⠀

⠀

Notably, this conservatism doesn’t have to be European in order to ring thematically true. You can model Lucien on a circa-1700 Chinese scholar-official, rather than a circa-1700 western-Europe librarian, and he’ll feel like an old archetype, and still be able to help build up a duality between the Dreaming’s old-fashioned rule-following and the waking world’s beautiful diversity and aliveness.

(I think it actually would have been cool to build the entire Dreaming around, e.g., old Chinese aristocratic tropes. Choosing a single overarching visual language can reinforce the traditionalism of these places and institutions.)

If you want the Dreaming to serve as a (complicated, slowly-starting-to-change-and-evolve) foil to the waking world’s aliveness and modernity and freedom, the one archetype you can’t lean into for characters like Lucien is a really new and refreshing archetype like Exciting Hip Culture-Straddling Person Who Looks Like A Denizen Of A Badass Timehopping Shadowrun Campaign.

If you want to nail the core conflict in the original comics, you can’t make Lucien a rebellious exciting fresh new counterpoint to Morpheus — especially not one whose vibe matches a just-starting-to-become-popular-in-the-21st-century new media archetype like Strong Female Queer-Fashion-Icon Supporting Character With Attitude.

⠀

⠀

Thematically, you can very much do that for the new kid in town, Matthew. For similar reasons, gender-swapping Constantine and race-swapping Rose, Jed, and Unity seems actively good to me, since (to my eye) it makes those characters feel less like standard Anglophone storytelling tropes.

(If Constantine feels too traditionally noir and rugged-protagonist-y, he’ll undermine the distinction between Archetypal Dreams and Messy, Decidedly Non-Archetypal Humans.)

The progressive, queer, humanistic, etc. stuff mostly goes in the human world. If you lean into this for the old, important, influential non-humans we meet early in the series, you’re liable to make the Dreaming feel like a vibrant cosmopolitan futurist utopia, as opposed to the complicated, disturbing, antiquated, foggy, nightmare-y castle-and-grounds it is in the source material.

⠀

⠀

Compare the voice-over that introduces the Dreaming in the Netflix adaptation (“When the waking world leaves you wanting, and weary, sleep brings you here — to find freedom and adventure!”) to its introduction in the comics:

⠀

⠀

If you go the route of making the Dreaming just another cool wonder-inspiring gee-whiz fantasy setting, like you might see in movie-Narnia or Doctor Who or the MCU, then you do get to open the TV series with an exciting hopeful fantasy-paradise CGI sequence.

⠀

⠀

The standard Hollywood visual language is popular for a reason!

But if you make a habit of making things normal/standardized in this way, and you don’t have an offsetting habit of making other things more weird than they were in the source material, you end up with just another fantasy/superhero TV show, a minor variation on the established theme.

Death is deliberately written as an exception to the “non-humans are old and stodgy and rule-bound” rule, and you can add other exceptions if you want — especially if they show up a lot later in the story, or if they’re new kids on the block like Matthew. But if you pile on more and more exceptions to the rule from the very start, then the rule just doesn’t exist and the non-human world turns into “just more world”. (Albeit with more neat CGI than the human world.)

For some quick thoughts on how I might have started a series like this if I were King Of Sandman Adaptations, see this Google Doc.

Elizabeth van Nostrand wrote an Aug. 30 blog post, Long Covid Is Not Necessarily Your Biggest Problem, concluding that “for vaccinated people under 40 with <=1 [comorbidity], the cognitive risks of long covid are lost in the noise of other risks they commonly take”.

She also concludes that

[…] your overall risk of long covid is strongly correlated with the strength of the initial infection. […]

Van Nostrand estimates estimates that the risk of hospitalization for a vaccinated person who catches Delta is:

• 0.38% for a healthy 30yo man;
• 0.24% for a healthy 30yo woman;
• 0.58% for an asthmatic 25yo man;
• 0.92% for a 40yo obese woman.
  

And:

[…] My tentative conclusion is that the risks to me of cognitive, mood, or fatigue side effects lasting >12 weeks from long covid are small relative to risks I was already taking, including the risk of similar long term issues from other common infectious diseases. Being hospitalized would create a risk of noticeable side effects, but is very unlikely post-vaccine (although immunity persistence is a major unresolved concern).

I want to emphasize again that ‘small relative to risks you were already taking’ doesn’t necessarily mean ‘too small to worry about’. For comparison, Josh Jacobson did a quick survey of the risks of driving and came to roughly the same conclusion: the risks are very small compared to the overall riskiness of life for people in their 30s. Josh isn’t stupid, so he obviously doesn’t mean ‘car accidents don’t happen’ or ‘car accidents aren’t dangerous when they happen’. What he means is that if you’re 35 with 15 years driving experience and not currently impaired, the marginal returns to improvements are minor.

[…] What this means is not that covid is safe, but that you should think about covid in the context of your overall risk portfolio. Depending on who you are that could include other contagious diseases, driving, drugs-n-alcohol, skydiving, camping, poor diet, insufficient exercise, too much exercise, and breathing outside [during wildfire season]. If you decide your current risk level is too high, or are suddenly realizing you were too risk-tolerant in the past, reducing covid risk in particular might not be the best bang for your buck. Paying for a personal trainer, higher quality food, or a safer car should be on your radar as much as reducing social contact, although for all I know that will end up being the best choice for you personally.

In Long COVID: Much More Than You Wanted To Know, Scott Alexander expresses stronger worries about long COVID (albeit with a broader definition of ‘long COVID’ that includes very mild symptoms like ‘reduced sense of smell’):

The prevalence of Long COVID after a mild non-hospital-level case is probably somewhere around 20%, but some of this is pretty mild.

[…]

Vaccination probably doesn’t change the per-symptomatic-case risk of Long COVID much

Alexander’s Fermi estimate:

About 25% of people who get COVID report long COVID symptoms. About half of those go away after a few months, so 12.5% get persistent symptoms. Suppose that half of those cases (totally made-up number) are very mild and not worth worrying about. Then 6.25% of people who get COVID would have serious long-lasting Long COVID symptoms.

[…] I’m going to round all of this off to about 1% – 10% per year of getting a breakthrough COVID case (though obviously this could change if the national picture got better or worse). Combined with the 0.4% to 6.25% risk of getting terrible long COVID conditional on getting COVID, that’s between a 1/150 – 1/25,000 chance of terrible long COVID per year.

[…] I find the 1/150 risk pretty scary and the 1/25,000 risk not scary at all, so, darn, I guess there’s not yet enough data to have a strong sense of how concerned I should be.

Zvi Mowshowitz comments on Alexander’s post in Covid 9/2: Long Covid Analysis:

[…] What I’m confused by is how he uses the data he reports in this section to end up at 20%, since he quotes studies where (Long Covid percent in Covid group minus Long Covid percent in control group) is respectively at most 28%, 12%, 17%, 13% and 13%, two of which lack a control group. If we naively average that we get 17% minus a few percent for the missing control groups, so maybe 15%. Scott seems to be buying that ‘any symptom at all’ is a reasonable standard here, and that asking ‘did you have Long Covid?’ is ripe with false negatives.

[…] I think we can safely throw out the upper part of [Scott’s] range, as I think a 10% chance of breakthrough symptomatic Covid within a year isn’t reasonable if you do a little math, and it’s starting at 25% which seems higher than the studies referenced above would suggest, so I think the range here would be more like 1 in 1,000 to 1 in 25,000.

[…] Long Covid seems legitimate, and worth a nonzero amount of effort to minimize, but my model says it is mixing a lot of things together, is largely typical of what happens after being sick, is protected against by vaccines similarly to how they protect against symptomatic disease, and in many studies they go on a fishing expedition for symptoms then attribute everything that happens chronologically after Covid to Covid.

Van Nostrand summarizes her disagreements with Alexander in Alternate Views on Long Covid:

– […] I think his studies are too small and sample-biased to be meaningful.

– He thinks my studies (especially Taquet) didn’t look at the right sequelae.

– I was only looking at cognition (including mood disorders), whereas he looked at everything.

Scott also didn’t do age-specific estimates, although I’m that’s not a crux because I expect other post-infection syndromes to worsen with age as well.

I intended to include fatigue in my analysis of cognitive symptoms but in practice the studies I weighted most highly didn’t include them. Scott’s studies, which he admits are less rigorous although we differ on how much, did include them. Why the hell aren’t the large, EHR-based studies with control groups looking at fatigue? […]

• Institutional failure.

Chinese officials suppressed early information about the virus. The WHO and the US CDC consistently spread misinformation and shoddy science throughout the course of the pandemic, and showed a shocking inability to understand and communicate basic distinctions like ‘we don’t know whether X’ versus ‘we know that not-X’.

World governments banned challenge trials for a full year based on imagined fears that they might prove unpopular, only to learn that they were very popular with the public once we bothered to check.

The US FDA banned COVID-19 testing and research during the critical early days of the pandemic in the US, and caused tens of thousands of deaths by refusing to approve well-tested vaccines in wide usage in the rest of the world. The developed world (and especially the European Union) massively under-invested in vaccines, spending thousands of dollars in human life and welfare to save pennies.

Most remarkably, many of these errors recur across many different countries, suggesting deep dysfunction in the way global elites generate, evaluate, and propagate ideas.

• Paranoid passivity.

A common theme in many of the above dysfunctions is a willingness to kill hundreds of thousands of people through inaction, before decisionmakers are willing to risk taking any unpopular action.

The Copenhagen Interpretation of Ethics points to one possible explanation, but it shouldn’t be forgotten that leaders are to a large extent giving the public what they want in all of this — it’s just that the public has pathologically low standards and a bizarre level of change aversion.

• Nationalism.

… But all of that may turn out to be a footnote in light of recent events in India. History may instead remember COVID-19 as a pandemic whose death toll largely occurred after vaccines were widely available, and one that mostly afflicted the poorest parts of the world.

The story of the pandemic may be: ‘The developed world made the strategic decision to prioritize themselves over the developing world. An effective genocide ensued. Crematoria spit their smoke into the sky while tens of millions of unused vaccines sat where they had been for months, gathering dust in storage in the US, useless even to Americans because their FDA refuses to approve the vaccines for domestic use too. They just sat there.’

• Biotech revolution.

… Or even that may turn out to be a footnote. History may remember COVID-19 like this:

‘By spurring the world to experiment with new vaccine tech, the COVID-19 pandemic ended up saving vastly more lives than it cost.’

This is even more uncertain, but if true, it raises major questions about why we couldn’t act sooner. Illnesses that kill millions of people don’t become less deadly just because we’re used to them. Yet somehow, it took a pandemic for human civilization to start taking human death and disease seriously to this degree.

Excerpts from Julia Galef’s interview of Bryan Caplan:

Bryan Caplan: [… T]here’s a standard story that almost everyone tells about why education pays in the labor market, and it just says: you go to school, they pour some skills into you, you’re better at your job, and so you get paid more. What’s the problem?

And I’m happy to say, sure, that’s part of the story.

But I say there’s also a much bigger part of the story that rarely gets discussed, and that is that when you do well in school, you impress others. You get certification. You get stamped with a sign of approval saying “Grade-A Worker”. And my story is that the majority—in fact, a large majority—of the payoff from education actually comes from this.

Selfishly speaking, that doesn’t matter so much. But from a social point of view, it matters tremendously. Because if the reason why people get paid more for school is because they learn more skills, then basically it’s a way that taxpayers invest in our productive capacities and then we produce the very wealth that we are being paid for.

But on the other hand, in the signaling story, the main thing that’s going on is that you’re getting paid because you’ve impressed employers. And if everyone has a bunch of stickers on their head, this doesn’t mean everyone gets good jobs or gets paid a lot. It just means that you need a lot of stickers in order to get a job.

So the biggest sign of this, I would say, is what’s called a credential inflation, which is you just now need more education to get a job that your dad or grandfather could’ve gotten with one or two fewer degrees.

Julia Galef: And what kind of signal are you mostly pointing at? Is it the signal that someone was good enough to be accepted into a college, or the signal that someone was good enough to graduate with the grades that they did?

Bryan Caplan: Yeah, so the graduation seems like it’s a lot more important. Because if it were the first story, if it were just you get a great signal by being accepted, then people would take their admission letters and shop them around employers saying, “I got into Harvard and Stanford, so what are you going to offer me, Goldman Sachs?” And in practice, that doesn’t seem to work very well.

So I think if you’re wondering why, I would say that there’s something very odd about a person who tries to do that. They seem like they’re trying to skip out on this sacred institution of our society. So, yeah, employers are understandably nervous about someone so weird that they would get into Harvard and then try to weasel out of it.

So in terms of what is it people are signaling, I’d say it’s really a big package of different traits. Intelligence, obviously, but it’s not just that. That’s too easy to measure by itself. It’s also work ethic. And then finally, sheer conformity, which again, is very important on the job. Someone could be really smart and really hard-working, but if they’re defiant, if they don’t play as part of the team, then they’re almost useless to you. And I say really to understand a lot of what’s going on with education, we have to focus on this conformity signaling.

[… T]he whole idea of signaling is that if you come up with a really cheap way of signaling, the result isn’t that you get your signal across at a low cost, but rather that you just have to do more of it. A key idea in the signaling model is if you found a way of cutting the cost of signaling in half, the result wouldn’t be that we do half as much signaling. The result would be that we signal for twice as long. […]

My favorite example of this is suppose that someone comes up with a new way of making synthetic diamonds at 10% of the current cost. And my question is, how long would it take before people either stopped giving diamond engagement rings, or they started giving rings that were enormous?

And the key point is that since what you’re signaling with that ring is that you’re willing to go and put in a lot of money into something to indicate your devotion, if the cost per carat of diamond were to fall, it’s not that we would just keep giving the same diamonds that we’re currently giving. Instead, people would say, “Well that doesn’t really convince people very much anymore. It doesn’t say much anymore. I’d better go and either get an even bigger diamond or give something that can’t be synthesized.”

And a lot of it is really the same for education. If you were to go and have, say, free college for all, the result wouldn’t be that everybody with a college degree can get the kind of jobs that people get with it now. Instead there’d be an army of extra people going, and then you might need a Master’s degree or another advanced degree to be considered worthy of an interview.

[…M]ost specialists in both education and labor economics, they’re only looking at income. So they’re looking at the effects of education, and then there is this really circular effort to say, “Well, since there’s a big effect on the income of the person, they must’ve learned something useful,” and you say, “Yes, but the signaling model predicts the very same thing.” So that’s a big issue.

And there is an idea of, “Well of course we all know that the people are learning tons of useful stuff.” And then when you say, “Well, actually, they’re learning a ton of stuff they’re never gonna use.” And this is then where economists will often retreat to, “Oh, well, they’re learning how to learn, learning critical thinking, it doesn’t really matter what the subject is.”

And then I’ll say there’s something they really don’t know about, which is: in educational psychology, they’ve been studying this very issue for a hundred years. They want to find evidence of learning how to learn. They want to find evidence that critical thinking is being successfully taught. And yet, after a hundred years, they’re really pretty shell-shocked and say, “Look, we’re just not finding much sign of this broad, general inculcation of thinking skills that educators love to believe is actually happening.” So that’s the stuff I’d say most economists are just totally unaware of.

[… S]tudents seem so focused on getting easy As. If you were in school to acquire skills, this is pretty perverse. But if you’re in school to impress employers, then it’s pretty easy to understand why you want an easy A, because the employer doesn’t know that it was an easy A. If you find the easiest teacher of real analysis in the country, get an A+ in exchange for doing some arithmetic, people look at that and say, “Wow, he’s got an A plus in real analysis. Wow, look at that guy.” So that makes sense.

The practical implication: if, e.g., college as it exists in the real world is largely a zero-sum arms race to signal pre-existing traits (like intelligence and disposition to conform / accept instructions), rather than a positive-sum opportunity to actually learn useful or enriching material, then causing more people to go to college doesn’t improve people’s lives in aggregate.

Just the opposite, since college is expensive in time and money. If you get another 10% of people to go to college, then everyone else has to burn that many more resources to keep up in the signaling competition, but there’s still the same pool of new jobs, and people are still roughly as good at those jobs as they would have been without the education. Which means that everyone is now burning more resources just to not fall behind relative to everyone else in the ‘signal you’re a good worker’ game. Like forcing everyone in a race to run twice as fast, without doing anything to increase the reward for absolute performance at the end.

So, for example, subsidizing college education is a terrible idea that actively hurts people. A better case could be made, if anything, for taxing it as a source of net harm to society, to try to reduce how much time people spend at college etc. and thereby put more time, money, and other resources in people’s hands. Forcing poor people to get more years of education doesn’t appear to materially benefit them at all in aggregate, but handing them back money and free years of their life certainly does.

If you want to enrich people with cool ideas as an end in itself, because cool ideas are cool, then give people Internet access and free time and let them decide how to use that time. Don’t force them into camps where they have to learn classics and jump through hoops in order to be able to pay medical bills, start a family, etc. later in life.

From Caplan’s book:

Higher education is the only product where the consumer tries to get as little out of it as possible. […]

Some big blatant facts are inexplicable without the signaling model.

[1.] The best education in the world is already free. All complaints about elite colleges’ impossible admissions and insane tuition are flatly mistaken. Fact: anyone can study at Princeton for free. While tuition is over $45,000 a year, anyone can show up and start attending classes. No one will stop you. No one will challenge you. No one will make you feel unwelcome. Gorge yourself at Princeton’s all-you-can-eat buffet of the mind. Colleges do not card. I have seen this with my own eyes at schools around the country.

If you keep your learn-for-free scheme to yourself, professors will assume you’re missing from their roster owing to a bureaucratic snafu. If you ask permission to sit in, most professors will be flattered. What a rare pleasure to teach someone who wants to learn! After four years of ‘guerrilla education,’ there’s only one thing you’ll lack: a diploma. Since you’re not in the system, your performance will be invisible to employers.

[… 2.] Failing versus forgetting. You’ve studied many subjects you barely remember. You might have motivated yourself with, ‘After the final exam, I’ll never have to think about this stupid subject again.’

[… 3.] Easy As. Students struggle to win admission to elite schools. Once they arrive, however, they hunt for professors with low expectations. A professor who wants to fill a lecture hall hands out lots of As and little homework.

[… 4.] Cheating. According to human capital purists, the labor market rewards only job skills, not academic credentials. Taken literally, this implies academic cheating is futile. Sure, a failing student can raise their grade by copying an A+ exam or plagiarizing a term paper from the Internet. Unless copying and plagiarizing make people more productive for their employer, however, the human capital model implies zero financial payoff for the worker. […]

The human capital model doesn’t just imply all cheaters are wasting their time. It also implies all educators who try to prevent cheating are wasting their time. All exams might as well be take-home. No one needs to proctor tests or call time. No one needs to punish plagiarism—or Google random sentences to detect it. Learners get job skills and financial rewards. Fakers get poetic justice.

[… 5.] Teachers have a foolproof way to make their students cheer: cancel class. If human capital purists are right, such jubilation is bizarre. Since you go to school to acquire job skills, a teacher who cancels class rips you off. You learn less, you’re less employable, yet your school doesn’t refund a dime of tuition. In construction, contractors don’t jump for joy if their roofers skip shingling to go gambling. In school, however, students jump for joy if their teachers cancel class to attend a conference in Vegas.

When students celebrate the absence of education, it’s tempting to blame their myopia on immaturity. Tempting, but wrongheaded. Once they’re in college, myopic, immature students can unilaterally skip class whenever they like. Why wait for the teacher’s green light? For most students, there’s an obvious answer: When you skip class, your relative performance suffers. When you teacher cancels class, everyone learns less, leaving your relative performance unimpaired.

Human capital purists must reject this ‘obvious answer.’ Employers reward you for your skills, not your skills compared to your classmates’. Signaling, in contrast, takes the ‘obvious answer’ over the finish line. Why do students cheer when a teacher cancels class? Because they’ve escaped an hour of drudgery without hurting their GPA.

And another excerpt—the following is a relatively minor argument in a big 400-page book, but I’ve come back to it a few times, so I’ll put it here too. As a philosophy major, I get to do this without looking like I’m lording my major over others…

We can ballpark the practicality of higher education by looking at the distribution of majors. Table 2.1 breaks down all bachelor’s degrees conferred in 2008-9 by field of study—and rates their usefulness.

High usefulness: Defenders of the real-world relevance of education love to invoke engineering. Engineering students learn how to make stuff work; employers hire them to make stuff work. Engineering has well-defined subbranches, each with straightforward applications: electrical, mechanical, civil, nuclear. Before we get carried away, we should accept a key act: Engineering is a challenging, hence unpopular, major. Psychologists outnumber engineers. Artists outnumber engineers. Social scientists plus historians outnumber engineers almost two to one. […]

Medium usefulness: Majors like business, education, and public administration sound vaguely vocational and funnel students toward predictable occupations after graduation. At the same time, they teach few technical skills, and nonmajors readily compete for the same jobs. While you could dismiss these majors as Low in usefulness, let’s give them the benefit of doubt. You don’t need a business degree to work in business, but perhaps your coursework gives you an edge. You don’t need an education degree to land a teaching job, but explicitly studying education could enhance your teaching down the road. […] By this standard, about 35% of majors end up in the Medium category. […]

Low usefulness: The status of most of the majors in this bin [which contains 40% of all bachelor’s degrees]—fine arts, philosophy, women’s studies, theology, and such—should be uncontroversial. Liberal arts programs uphold the ideal of ‘knowledge for knowledge’s sake.’ Few even pretend to prepare students for the job market. You could argue I underrate the usefulness of communications and psychology. Don’t they prepare students to work in journalism and psychology? Yet this objection is almost as naive as, ‘Don’t history programs prepare students to work as historians?’ Psychology, communications, and history’s usefulness is Low because they prepare their students for fields where paying jobs are almost impossible to get. In 2008-9, over 94,000 students earned their bachelor’s in psychology, but there are only 174,000 practicing psychologists in the country. In the same year, over 83,000 students earned their bachelor’s degree in communications. Total jobs for reporters, correspondents, and broadcast news analysts number 54,000. Historians, unsurprisingly, have the bleakest prospects of all. There were over 34,000 newly minted history graduates—and only 3,500 working historians in the entire country. […]

The staunchest defenders of education reject the idea of sorting subjects and majors by ‘usefulness.’ How do you know Latin, trigonometry, or Emily Dickinson won’t serve you on the job? A man told me his French once helped him understand an airport announcement in Paris. Without high school French, he would have missed his flight. Invest years now and one day you might save hours at an airport. See, studying French pays!

These claims remind me of Hoarders, a reality show about people whose mad acquisitiveness has ruined their lives. Some hoarders collect herds of cats, others old refrigerators, others their own garbage. Why not throw away some of their useless possessions? Stock answer: ‘I might need it one day.’ They ‘might need’ a hundred empty milk cartons.

Taken literally, the hoarders are right: there is a chance they’ll need their trash. The commensense reply is that packing your house with trash is almost always a bad idea. You must weigh the storage cost against the likely benefits. […] ‘No one knows if this trash will come in handy’ is a crazy argument for hoarding trash. ‘No one knows if this knowledge will come in handy’ is a crazy argument for hoarding knowledge.

More discussion from Julia Galef’s podcast:

Julia Galef: You mentioned the case in which a teacher cancels class and the students are all happy about that and say like, “Jeez, if it was really about gaining skills that they expect to increase their productivity and value to future employers, then why would they be happy?”

You know, they already paid for tuition, and now they’re just getting less for their money. Which I do think is a suggestive and striking fact about the world.

But I felt like you didn’t quite give enough space to the alternate explanation of that—which is just, you know, people buy gym memberships because they want to lose weight or get fit, and then they find excuses not to go to the gym, or they’re happy when there’s a holiday and the gym is closed, so they don’t have to go to the gym.

It just feels like there’s this common phenomenon of a tension, of struggle between your present self’s interests and your future self’s interests, and this leads to a lot of behavior that otherwise looks irrational.

Bryan Caplan: Yeah, so I think I did have a couple sentences on that point, but you’re right, I could’ve talked more about it. […] But the main thing I say is that this myopia can explain why students don’t show up on a regular day. And yeah, typical college class in the middle of the semester, barely half the students are showing up. And that, I think you might say, “Well, it’s just myopia,” because they’re going and putting this money in, and they’re gonna get worse grades, and their life is going to be worse as a result.

But of course, there’s all the students who do show up, and why is it that those students are also happy when you cancel class? And that one seems to be that well, then I get to have this holiday without having to worry about the material that I failed to learn and that is going to lead me to get lower grades.

So yeah, I think in terms of just low attendance, you can explain it with myopia. But why people see a big difference between skipping class when everyone else is doing it, and skipping class when only half of the people are doing it, or when only you’re doing it—that’s where I think that you can detect the signaling element. It’s like, “I don’t mind missing it if everyone else misses it, but if I’m the only one missing it, then I’m dead, so no. I’ll go.” […]

Julia Galef: I just want to zoom out for a moment to note that I’ve been honing in on the parts of your argument that I find relatively less convincing—but I actually do find your argument overall pretty convincing. And if you’re correct about the standard [view] being closer to 10% signaling, I’m closer to 80% than 10%.

But yeah, I don’t know, I’ve just been thinking during our conversation about cruxes of disagreement between you and me. And I think probably one of them is I just expect that companies are less rational than you expect they are. And so I would just be less surprised if they were leaving large amounts of money on the table. Or less surprised if societal inertia or irrational biases were doing a lot of the work here. Which just changes the whole way you make sense of what’s happening.

Bryan Caplan: I mean, what’s funny is for an economist, I’d say I’m very open-minded about this stuff, and there are a bunch of cases where I’ll say, “Yeah, it looks like firms are actually not maximizing profits,” or, “They’re leaving money on the table.” But again, I think the cases that are well-documented are ones where it’s more marginal.

And there is actually a big body of literature on how firms that don’t maximize profits and have low productivity per worker have much higher attrition rates than other firms. And on the other end, the firms that have usually high productivity are just more likely to not only survive, but also to expand. It’s another thing to say, your firms are leaving money on the table for five years, but to say that it’s gone on for decades, again, this seems to go against most of what we know about selective attrition in growth of firms.

Julia Galef: Okay. Well, that’s a way bigger crux of disagreement than we can resolve in two minutes, so I’ll leave it at that. I just thought it was interesting to point out.

Bryan Caplan: Yeah, totally.

Julia Galef: And I want to make sure that I don’t forget to tell you about an ironic thing that I’ve noticed, that’s very relevant to your case, which is: Philosophy departments, in their “Why you should be a philosophy major” page on their departmental website, they always cite statistics about how there’s a high return to a philosophy major, in terms of the starting salaries you get offered. And they say, “See, this is proof that philosophy majors teach you critical thinking skills!”

Bryan Caplan: Yeah, that’s terrible.

Julia Galef: Which is especially ironic, because they’re confusing correlation and causation, which is an example of poor thinking skills in their very argument! That just struck me as a very Bryan-flavored observation.

Bryan Caplan: Yeah. Plus it’s not even true that a philosophy major is well-paid.

Julia Galef: Oh?

Bryan Caplan: It’s not at the bottom of the distribution by any means, but… actually, now that I think about it, normally the numbers that I look at actually correct for test scores. So it might be that, yeah, philosophers do come in with very high test scores. So it might be that if you just look at raw means, what you’re saying is true.

Julia Galef: Right, yeah.

Bryan Caplan: But if you go and look at how people who had the same test scores but who majored in something else do, then I think philosophy does pretty poorly. Especially if you’re not looking at people who go on to get a law degree or something like that. Those people are probably pulling up the average a lot.

Students who do unschooling seem to do totally fine; and differences between education approaches mostly don’t seem to change much. So on the face of it, mandating decades of universal formal education seems to just be burning value.

There may be subtle society-wide effects of forcing people to spend a large chunk of their life in something like a well-intentioned labor camp; but the balance of these effects might be on things like “how conformist society is in general” that I don’t think are good things to optimize for.

Some related discussion by Scott Alexander:

• “Against Tulip Subsidies“
• “SlateStarCodex Gives a Graduation Speech“

___________________________________________

The first COVID-19 human challenge trials have been approved… in February 2021, instead of February 2020.

If this is a good idea now, when the benefits are vastly lower and the risks are only slightly lower, then it was probably also a good idea a year ago. Opponents of human challenge trials should think hard about the background heuristics that caused them to get this one wrong, so we don’t have to repeat this tragic error again.
___________________________________________

From Zvi Mowshowitz on Feb. 11:

The good news is that Johnson & Johnson has applied for emergency use authorization, and they are going to get it. There’s going to be a sprint to review the data in which every second will be used! Which started right after the official application, because you can’t review data that hasn’t been submitted in a completed application, that’s physically impossible, what are you even talking about. 

The bad news is that it’s going to take three weeks to get to the meeting, likely with additional time after the meeting before we can distribute the vaccine, but hey:

Last week I went over how we know the Johnson & Johnson vaccine is safe and effective, and there are millions of doses waiting to be distributed, and there’s no good reason we can’t start that process yesterday.

I do realize that there is a difference between, as Scott Alexander discusses, the FDA’s need to be legible and reliable, and follow proper procedures, versus my ability to apply Bayesian reasoning. 

That doesn’t mean this needs to happen three weeks after application, and attempts to justify that timeline are obvious nonsense. 

Mostly, it’s a call and response. You say ‘why are we letting people die for no reason?’ and they say ‘Thalidomide!’ and ‘people won’t trust it.’

So basically, one time someone had a drug that wasn’t safe. We didn’t approve that drug because our existing review process made it look unsafe, so in response to that we created a more involved and more onerous process, as opposed to noticing that the previous process actually worked in this case exactly as designed. Then we use this as a fully general excuse to freak everyone out about everything that hasn’t gone through this process, and then use that freak out (that, to the extent it exists which it mostly doesn’t, is directly the result of such warnings) as our reason to force everything through the process. Neat trick. 

Oh, and did I mention that the ‘safety data’ that requires three weeks to review is, and I quote it in its entirety, ‘nothing serious happened to anyone at all, and no one was struck by lightning.’ Either J&J has created a safe vaccine, or J&J  is committing a fraud that will be caught and get everyone involved arrested within three weeks, or they’re committing a fraud so effectively that the review won’t catch the fraud and won’t help. Those are the only possibilities. If the data isn’t fraudulent then the drug is safe, period. […]

On the actual J&J vaccine, I don’t know what more there is to say. As with Moderna and Pfizer, they’ve already done the actual approval process and confirmed that it’s going to get approved before they applied, and now we’re delaying in order to make it clear we are Very Serious People who Follow Proper Procedure and are not In Bed With Industry and Putting People At Risk or Destroying Trust in Vaccines by going ‘too fast.’ Or something like that. 

We have now done this three times. It’s one thing to have the first vaccine application point out that there’s weeks of lost time. It’s another thing to not have fixed the problem months later.

___________________________________________

From Zvi on Mar. 4:

Meanwhile, now that we were provided a sufficiently urgent excuse that we were able to show that mRNA vaccines work, we’ve adopted them to create a vaccine for Malaria. Still very early but I consider this a favorite to end up working in some form within (regulatory burden) number of years. It’s plausible that the Covid-19 pandemic could end up net massively saving lives, and a lot of Effective Altruists (and anyone looking to actually help people) have some updating to do. It’s also worth saying that 409k people died of malaria in 2020 around the world, despite a lot of mitigation efforts, so can we please please please do some challenge trials and ramp up production in advance and otherwise give this the urgency it deserves? And speed up the approval process at least as much as we did for Covid? And fund the hell out of both testing this and doing research to create more mRNA vaccines? There’s also mRNA vaccines in the works for HIV, influenza and certain types of heart disease and cancer. These things having been around for a long time doesn’t make them not a crisis when we have the chance to fix them.

___________________________________________

From Bryan Caplan on Feb. 16: Bioethics: Tuskegee vs. COVID. Excerpts:

[…] Why do bioethicists habitually invoke the Tuskegee experiment?  To justify current Human Subjects Review.  Which is bizarre, because Human Subjects Review applies to a vast range of obviously innocuous activities.  Under current rules, you need approval from Human Subjects merely to conduct a survey – i.e., to talk to a bunch of people and record their answers.

The rationale, presumably, is: “You should only conduct research on human beings if they give you informed consent.  And we shouldn’t let researchers decide for themselves if informed consent has been given.  Only bioethicists (and their well-trained minions) can make that call.”

On reflection, this just pushes the issue back a step.  Researchers aren’t allowed to decide if their human experiment requires informed consent.  However, they are allowed to decide if what they’re doing counts as an experiment.   No one submits a formal request to their Human Subjects Review Board before emailing other researchers questions about their work.  No professor submits a formal request to their Human Subjects Review Board before polling his students.  Why not?  Because they don’t classify such activities as “experiments.”  How is a formal survey any more “experimental” than emailing researchers or polling students?

[…] The safest answer for bioethicists, of course, is simply: “They should get our approval for those activities, too.”  The more territory bioethicists claim for themselves, however, the more you have to wonder, “How good is bioethicists’ moral judgment in the first place?”

To answer this question, let me bring up a bioethical incident thousands of times deadlier than the Tuskegee experiment.  You see, there was a deadly plague called COVID-19.  Researchers quickly came up with promising vaccines.  They could have tested the safety and efficacy of these vaccines in about one month using voluntary paid human experimentation.

[…] In the real world, researchers only did Step 1, then waited about six months to compare naturally-occurring infection rates.  During this period, ignorance of the various vaccines’ efficacy continued, almost no one received any COVID vaccine, and over a million people died.  In the end, researchers discovered that the vaccines were highly effective, so this delay really did cause mass death.

How come no country on Earth tried voluntary paid human experimentation?*  As far as I can tell, the most important factor was the formal and informal opposition of bioethicists.  In particular, bioethicists converged on absurdly (or impossibly) high standards for “truly informed consent” to deliberate infection. Here’s a prime example:

“An important principle in human challenge studies is that subjects must give their informed consent in order to take part. That means they should be provided with all the relevant information about the risk they are considering. But that is impossible for such a new disease.”

Why can’t you bluntly tell would-be subjects, “This is a very new disease, so there could be all sorts of unforeseen complications.  Do you still consent?”  Because the real point of bioethics isn’t to ensure informed consent, but to veto informed consent to whatever gives bioethicists the willies.

[…] I’ve said it before and I’ll say it again: Bioethics is to ethics as astrology is to astronomy.  If bioethicists had previously prevented a hundred Tuskegees from happening, COVID would still have turned the existence of their entire profession into a net negative for humanity.  Verily, we would be better off if their field had never existed.

If you find this hard to believe, remember: What the Tuskegee researchers did was already illegal in 1932.  Instead of creating a pile of new rules enforced by a cult of sanctimonious busybodies, the obvious response was to apply the familiar laws of contract and fiduciary duty.  These rules alone would have sent people like the Tuskegee researchers to jail where they belong.  And they would have left forthright practitioners of voluntary paid human experimentation free to do their vital life-saving work.

In a just world, future generations would hear stories of the monstrous effort to impede COVID-19 vaccine research.  Textbooks and documentaries would icily describe bioethicists’ lame rationalizations for allowing over a million people die.  If the Tuskegee experiments laid the groundwork for modern Human Subjects Review, the COVID non-experiments would lay the groundwork for the abolition of these deadly shackles on medical progress. […]

___________________________________________

General COVID-19 thoughts, from me (someone with no relevant medical background):

1. I’ve head reports of people getting seriously ill or dying from various preventable illnesses, because they’re too scared to go to the hospital for various non-COVID-related ailments. In general, I think people are unduly scared of hospitals: going to the hospital is risky, but not catastrophically so. I’d advise people to stop going into grocery stores (if they can avoid it) long before I advised avoiding hospitals (in cases where they’re worried something might be seriously wrong).

Obviously, now is not the time to go in for routine check-ups, and video calls with doctors are a good first step in most cases, etc.

2. I’ve updated toward thinking it won’t be that hard to avoid catching COVID-19 in March/April in spite of the new strain, if you’re the kind of person who’s in a social network of very cautious people who have ~all avoided catching COVID-19 thus far. A large number of people are taking few or no precautions, and the bulk of COVID-19 exposures has been (and will continue to be) drawn from that group.

If you’re young and your whole social network has almost completely avoided anyone catching COVID-19 thus far, it’s more likely your social network is being over-cautious.

3. A lot of sources have been exaggerating the risk that you’ll be infected, or infect others, even if you’ve previously caught COVID-19 or been vaccinated. I think most people who’ve recovered from COVID-19 should mostly act as though COVID-19 doesn’t exist at all, at least for the next few months (in areas where the Brazil and South Africa strains aren’t widespread yet).

I’d say the same for people who have had two shots of the Pfizer or Moderna vaccine, as long as it’s been ~2 weeks since you had your second shot. For more detailed risk assessments than that, I recommend using the microCOVID website.

4. This is the home stretch. Universal vaccine availability is on the horizon, and our vaccines seem amazingly effective (especially for preventing deaths and hospitalizations). It goes without saying that it’s extra-unfortunate to catch COVID-19 shortly before you would have gotten vaccinated.

___________________________________________

From Scott Alexander’s March link post:

CTRL+F “Blackrock” in this Matt Levine column for a discussion of how we accidentally stumbled into true communism for the good of all. The short version: an investing company called Blackrock owns so much of the economy that it’s in their self-interest to have all companies cooperate for the good of the economy as a whole. While they don’t usually push this too hard, the coronavirus pandemic was a big enough threat that “BlackRock is actually calling drug companies and telling them to cooperate to find a cure without worrying about credit or patents or profits”.

___________________________________________

Also from Scott’s link post:

The class-first left’s case for why the Sanders campaign failed: he tried too hard to reinvent himself as a typical liberal to fit in, but people who wanted typical liberals had better choices, and it lost him his outsider energy (see especially the description of his “astoundingly dysfunctional” South Carolina campaign – “not only did basic tasks go unfulfilled, phone-banking and canvassing data were outright fabricated” – the article claims nobody was able to fix it because it was run by social justice activists who interpreted any criticism of them as racist/sexist. Interested to hear if anyone knows of other perspectives on this). Counterpoint: South Carolina was always going to be hostile territory for him, and maybe he didn’t reinvent himself as a typical liberal enough. I cannot find any other source confirming the South Carolina campaign allegations; interested in hearing what people think.

___________________________________________

Marginal Revolution (also linked in Scott’s roundup) quotes Tanaya Devi and Roland Fryer’s “Policing the Police: The Impact of ‘Pattern-or-Practice’ Investigations on Crime”:

This paper provides the first empirical examination of the impact of federal and state “Pattern-or-Practice” investigations on crime and policing. For investigations that were not preceded by “viral” incidents of deadly force, investigations, on average, led to a statistically significant reduction in homicides and total crime. In stark contrast, all investigations that were preceded by “viral” incidents of deadly force have led to a large and statistically significant increase in homicides and total crime. We estimate that these investigations caused almost 900 excess homicides and almost 34,000 excess felonies. The leading hypothesis for why these investigations increase homicides and total crime is an abrupt change in the quantity of policing activity. In Chicago, the number of police-civilian interactions decreased by almost 90% in the month after the investigation was announced. In Riverside CA, interactions decreased 54%. In St. Louis, self-initiated police activities declined by 46%. Other theories we test such as changes in community trust or the aggressiveness of consent decrees associated with investigations — all contradict the data in important ways.

Neal Zupancic comments:

The authors seem to suggest it is mostly the investigations themselves causing the increase in crime, rather than any particular policy changes. The mechanism they propose is that police officers greatly reduce their quantity of policing when under federal investigation after a “viral” incident, but there is little indication that this comes about as the result of any particular policy reform – the suggestion is that police are either reducing public contact in an effort to avoid having their own actions scrutinized, or are trying to make a point (in the case of deliberate strikes and slowdowns/sickouts). There’s also a section (page 27) where the authors talk about the possible impact of increased paperwork, and estimate it might account for about 20% of the reduction in police activity in one city. I’m not sure if we’re calling this “reform” but even if we do it’s a small proposed effect.

My first attempt:

[epistemic status: playing around with narratives]

___________________________________________________________________

Coeptis – The nietzschean house. Some combination of ‘everyone benefits when we stay out of the way and let the cream rise to the top’, and ‘if you don’t pull yourself up by your boot-straps, well, that’s on you’. Believes in advancing, gaining power, and letting power concentrate in the hands of a few super-competent elites.

Believes in cowboys, superheroes, lone vigilantes. Stubbornly refuses to take orders or conform, where it disagrees with their conscience or taste. Persnicketiness.

Believes the world is intelligible; so hand the world to the best and brightest, and let them figure it out.

___________________________________________________________________

Pluribus – The wisdom-of-crowds house. Believes in the elegance of democratic and (competitive, monopoly-free) market-based solutions, and believes in the nobility of respecting others’ autonomy and agency. Live, and let live. Distributed decision-making, and aesthetic appreciation for the dizzying variety of different individuals’ life-projects.

Pluribus is skeptical of Coeptis’ belief that any one individual can model and optimize the world. The world is too messy for that; it requires diverse and distributed optimization. But Pluribus agrees with Coeptis that individuals’ free action is the special sauce of civilization (albeit en masse, not through an elite).

Let a hundred flowers bloom. Don’t just leave me be; leave people be. See America for what it is, not just what you wish it were. Respect what it is. Respect who we are.

___________________________________________________________________

Novus – The compassionate utopian consequentialist house. Do whatever it takes to protect people and save lives. Break the rules and encroach on apparent ‘rights’ when doing so actually works and improves welfare.

Radicalism; idealism; willingness to work hard for fundamental change. Paternalism. Progress. Deliberately moving toward a brighter future.

Coeptis breaks the rules out of stubbornness and frustration with the idiots who designed things wrong. Novus breaks the rules because people are starving and in need.

Novus is confident that somehow this can be fixed, even if it’s less certain of methodology (more pragmatic, willing to experiment, break eggs, be inelegant) than the other three houses.

___________________________________________________________________

Unum – The house of coordination and tradition.

1. Deferring to authority. Forming tight-knit high-trust alliances. Achieving great things via teamwork and the superpower of acting in lock-step.

2. Preserving and absorbing established scholarship. Learning the lessons of history. Minding Chesterton’s fences and the wisdom of old, evolved systems. Approaching risky new ideas with caution.

1 and 2 are related: strong coordination requires everyone to know with confidence what everyone else in the group believes and wants, which requires relatively stable, uniform, uncertainty-minimizing culture. A marching band can’t be confused about what their orders are, or be perpetually uncertain about whether the left side will suddenly decide to go off and do its own thing.

3. Rule of law, since law is much of what makes society legible and enables cooperation. Applying the law consistently. Resisting corruption. Order.

Brown, Wai, and Chabris – Can You Ever Be Too Smart for Your Own Good? Linear and Nonlinear Effects of Cognitive Ability

Despite a longstanding expert consensus about the importance of cognitive ability for life outcomes, contrary views continue to proliferate in scholarly and popular literature. This divergence of beliefs among researchers, practitioners, and the general public presents an obstacle for evidence-based policy and decision-making in a variety of settings. One commonly held idea is that greater cognitive ability does not matter or is actually harmful beyond a certain point (sometimes stated as either 100 or 120 IQ points). We empirically test these notions using data from four longitudinal, representative cohort studies comprising a total of 48,558 participants in the U.S. and U.K. from 1957 to the present. We find that cognitive ability measured in youth has a positive association with most occupational, educational, health, and social outcomes later in life. Most effects were characterized by a moderate-to-strong linear trend or a practically null effect (mean R^2 = .002 to .256). Although we detected several nonlinear effects, they were small in magnitude (mean incremental R^2 = .001). We found no support for any detrimental effects of cognitive ability and no evidence for a threshold beyond which greater scores cease to be beneficial. Thus, greater cognitive ability is generally advantageous—and virtually never detrimental.

Boxell, Gentzkow, and Shapiro – Cross-Country Trends in Affective Polarization

Politics seems to have gotten more polarized over the last 40 years in the US and to a lesser extent Canada and Switzerland, but polarization has gone down in Germany, Sweden, and Norway, and hasn’t changed much in New Zealand, Australia, or Britain.

Stefan Schubert comments:

As always, one should be careful with making generalisations based on American observations alone. The fact that the trend isn’t universal speaks against American polarisation being due to technological changes, since those changes also have affected the other countries in the study.

Burum, Nowak, and Hoffman – An Evolutionary Explanation for Ineffective Altruism

We donate billions to charities each year, yet much of our giving is ineffective. Why are we motivated to give but not to give effectively? Building off evolutionary game theory models, we argue that donors evolved (biologically or via learning) to be insensitive to efficacy because efficacy is difficult to socially reward, as social rewards tend to depend on well-defined and highly observable behaviors. We present five experiments testing key predictions of this account that are difficult to reconcile with alternative accounts based on cognitive or emotional limitations. Namely, we show that donors are more sensitive to efficacy when helping (i) themselves or (ii) their family. Moreover, (iii) social rewarders don’t condition on efficacy or (iv, v) other difficult-to-observe behaviors, like amount donated.

Schubert on rationality:

Rationality has many aspects. It seems to me that the rationalist community often focuses on the fun bits, such as self-improvement, musings on one’s own thought-processes, and speculative theorising (though no doubt there are important exceptions). What then gets a bit lost is that rationality is to a large extent about discipline, restraint, and rigour: things that aren’t necessarily fun for most people. This is maybe natural given that the community is at least partly built around an intrinsic interest in rationality – they normally don’t provide strong extrinsic incentives (e.g. degrees, money) to students of rationality. Nevertheless, I think a stronger emphasis on these less intrinsically appealing aspects of rationality is important.

Huemer – The Failings of Analytic Philosophy

[…] Analytic philosophers used to think that philosophy was or ought to be a body of analytic knowledge, and that analytic knowledge was essentially about the meanings of words, or the relationships between concepts, or something like that, and did not concern substantive, mind-independent facts. […]

I don’t know how many people still think the job of philosophy is to analyze language/concepts. I don’t think it’s very many. But the field retains leftover influences of that early doctrine. And the central problem with this is that most questions that are amenable to typical analytic-philosophy methods are just not very interesting.

More specifically, I see three things that we’re doing too much of.

1. Fruitless Analysis [- …] Perhaps the value of these analyses is purely for the theoretical understanding of philosophers. But understanding of what — how a specific word is used in a specific language? The exact contours of a conventionally defined category? […]

2. Semantic Debates [- …] Ex.: Reliabilists (the most common kind of externalists) sometimes say that a belief is “justified” as long as the subject formed it in a reliable way, whether or not the subject knows or has reason to believe that the belief-forming method is reliable. Internalists say this is not enough. That looks to me semantic. As an internalist, I don’t deny that reliability exists or is good. I just don’t think that’s what “justification” refers to. […]

3. Defining Down the Issue [-] Okay, here is my biggest complaint. Philosophers will actually decide what questions to ask based on the consideration of to what questions they can apply purely a priori methods, especially conceptual analysis and deductive arguments. This often involves shifting attention away from questions that matter, to questions that are in the vicinity but that in fact do not matter at all. […]

When I worked as a TA in grad school, some of the classes covered the Problem of Evil. […] Here is a possible response: maybe God isn’t all-powerful after all. (Or he could fail to be all-knowing, or maximally good, but the ‘all-powerful’ attribute is the one theists are most likely to give up.)

I saw this discussed in one of these introductory philosophy textbooks that the students were reading. The author (who was defending atheism based on the Problem of Evil) said something like ‘we are merely bored by such replies’ — I guess because it’s not interesting to defend a thesis by redefining it. (Well of course you can defend the existence of ‘God’ in some sense of that word!)

I found this kind of amazing. So if it turns out that there is an extremely powerful, intelligent, and good being who created the physical universe, but the being isn’t capable of all logically possible actions, then that would be completely uninteresting to a philosopher, because … it doesn’t satisfy the definition of a certain word that we stipulated at the start? That sounds to me like caring more about word games than about reality. […]

Huemer – Against History

In my previous two posts, I attacked Continental philosophy and Analytic philosophy, respectively. But some philosophers remain unoffended, so now it’s time for me to attack the third main thing that people do in philosophy departments: the history of philosophy. I don’t understand why we have history of philosophy. I’ve taken several courses in history of philosophy, and listened to many lectures on it over the years, and occasionally I have raised this question, but no one has ever told me why we have this field.

Don’t get me wrong. I understand why we read historical figures, and why we cover them in classes — because the famous philosophers of the past are usually interesting, and they gave canonical formulations of very important views that are often still under discussion today. They also tended to have a breadth of scope and a boldness missing from most contemporary work.

What I don’t understand is why we have history of philosophy as a field of academic research. For those who don’t know, philosophers in the English-speaking world have whole careers devoted to researching a particular period in the history of philosophy (almost always within Western philosophy), and sometimes just a single philosopher.

What are these scholars trying to find out? Are they looking for more writings that have been lost or forgotten? Are they trying to trace the historical roots of particular ideas and how they developed over the ages? Or are they perhaps trying to figure out whether particular theories held by historical figures were true or false?

No, not really. Not any of those things. Scholarship in the history of philosophy is mainly like this: there are certain books that we have had for a long time, by a certain list of canonical major figures in philosophy. You read the books of a particular philosopher. Then you pick a particular passage in one of the books, and you argue with other people about what that passage means. In making your arguments, you cite other things the philosopher said. You also try to claim that your interpretation is “more charitable” than some rival interpretation, because it attributes fewer errors, or less egregious errors, to the great figure.

What you most hope to do is come up with some startlingly new way of interpreting the great philosopher’s words, one that no one thought of before but that turns out to be surprisingly defensible. It’s especially fun to deny that the philosopher said one of the main things that he’s known for saying. For instance, wouldn’t it be great if you could somehow argue that Kant was really a consequentialist?

Hausfather and Peters – Emissions – the ‘business as usual’ story is misleading

Stefan Schubert: “Nature comment says 5 °C warming scenario by the end of the century is looking increasingly unlikely.”

Another paints a dystopian future that is fossil-fuel intensive and excludes any climate mitigation policies, leading to nearly 5 °C of warming by the end of the century. That one is named RCP8.5. […] Happily — and that’s a word we climatologists rarely get to use — the world imagined in RCP8.5 is one that, in our view, becomes increasingly implausible with every passing year. Emission pathways to get to RCP8.5 generally require an unprecedented fivefold increase in coal use by the end of the century, an amount larger than some estimates of recoverable coal reserves. It is thought that global coal use peaked in 2013, and although increases are still possible, many energy forecasts expect it to flatline over the next few decades. Furthermore, the falling cost of clean energy sources is a trend that is unlikely to reverse, even in the absence of new climate policies.

Assessment of current policies suggests that the world is on course for around 3 °C of warming above pre-industrial levels by the end of the century — still a catastrophic outcome, but a long way from 5 °C. We cannot settle for 3 °C; nor should we dismiss progress.

Some researchers argue that RCP8.5 could be more likely than was originally proposed. This is because some important feedback effects — such as the release of greenhouse gases from thawing permafrost — might be much larger than has been estimated by current climate models. These researchers point out that current emissions are in line with such a worst-case scenario. Yet, in our view, reports of emissions over the past decade suggest that they are actually closer to those in the median scenarios. We contend that these critics are looking at the extremes and assuming that all the dice are loaded with the worst outcomes.

[…]

Those who are tasked with taking climate action on the basis of information from model scenarios are increasingly calling for a more risk-based approach to help with adaptation and mitigation. This approach accounts for the relative likelihood of different outcomes. Controversially, it requires researchers to assign probabilities to scenarios. Critics don’t want to do this, because many see it as an arbitrary process. But when specialists refuse to assign probabilities, users often do so themselves. Most do so poorly because they do not have a deep understanding of the assumptions that underpin these scenarios.

International Energy Agency – Defying expectations of a rise, global carbon dioxide emissions flatlined in 2019

Despite widespread expectations of another increase, global energy-related carbon dioxide emissions stopped growing in 2019, according to IEA data released today.

After two years of growth, global emissions were unchanged at 33 gigatonnes in 2019 even as the world economy expanded by 2.9%. This was primarily due to declining emissions from electricity generation in advanced economies, thanks to the expanding role of renewable sources (mainly wind and solar), fuel switching from coal to natural gas, and higher nuclear power generation. Other factors included milder weather in several countries, and slower economic growth in some emerging markets.

“We now need to work hard to make sure that 2019 is remembered as a definitive peak in global emissions, not just another pause in growth,” said Dr Fatih Birol, the IEA’s Executive Director. “We have the energy technologies to do this, and we have to make use of them all. The IEA is building a grand coalition focused on reducing emissions – encompassing governments, companies, investors and everyone with a genuine commitment to tackling our climate challenge.”

A significant decrease in emissions in advanced economies in 2019 offset continued growth elsewhere. The United States recorded the largest emissions decline on a country basis, with a fall of 140 million tonnes, or 2.9%. US emissions are now down by almost 1 gigatonne from their peak in 2000. Emissions in the European Union fell by 160 million tonnes, or 5%, in 2019 driven by reductions in the power sector. Natural gas produced more electricity than coal for the first time ever, meanwhile wind-powered electricity nearly caught up with coal-fired electricity. Japan’s emissions fell by 45 million tonnes, or around 4%, the fastest pace of decline since 2009, as output from recently restarted nuclear reactors increased. Emissions in the rest of the world grew by close to 400 million tonnes in 2019, with almost 80% of the increase coming from countries in Asia where coal-fired power generation continued to rise.

Across advanced economies, emissions from the power sector declined to levels last seen in the late 1980s, when electricity demand was one-third lower than today. Coal-fired power generation in advanced economies declined by nearly 15% as a result of growth in renewables, coal-to-gas switching, a rise in nuclear power and weaker electricity demand.

“This welcome halt in emissions growth is grounds for optimism that we can tackle the climate challenge this decade,” said Dr Birol. “It is evidence that clean energy transitions are underway – and it’s also a signal that we have the opportunity to meaningfully move the needle on emissions through more ambitious policies and investments.”

Yudkowsky – Dunbar’s Function

Old LessWrong post.

Samuel, Piper, and Matthews – 19 big predictions about 2020, from Trump’s reelection to Brexit

Vox staff’s predictions for 2020.

Chalmers gives various arguments for thinking that phenomenal consciousness isn’t reducible to merely physical facts. A complete reductive explanation must make it logically impossible for the reduced entity to differ in any way unless the thing you’re reducing it to also differs in some way. Chalmers argues that reductions of consciousness to physical facts can never be complete in this sense, because there is some aspect of consciousness that could in principle vary without varying any physical fact. This aspect is the first-person, subjective, phenomenal character of consciousness; what it actually feels like “from the inside” to instantiate conscious states.

I accept Chalmers’ arguments, for reasons I detail in an earlier section of the paper but won’t go into here. Rather, I agree with him that phenomenal reductionism is probably false; but Chalmers’ own position, which I call phenomenal fundamentalism (the idea that there are irreducible phenomenal states), also commits us to absurdities.

Eliezer Yudkowsky’s “Zombies! Zombies?” does a good job of articulating the core problem with non-interactionist fundamentalism, though I didn’t really understand this argument’s force at the time. Sean Carroll’s “Telekinesis and Quantum Field Theory” dispenses with interactionist fundamentalism.

By process of elimination, I conclude that phenomenal anti-realism, or eliminativism, is probably true: phenomenal consciousness is neither reducible nor irreducible (in our universe), because it doesn’t exist.

This idea seems absurd, so I endorse it only grudgingly: it’s absurd, but less absurd than the two alternatives.

There are a number of obvious objections to the idea. While I think some of these objections are partly successful, I think on the whole they aren’t successful enough to make eliminativism a worse option than reductionism and fundamentalism. Here, I’ll try to systematically address a large number of possible objections. In the process I’ll hopefully clarify for some people what I mean by “eliminativism.”

Be warned that the following is not my standard fare. It’s very much written for an audience of professional analytic philosophers, and is pretty relentless about pursuing fine distinctions and subtle counter-arguments. I think this is warranted by the fact that eliminativism is such a strange view. Philosophers to date have reasonably complained that anti-fundamentalists like Dennett and Yudkowsky have been needlessly sloppy and imprecise. My view is that the arguments of anti-fundamentalists have exhibited less rigor than those of fundamentalists for contingent historical reasons, and this shouldn’t be taken to indicate that the underlying idea is fragile and liable to collapse under close scrutiny.

_________________________________________________________

I have suggested that there’s a good initial case to be made that physicalism is true. I’ve also said, contra ‘introspective’ and ‘skeptical’ phenomenal anti-realism, that eliminativism isn’t obviously true. But I’ve argued as well that reductionism is untenable. Given the difficulties phenomenal states raise, we should take seriously the hypothesis that there are no such states, even though there initially seem to be. I call this position ‘theoretical’ phenomenal anti-realism.

I will defend my own version of anti-realism, which grants a large number of the fundamentalist’s intuitions—that our introspection seems to give phenomenal realism some immediate plausibility, that phenomenal properties are invertible, etc.—but denies that any properties that could satisfy all these intuitions are instantiated.

In addition to modeling its environment, the brain models its own internal states; but systematic error is possible in the latter case just as it is in the former. Hypothesizing that a persistent error in internal representation is responsible for intuitions like “color qualia outstrip their functional roles in behavior and cognition” helps explain why fundamentalism could seem so plausible in a purely physical world.

This treatment of qualia as intentional objects is similar in several respects to Richard Hall’s treatment of phenomenal properties as “dummy properties”:

[I]t (the experience) doesn’t instantiate the phenomenal properties but rather represents – represents_int – the perceived object as instantiating them. That is, the phenomenal properties only occur in the experience as content, as properties the object of perception is represented as having, not as properties instantiated in the experience itself. (Hall 2007, 205)

Hall’s account differs from my own mainly in that he treats qualia as properties of objects in one’s environment, whereas I treat them as properties of one’s experiences, as is more common in hard-problematic debates. Hall’s main argument is in fact compatible with fundamentalism, since his primary concern is with whether environmental objects have irreducible qualia—an easy idea to reject (Hall 2007, 205-206). He refutes certain representationalist arguments against physicalism, but Chalmersians are free to grant that our environment is devoid of the (Edenic) properties Hall thinks we represent it as having, while insisting that our experiences too seem to bear such properties—not just as intentional content, but as occurrent constituents (Chalmers 2006/2010a).

If both our environments and our minds seem to violate physicalism, it is consistent to say that the latter experience is veridical while the former is not. Hall’s view may also lead to broader epistemic worries than mine, since it plagues our ordinary perceptions of the physical world, not just our representations of our own and others’ minds (cf. O12 below).

Interestingly, fundamentalists like Chalmers can largely accept my account, as a theory of how our zombie duplicates come to express belief in fundamentalism. But Chalmers will insist that I have failed to explain our own epistemic situation. I will seek to further unpack and defend my view in the course of responding to this objection and a variety of others, which I’ll number “O1” through “O20”.

O1: Seeming. You grant that prima facie we seem phenomenally conscious, while insisting that this is an illusion. But phenomenal properties’ being just is their seeming. If Mary seems to see a table, she may be mistaken; there is a metaphysical and inferential gap between her mental image of a table and the worldly referent. But if she seems to experience phenomenal redness, this semblance just is phenomenal redness. There is no gap between a mental appearing of phenomenal character and the character itself, because the appearances are the phenomenal character. For qualia, there is no appearance-reality gap. So your view, ‘We don’t undergo phenomenal states, although we seem to undergo them’, is internally inconsistent.

My response is that this objection borrows a lot of its plausibility from an ambiguity in terms like ‘seeming’ and ‘appearance.’ Consider, by analogy, an optical illusion such as Shepard’s Tables (Shepard 1990):

When first confronted with this illusion, a reasonable observer is likely to be disposed by the overall evidence to believe—with good reason—that the horizontal table is shorter than the vertical one on the page. The horizontal table thus seems or appears shorter. However, an observer might then learn all about the illusion and its mechanisms, enough to correct for it in her beliefs in many other circumstances. It no longer seems to such an observer, in the aforementioned sense, that the horizontal table is shorter—the available evidence disposes the observer to think that the tables are equally long. Yet there is a sense in which the horizontal table still seems or appears longer to this observer, provided her vision works like most people’s. The tables still look as though they have different lengths, even after one has adjusted one’s beliefs so as to no longer misinterpret the visual impression.

We might say that the former kind of seeming is doxastic, while the latter is perceptual. Thus it might make sense to suppose that an oar partly submerged in water continues to perceptually seem bent—our visual experience still represents it that way—even after it ceases to doxastically seem bent, i.e., after the balance of evidence ceases to indicate that it is in fact bent (cf. Augustine c. 417/2002, 190-191).

I am happy to grant that there is a coherent notion of seeming we might call phenomenal seeming—a perceptual or quasiperceptual appearing analogous in some ways to sensory or somatic appearing, but characteristic of phenomenal consciousness qua phenomenal. And I am happy to grant that there is no gap between phenomenal seeming and phenomenal being; it seems to be part of our concept of phenomenal states that undergoing such states has two inextricable aspects, epistemological and ontological. That is, phenomenal properties involve a primitive, pre-doxastic (perception-like) seeming as well as a being, and these two always co-occur. But, this being the case, my view is simply that phenomenal seeming, like phenomenal being, is an illusion. Nothing in reality actually phenomenally seems any way to any subject.

Hence it is unproblematic that such seeming would necessarily go hand-in-hand with the actual instantiation of the indicated phenomenal properties. Instead, the theoretical eliminativist will insist, at a minimum, that we sometimes epistemically seem to instantiate phenomenal properties, while denying that epistemic seeming of this sort entails phenomenal seeming.

I feel especially confident in denying that if we doxastically seem to instantiate some phenomenal property, we therefore instantiate it. In conversation, Kirk Ludwig noted a clear example of this: A subject is told that she’ll be touched from behind with a hot poker, then is touched instead with an ice cube (cf. Smithies 2012, 265). It can epistemically seem to such a person, misled by his expectations, as though she has a burning sensation, i.e., qualia characteristic of being burned. This is possible even if she’s really instantiating cold qualia and her perception straightforwardly (though not particularly distinctly) represents it as cold.

O2: (Quasi)perception. Epistemic seeming isn’t enough. Our prima facie evidence for our own phenomenal consciousness doesn’t stem from a description, as in the case of inferentially mistaking a cold sensation for a hot one; it stems from something perception-like, some sort of immediate acquaintance. Eliminativism can account for the subjects’ tendencies to form certain beliefs about their consciousness. But the perceptually or quasiperceptually apparent presence of phenomenal states is a further datum, a further explanandum, not exhausted in any attendant judgments, beliefs, or verbal reports (Chalmers 1996, 186-189). (We can imagine, for instance, neurosurgically altering a subject to consistently form false judgments about her experiences’ phenomenal character or perceptual content, without thereby altering her predoxastic
perceptions to correspond to these spontaneous judgments.) Since eliminativism ignores
this datum, it cannot provide an intellectually satisfying response to the hard problem.

I think this is a very good argument against some forms of phenomenal anti-realism. Dennett, for instance, at times endorses something like perceptual anti-realism, asserting that nothing is introspectively evident except belief-like states. Writes Dennett:

Since I hold that we have privileged access only to judgments, and since I cannot make sense of any claim to the effect that something to which I do not have privileged access is an element of my immediate conscious experience, I am left defending the view that such judgments exhaust our immediate consciousness, that our individual streams of consciousness consist of nothing but such propositional episodes, or better: that such streams of consciousness, composed exclusively of such propositional episodes, are the reality that inspires the variety of misdescriptions that pass for theories of consciousness, both homegrown and academic. […]

You may be wondering if you even have judgments. Typically these episodes are the momentary, wordless thinkings or convictions (sometimes misleadingly called conscious or episodic beliefs) that are often supposed to be the executive bridges leading to our public, worded introspective reports from our perusal or enjoyment of the phenomenological manifold our reports are about. My view, put bluntly, is that there is no phenomenological manifold in any such relation to our reports. There are the public reports we issue, and then there are episodes of our propositional awareness, our judgments, and then there is—so far as introspection is concerned—darkness. What lies beyond or on the interior of our judgments of the moment, what grounds or causes or controls them, is wholly a matter for science or speculation—in any event it is not a matter to which we have any privileged access at all. (Dennett 1979, 95)

The infallibilist line on qualia treats them as properties of one’s experience one cannot in principle misdiscover, and this is a mysterious doctrine (at least as mysterious as papal infallibility) unless we shift the emphasis a little and treat qualia as logical constructs out of subjects’ qualia-judgments: a subject’s experience has the quale F if and only if the subject judges his experience to have quale F. We can then treat such judgings as constitutive acts, in effect, bringing the qualia into existence by the same sort of license as novelists have to determine the hair color of their characters by fiat. We do not ask how Dostoevski knows that Raskolnikov’s hair is light brown. (Dennett 1988, 55)

You seem to think there’s a difference between thinking (judging, deciding, being of the heartfelt opinion that) something seems pink to you and something really seeming pink to you. But there is no difference.
There is no such phenomenon as really seeming – over and above the phenomenon of judging in one way or another that something is the case. (Dennett 1991, 364)

Although this view is not incoherent, I see no reason to endorse it. Phenomenologically, our inner lives seem to divide (though perhaps not cleanly or crisply) into propositionally organized states (beliefs, desires, concerns, suspicions…) and non-propositional ones (pains, pressures, emotions, sounds, smells…) (Chalmers 1996, 189-191). This gives us prima facie reason to accept such a division, just as prima facie we have introspective warrant for phenomenal realism — but without a hard problem to problematize the former division. Accordingly, I think that some of our experiences are not structured like beliefs in all respects, though they may share some important features in common with beliefs. (For instance, perceptual states may represent states of affairs, albeit more in the manner of paintings than of sentences.)

My form of anti-realism allows that our doxastic prima facie warrant for phenomenal realism may be grounded in a pre-doxastic, perception-like apparent phenomenality. Phenomenal anti-realism does not entail perceptual anti-realism. No hard problem is raised if a computer processes information sometimes in a linguiform fashion, sometimes in a pictiform (or tactiform, or…) fashion. The same holds for a brain. We both doxastically (prima facie) seem and (quasi)perceptually seem to bear phenomenal properties, though we do not phenomenally seem to bear such properties.

To defend the distinction between phenomenal and perceptual seeming, I can note that zombies, to the extent that they have beliefs or perceptions at all, intuitively seem as though they would be susceptible to both doxastic and perceptual misrepresentation (e.g., they can succumb to optical illusions), though ex hypothesi they cannot phenomenally represent or misrepresent anything.

This is of course only an analogy. Visual perception is significantly unlike our quasiperceptual access to our own mental states. However, once it is granted that both doxastic and perceptual error is possible in the absence of the phenomenal, I see no principled way to rule out the possibility that zombies are capable of quasiperceptual metarepresentations as well, including erroneous ones. And such metarepresentations, I believe, do hold some promise of explaining the apparent presence of phenomenal properties in our stream of consciousness, as opposed to ignoring or flatly denying this pre-doxastic semblance.

O3: Phenomenality as explanandum. The eliminativist is still simply denying the data, not taking it seriously and attempting to explain it. Granting that we sometimes doxastically and (quasi)perceptually seem to be phenomenally conscious is not enough, so long as ‘doxastic seeming’ and ‘(quasi)perceptual seeming’ are given so thin a reading that even a zombie could attain this cognitive and epistemic standing. The intuition that we are phenomenally conscious is the intuition that we differ from zombies, and the project of responding to the hard problem is precisely the project of explaining why we aren’t zombies. Any characterization of the data that remains neutral regarding whether we are the same as zombies has not even begun to address the task at hand, which is to explain phenomenal consciousness itself, not to explain the blind cognitive processes correlated with phenomenal consciousness (Chalmers 1996, 107, 110-114; Chalmers 2002/2010, 112-113).

Here I concede that the eliminativist denies some of (what seem to be) the data. That much is unavoidable. Eliminativism calls for a Copernican revolution in our view of ourselves. This requires us to question our past givens just as Copernicus made us question as obvious a datum as ‘the Earth remains still while the Sun moves across the sky’. If we replace my formulation of the hard problem with ‘the hard problem given phenomenal realism’, then it trivially follows that the anti-realist, as anti-realist, cannot respond to this problem. But as long as it is possible that we have misconstrued the data itself, we cannot rule out all hypotheses that imply such a misconstrual.

Theorists who are already fully convinced of their views have no reason to carry on a dialogue (cf. O7-8). It is only the more moderate realists and anti-realists, who are open to being convinced, to whom discussions like these are of interest. These discussants much allow, at least for argument’s sake, a theory-neutral interpretation of the data. Dismissing anti-realism because one’s first impulse is to treat phenomenal properties as givens amounts to a refusal to seriously engage with anti-realism, just as dismissing Berkeleian idealism because one initially counted mind-independent rocks and horses as undisputed data would amount to a refusal to seriously engage with idealism. In the case of phenomenal anti-realism, serious evaluation is warranted jointly by the arguments for physicalism and against reductionism.

O4: Non-inference. The phenomenal is known by some form of direct quasiperceptual acquaintance, not by inference (cf. Chalmers 2002/2010, 114). Hence this knowledge cannot be wrong.

This objection is refuted by non-inferential misrepresentations, e.g., optical illusions. To revive this objection, one must either deny that perceptions can misrepresent anything (O6), or deny that phenomenal knowledge-by-acquaintance is representational at all (O5).

O5: Non-representation. Our acquaintance with the phenomenal is not a perception-like representation, because it isn’t a representation. We are in some sort of metaphysically and epistemically intimate contact with the phenomenal, as opposed to being in contact with some sort of depiction of the phenomenal. Representational states can be illusory, but bare acquaintance cannot.

I think this is a reasonable initial objection. My first-person experience in general does initially seem to me to be simply and immediately ‘present’, as opposed to being a model, accurate or inaccurate, of lower-level patterns of neural activity. But I ultimately reject this idea on theoretical grounds. Our access-conscious states do seem to serve a self-modeling functional role, and treating the illusion of phenomenal consciousness as an artifact of this process provides a simple means for me to articulate and defend physicalism. If such a move is decisively off-limits, I gather that it must be for reasons involving O7.

O6: The possibility of misperception. Perhaps we have some perception-like representations of our own phenomenal properties. But there isn’t any such thing as ‘perceptual misrepresentation’. Only beliefs can misrepresent the world. For instance, there is no nonveridical element in our perception of Shepard’s Tables independent of our tendency to form false beliefs therefrom. (Perhaps we only think that there is some lingering misrepresentation because the false beliefs the tables inspire remain so vividly seductive even to those in the know.) Since our beliefs must then be the locus of any errors and illusions on our parts, O2 and O4 regain their full force as rebuttals of phenomenal anti-realism.

I disagree that non-doxastic states are always veridical (or simply non-representational). Such a view clashes with how we usually think about perceptual experience. Perhaps perceptual states are only veridical or nonveridical in virtue of the judgments they tend to produce or justify. For instance, perhaps the reason the pseudophenomenality-generating quasiperceptual self-monitoring mechanism I’ve posited counts as yielding misrepresentations is that it tends ordinarily to produce and/or ground relevantly false judgments. Or perhaps perception has its own, independent representational content (Pryor 2004, 355-357). I don’t have a settled view on which account of misperception is right, but as long as the distinction is granted at all, my view remains importantly distinct from the doxastic accounts vulnerable to O2.

O7: Self-evidence. Contra O4, the non-inferential character of our immediate phenomenal knowledge doesn’t establish its certainty. But, nevertheless, our immediate phenomenal knowledge is maximally self-verifying. It isn’t possible that we’ve misconstrued the data so radically that we’re deluded about phenomenal consciousness altogether, because the phenomenal character of experience is introspectively manifest, in the strong sense of being beyond any doubt. Our acquaintance with it may in some fashion depend on a perception-like representation, but there is simply no way for us to misperceive it. No theory, no matter how appealing, can be compelling enough to override the sheer obviousness of our own consciousness.

What does this perfect manifest accuracy consist in? If its presence means that I cannot be doxastically mistaken about my phenomenal states, then it is surely absent, as shown by the hot poker example. Moreover, some people accept phenomenal realism, others anti realism. That in itself shows that some people are deeply mistaken about their own phenomenology.

Perhaps the phenomenal inerrantist is appealing to a stronger version of O5. Or perhaps she means that my (quasi)perceptual representations of my phenomenal states are inerrant, even if none of my subsequent judgments about my phenomenology are inerrant. In O1-2, the realist’s point was that one cannot give up phenomenal being without giving up phenomenal seeming too. Here, the claim is much stronger: One simply cannot reasonably give up phenomenal seeming. The phenomenal’s manifest presence rationally obliges us to assent to realism. Of course, Chalmers’ paradox of phenomenal judgment shows that a zombie would reach effectively the same conclusion in the absence of the phenomenal. But that doesn’t change the fact that we just do have an immediate acquaintance with our phenomenal consciousness (Chalmers 1996, 172-189). The fallibility of judgment can do nothing to undermine that, provided we do not reduce phenomenal quasiperception itself to a kind of judgment.

Like most of the objections here, I think this one has intuitive force, even for physicalists. As such, I want to do as little violence to the underlying intuitions as possible. Just as I granted that we quasiperceptually seem to have phenomenal properties—and quasiperceptually seem to phenomenally seem to have them!—I will also grant that these properties quasiperceptually (and doxastically, prima facie) seem manifest, self-evident, pretheoretically obvious, infallibly self-revealing. But just as I denied that we really do phenomenally seem to have phenomenal properties, I also deny that the phenomenal really is manifest. I grant a watered-down semblance, and seek to explain it. But I do not grant as much as the realist wishes.

The independent plausibility of eliminativism (via physicalism and anti-reductionism) is not the only reason to doubt this sort of self-evidence, however. Inerrant self-knowledge is an extraordinarily strong claim, and does not seem to hold for any of our other perceptual or self-monitoring faculties. Humans in general have turned out to be much more error-prone than introspection and intuition originally indicated, in a wide variety of cases. For instance, people’s introspective judgments about the contents and dynamics of their memory and reasoning processes (Nisbett and Wilson 1977), as well as their perception and imagination (Schwitzgebel 2011; Dennett 1991, 344-362), are not consistently reliable. I remain skeptical that naturalistic fundamentalists can devise a plausible metaphysical mechanism for an epistemic status of this sort. Human consciousness is a product of many complex working parts, an evolved system; it would be astounding if such a system, even if accompanied by irreducible mental properties, could achieve a perfectly error-proof reckoning of itself, much less one where the error-proofness itself is introspectively manifest in an error-proof way.

Even if phenomenal properties are manifest in our experience, I doubt that the fact of quasiperceptual manifestness is itself quasiperceptually manifest. As I type this, I perhaps seem to be inerrantly acquainted with my own phenomenal states. But I don’t seem to inerrantly know by acquaintance that inerrancy itself; I don’t even notice this fact, if fact it is, until I have reflected about my experience and formed a preliminary theory of epistemic access and representation. So, on reflection, it is possible that the initial quasiperceptual indicators of my immediate acquaintance with the phenomenal are delusive, even if we grant that such a delusion initially doxastically or quasiperceptually seems impossible.

O8: Self-evidence as realist sine qua non. Inductive arguments for eliminativism are
question-begging. For O7’s doctrine of phenomenal self-evidence is indispensable to realism. If phenomenal properties occur at all, then they manifestly and with certainty occur. In doubting that phenomenal properties are perfectly self-evident, the eliminativism-entertainer will thereby have rejected the very existence of phenomenal properties. Inductive arguments will then be superfluous. All the work is done at the stage of deciding whether or not anti-realism is even an epistemically open possibility, one for which any subsequent evidence or argumentation could conceivably matter.

1. If someone is phenomenally conscious, then it is not epistemically possible (for that person) that he or she is not phenomenally conscious.

2. The eliminativist must assume, at the outset, that inductive arguments are relevant to the issue of phenomenal realism, i.e., that they can weaken one’s initial conviction in realism.

3. Therefore the eliminativist must assume, at the outset, that it is epistemically possible (for us) that we are not phenomenally conscious.

4. Therefore the eliminativist must assume, at the outset, that we are not phenomenally conscious.

If we have strong prior reason to accept realism, this then shows that anti-realism can’t be rationally justified. And even if we lack such reason, this argument shows that realist and antirealist are stuck in a stalemate. There is no way to non-question-beggingly argue for either view.

My primary response is that the truth of phenomenal realism does not entail the manifest epistemic impossibility of phenomenal anti-realism. Some phenomenal realists might claim inerrantism, but realism also admits of epistemically moderate views. Even Chalmers tends to speak of phenomenal anti-realism only as unmotivated and preposterous, rather than as outright impossible. Responding to type-A views such as eliminativism, he writes,

The obvious problem with type-A materialism is that it appears to deny the manifest. [… I]t seems to be a further truth that we are conscious, and this phenomenon seems to pose a further explanandum, one that raises the interesting problems of consciousness. To flatly deny the further truth, or to deny without argument that there is a hard problem of consciousness over and above the easy problems, would be to make a highly counterintuitive claim that begs the important questions. This is not to say that highly counterintuitive claims are always false, but they need to be supported by extremely strong arguments. So the crucial question is whether there are any compelling arguments for the claim that, on reflection, explaining the functions explains everything. (Chalmers 2002/2010, 112)

Chalmers does sometimes argue as though eliminativism is not an epistemically open possibility (e.g., in O10). Yet he seems to welcome new eliminativist arguments, and to entertain at least the abstract possibility that one could be successful!

I reject [eliminativism] as being in conflict with the manifest facts. Perhaps an extraordinary argument could establish that conscious experience does not exist, but I have never seen an argument that comes remotely close to making this case. (Chalmers 1996, 164)

If realism is supposed to be immune to any degree of reasonable doubt, then this openness is hard to understand. One explanation, suggested by Ludwig, is that while Chalmersian realists may suspect that they know realism with certainty, they don’t claim to know with certainty that they know realism with certainty. Some degree of confidence or warrant is lost in the shift to meta-knowledge. Alternatively, perhaps certitude is lost in the shift from immediate phenomenal knowledge-by-acquaintance to explicit phenomenal judgments.

I conclude that it is reasonable to insist on phenomenal fallibilism, at least for ordinary introspecters. Perhaps some wondrous philosophical or attentional feat could make us infallible introspecters; perhaps some discovery could truly, definitively, and non-question-beggingly rule out our zombiehood. But anyone willing to take anti-realism at all seriously should insist that this is not our present epistemic situation. This does not beg the question in anti-realism’s favor, because not all realisms appeal to phenomenal certitude, and even those that do may restrict this certitude to idealized introspecters, first-order knowledge, or some non-doxastic bare insight.

O9: Introspective evidence. The only way to seriously entertain eliminativism as a hypothesis is to deny that we are immediately acquainted with phenomenal properties, since granting that we are immediately acquainted with such properties would immediately rule eliminativism out. But if she denies this acquaintance, then the eliminativist will have already ruled out the possibility of recovering any evidence for phenomenal realism. As Chalmers notes, the only reason to think we aren’t zombies is that we do seem to be immediately acquainted with phenomenal properties (Chalmers 1996, 102). O8 insisted that doubting this self-evident acquaintance is ipso facto to rule out phenomenal realism. Here we instead make the weaker claim that excluding the phenomenal’s apparent self-evidence as evidence for phenomenal realism unfairly excludes any possible way, according to the realists themselves, of arriving at supporting evidence for phenomenal realism.

My response is that I accept that apparent manifestness is good evidence for phenomenal realism. I think things really do initially epistemically seem that way; and this is why I grant phenomenal realism the pre-theoretic advantage, even though I think that in the endgame, when we weigh the merits of the best realist and anti-realist theories, the anti-realist’s view will win out. Although my inductive arguments do require us to grant at the outset that it is an open epistemic possibility that we are zombies, they do not require us to completely exclude introspective appeals as relevant evidence. We can then weigh how compelling those appeals are against the virtues and shortcomings of our best physicalist theories.

O10: Zombic skepticism. We can’t rationally entertain radical error in our phenomenal
(quasi)perception as a skeptical hypothesis. Skeptical hypotheses by their very nature are about holding constant how things appear, while varying how they (otherwise) are. Thus we can entertain the hypothesis that we’re a brain in a vat because the existence of our experiences is logically compatible with two transexperiential causal stories: authentic sensation, and complex simulation. But this clearly does not hold for eliminativism. It is not consistent to hypothesize the world continuing to appear as it does, while we are not conscious of it.

I think that this argument can be reduced to O1-3. Skeptical hypotheses hold constant how things seem; but if they can legitimately rely on holding constant how things doxastically, perceptually, or quasiperceptually seem, without always assuming that there is a certain way things phenomenally seem for us to hold constant as well, then eliminativism will be a coherent hypothesis in the same sense as the others. (For instance, one could speak of a brain-in-a-vat scenario in which we hold constant the nonphenomenal component of our judgments and perceptual states, ignoring whether the phenomenal component is also preserved.)

If this is not viable, it will only be because of the strength of O7 and O8: the eliminativist hypothesis isn’t epistemically possible because the manifest experiential presence of the phenomenal is overwhelmingly, irresistibly obvious.

O11: All knowledge requires consciousness. Our access to any information about the world must be causally mediated by phenomenal consciousness. Alternatively: Our warrant for believing anything depends on phenomenal consciousness. The idea here is, perhaps, that the phenomenal dimension of mind is precisely what I have access to. I cannot know about my unconscious mind, except inasmuch as I can draw reasonable inferences about it via the conscious mind. The world beyond my experience is a mystery, except inasmuch as it in some fashion affects my experience. But it then follows that if we were completely unconscious, we would be completely oblivious of everything. Our unconscious mind might skillfully model and respond to its environment, but I would be totally in the dark regarding that fact—indeed, in the dark regarding everything.

I think this objection borrows a lot of its attraction from blurring the lines between phenomenal consciousness and access consciousness. A fundamentalist who makes use of Absent Qualia arguments must grant that the phenomenal is causally unnecessary for our cognitive functioning. Inasmuch as the conscious/unconscious divide is of great significance for global self-monitoring brain states that a zombie might possess, and not just for phenomenal consciousness, the eliminativist may appeal to it to justify the zombie’s beliefs. Since there will in general be a precise correspondence between the reliability of the zombie’s and angel’s cognitive faculties, and between the zombie’s and angel’s fulfillment of doxastic duties (though cf. O20), it will be very difficult to find points of asymmetry between the zombie and the angel that do not simply presuppose that knowledge is impossible without phenomenal consciousness.

(An “angel” is an otherworldly being that resemble us in all respects, except that it’s phenomenally conscious.)

If there are deeper reasons to think that zombies are epistemic voids, then I suspect that they stem from a general sense that a zombie double would lack my mind (O13 and O17), or a general difficulty imagining the (seemingly) phenomenal as a wellspring, not of epistemic insight, but of epistemic impoverishment or confusion (O18).

Smithies writes that perceptual experience plays a central role in epistemic justification

in virtue of its phenomenal character. It is because perceptual experience has the phenomenal character of confronting one with objects and properties in the world around me that it justifies beliefs about those objects and properties. (Smithies forthcoming, 8-9)

Smithies’ goal is to explicate our epistemic concepts in light of their relation to phenomenal realism, not to defend phenomenal realism by appealing to its ties to epistemology, so it may be a stretch to treat his arguments as arguments for realism. (Cf. Searle’s argument in O14.) But regardless, I do not find his arguments convincing.

Smithies suggests that access consciousness is not necessary for epistemic justification, because the contents of our perceptual fields can justify beliefs on our part even if we aren’t actively attending to them; and that it is not sufficient, because a case of “super-blindsight”, where one spontaneously formed detailed and accurate beliefs about one’s visual environment in a blind spot, would constitute access-consciousness but not a source of (immediate, non-inferential) justification (Smithies forthcoming, 13-14). But this only shows that a more fine-grained functional notion of epistemic access is needed here, one that not only demands that information be “poised to be used” in reasoning and deliberative action (Block 1995) but that privileges certain ways to be so poised.

In particular, we can privilege functionally normal (i.e., non-blindsighted) perceptually poised-for-use information. In this way, an epistemic asymmetry between perceptual and non-perceptual access to information can be functionally grounded, as can an epistemic continuum from information that’s both perceived and attended to, to information that’s perceived but not attended to, to information that isn’t perceived at all.

In general, it would be extremely surprising if any demonstration could be made that functional states are categorically non-justifying, because this would imply (quite
uncharitably!) that zombies lack any warrant for any of their beliefs.

O12: Most knowledge requires consciousness. Zombies (or at least zombie eliminativists) lack warrant for most of their beliefs, because if the phenomenal illusion is so all-pervading, then succumbing to (or, perhaps, discovering) it will dramatically undermine our justification for a variety of other beliefs. For instance, I believe that tomatoes refract light in the 620–740-nm range in large part because I believe that my mental images of tomatoes are phenomenally red. Thus anything that undermines the latter would partly undermine the former; and if a sufficiently large number of my phenomenal beliefs were undermined at once, it would have a cascading effect on my nonphenomenal beliefs. Since anti-realism is being justified largely on inductive, scientific grounds, this threatens to make anti-realism self-undermining.

In response, I deny that phenomenal properties are particularly important to our everyday beliefs, particularly about the physical world. Although it is true that my beliefs regarding the objective color of tomatoes depend in part on my beliefs regarding the subjective color of my visual experiences of tomatoes, the relevant aspects of this subjective experience are its functional relationships to other experiences and to the world, not its phenomenal character.

What matters most for our overall world-view is not that tomatoes produce this experience—as I attempt to ostend a red quale—but that they produce the same experience as that produced by stop signs, the opposite of that produced by a clear afternoon sky on an HSV color wheel, etc. This can be shown by considering someone born a color invert. Intuitively, the invert would have perfectly ordinary, veridical beliefs about tomatoes, erring only in her beliefs about others’ phenomenal states. If Inverted Qualia doesn’t endanger our general beliefs about the external world, about our functional mental states, etc., then I see no reason why Absent Qualia should vitiate the zombie’s nonphenomenal beliefs.

O13: Mind requires consciousness. The ordinary assertions of eliminativists, like ‘I believe in eliminativism’ or ‘I’ve seen that article’, depend for their truth on the existence of phenomenal states like belief and vision. Since eliminativists deny phenomenal properties, their routine assertions—perhaps even their assertions of eliminativism itself—commit them to a performative contradiction (cf. Wallace 2007, 14-15).

Whether states like belief seem on introspection to have a phenomenal character remains actively debated. Whether they have an essentially phenomenal character is even more questionable. However, even in cases like ‘sight’, the phenomenal anti-realist can simply stipulate that she has in mind a functional state in the vicinity of the relevant phenomenal one. Thus when I speak of perception or memory in ways that take these faculties for granted, I have in mind the functional states phenomenal realists and anti-realists both agree exist, semantically devoid of any essential appeal to the phenomenal. Whether as it happens such faculties metaphysically depend on phenomenal consciousness will then be left open to substantive investigation and debate.

Thus I establish the theory-neutrality of much of my mentalist talk by stipulation. I’m perfectly happy to grant that this practice may be idiosyncratic—that is, the ordinary English-language senses of these words may indispensably presuppose phenomenal consciousness. I adopt it anyway, for pragmatic reasons.

O14: Intentionality requires consciousness. Intentional states in general depend on phenomenal consciousness. Thus, in denying phenomenal consciousness, eliminativists are rationally committed to the meaninglessness of their own utterances.

There is no good reason to think that intentionality depends on phenomenal consciousness. Chalmers notes that Searle argues for the dependence of intentionality upon consciousness in The Rediscovery of the Mind (Chalmers 1996, 371). Searle argues as follows:

1. Intentionality occurs and is not reducible to the physical.
2. Intentionality is correlated with consciousness.
3. Consciousness occurs and is not reducible to the physical.
4. Therefore the simplest explanation of (1) is that intentionality’s physical irreducibility derives from that of consciousness.
5. Therefore intentionality depends on consciousness.

But there are good reasons to doubt these premises. On 1: It seems much easier to argue for the reducibility of intentionality than to make the corresponding case regarding phenomenal consciousness.

On 2: It’s true that we generally speak as though intentionality depended on
consciousness. But if we presuppose that intentionality is a nonfunctional, nonphysical
phenomenon, then we seem thereby to lose any empirical grounds for retaining this assumption. After all, how would the world look any different if rocks had intentional objects? Panintentionalism is easily motivated by premise 1, for the same reason panpsychism is easily motivated by phenomenal fundamentalism.

It’s true that if you already accept phenomenal fundamentalism, you thereby gain some inductive support for the fundamentality of other difficult-to-reduce phenomena, including intentionality. But the dependence of intentionality upon consciousness seems to derive its plausibility here from a mixture of background phenomenal fundamentalism and a Moorean rejection of views like panintentionalism. In contexts where we cannot presuppose phenomenal fundamentalism, we lose much of our reason to suspect that intentionality is likewise physically irreducible and ineliminable. In particular, nearly all of the arguments for the inexplicability of the intentional in physical or functional terms also suggest the inexplicability of the intentional in phenomenal terms.

O15: Charity. People simply can’t be guilty of such radical, pervasive error. A guiding principle of semantic theories is that they should make speakers’ statements come out mostly true. But if anti-realism is true, then speakers are pervasively wrong about their mental lives. Any semantic theory of what terms like ‘phenomenal’ and ‘experience’ mean that has such radical implications is more likely to be false than is the body of assertions in question. Inasmuch as we wish to reject perverse semantic theories, we must calculate tradeoffs between these theories’ simplicity, their predictive force, and how many statements of the language come out true. If we exclude the last criterion, then we are likely to end up with otherwise simple and explanatory-seeming theories that assign absurd meanings to ordinary-language terms.

In response, I note that there are many cases in which the best explanation for human behavior, including human language use, is that people are consistently wrong about a topic. Historically, most people thought the Earth flat and immobile, viewed macroscopic objects as mostly solid, and assumed an objective, reference-frame-independent temporal simultaneity. Reinterpreting those language users’ words for ‘Earth’ or ‘now’ or ‘solid’ to make their statements literally true is not merely misguided; it is entirely beside the point. The strict truth of statements is often irrelevant to most plausible hypotheses for why statements mean what they do. What matters is some relevant respect in which they are true. For instance, the mereological nihilist can explain why the word ‘table’ means what it does because atoms arranged tablewise share nearly every important feature in common with metaphysically real wholes. Likewise, as I noted in response to O12, which phenomenal properties we instantiate—indeed, to some extent that we instantiate them at all—is irrelevant to how and why people use their language, including even their mentalist language.

This is not to deny that phenomenal anti-realism has far-reaching implications for the truth of our statements and theories. It is only to suggest that strict ‘truth’ is too simple and one-dimensional a concept to capture an important constraint on semantic theories.

O16: Phenomenal concepts. You claim that we quasiperceptually meta-represent our first-order mental states, nonveridically presenting the first-order states as possessing phenomenal properties. Phenomenal property instances or states are intentional objects (like leprechauns or infinitesimals) that do not actually occur. But to represent such intentional objects will require grounding the apparently phenomenal in a physical substrate. Replacing actual phenomenality with mere depictions thereof accomplishes nothing, because these representations will either fail to explain why the phenomenal illusion seems exactly the way it does, or will be physically inexplicable themselves, raising a new explanatory gap. Ludwig frames the argument like so:

1. Your brand of anti-realism claims that we quasiperceptually represent phenomenal redness, and that no property satisfies this representation.
2. Representing phenomenal redness requires some phenomenal concept.
3. To possess a phenomenal concept, one must have actually instantiated the phenomenal property in question.
4. So your brand of anti-realism entails a contradiction.

An analogous argument can be given regarding our doxastic representations of phenomenal properties, one that generalizes to all eliminativists.

This is a powerful objection, because it appeals to the same intuition underlying Secret Qualia. If we cannot understand a bat’s phenomenology, then a fortiori zombies cannot either. If Mary lacks the phenomenal concept Ω^QR, then a fortiori Mary’s zombie twin does too.

The physicalist who gives in to this argument might adopt a quietist or non-cognitivist approach, granting that she has suddenly grasped the meaninglessness of ‘phenomenal property’ (and therefore of ‘phenomenal anti-realism’) but biting the bullet in a Wittgensteinian fashion (Wittgenstein 1922/1999, 108). However, I think this drastic step can be avoided.

On my theory, Mary does not acquire new knowledge when she first observes red objects firsthand. She acquires a new concept—she can represent (quasi)perceptually, imaginatively, and verbally, in a way that she could not before—but this concept is empty.

My view can accommodate the notion that she seems to encounter new, irreducible, ineffable properties. Similarly, it can accommodate the notion that these properties, upon reflection, seem of the sort that a physical system could not represent them. Intuitively, redness is too specific for the purely physical and functional even to falsely portray. Reapplying the Inverted Qualia intuition, it seems as though our pseudoquale
could have (falsely) seemed to have a different phenomenal character, and the rest of my account remain just as adequate in every detail.

But, while I grant that this seems intuitively so, once again I must deny that this semblance is veridical. A basic feature of pseudophenomenality, as I understand it, is that the properties themselves seem to have a qualitative richness that outstrips any functional specification. ‘Functional specification’ not only includes human theories, but also patterns of neural firing. For the same reason that no possible description seems up to the task of fully expressing phenomenal redness, no possible computer seems up to the task of picking it out in a model or simulation, whether veridically or unveridically.

My anti-realism treats both intuitions as illusory. If we are in the grip of an illusion that that which is quasiperceptually represented is present and irreducible, then we are also in the grip of an illusion that that which is quasiperceptually representing is present and irreducible. I deny the representatum’s presence, whereas I deny the representational content’s reducibility.

It may seem that I should either eliminate both or eliminate neither. But eliminating the representational content would overcomplicate any representationalist response to the hard problem, and might even lead to a regress. The representational content itself does not seem present in our experience; rather, it is a theoretical posit of mine. Since it is a novel posit, and since it does not seem in its own right to have a (quasi)perceptually vivid ‘aura of irreducibility’, I feel freer to specify my concept of it in such a way that I ensure its physical reducibility.

The opposite is the case for the phenomenal. I take ‘phenomenal property’ to have an established meaning along the lines of ‘those properties whose instances I ostend in my field of awareness that are (or persistently appear to be) ineffable, intrinsic, self-revealing, qualitative, etc.’. Since I think that there are no such property instances, but I think that there’s some aspect of experience that naturally leads us to think that there are (and that we can ostend them), I take the ostensive acts to be successfully ‘directed toward’ such instances (or at the ‘spots’ where they would be). They are unsuccessful only inasmuch as the phenomenal property instances aren’t actually there.

I could have redefined ‘phenomenal’ to refer to some state that I do think occurs, but then I would have had no word for the kind of property philosophers like Nagel, Chalmers, Searle, Levine, and Jackson falsely think they instantiate. So too, I would have had no word for the illusory aspect of our second-order experience that so powerfully motivates the dualist’s intuitions.

The higher-order representational content itself, on the other hand, is at best known only indirectly, and inasmuch as it too seems on reflection to be irreducible, it is only indirectly (as a side-effect of the irreducibility of the phenomenal itself) and inferentially.

Suppose that the fundamentalist digs her heels in and insists that it is a conceptual truth,
for her concept of ‘phenomenal property’, that bearing the concept requires one to have
instantiated the property. My response would be that that’s not a concept anyone actually has, given that the property in question is irreducible and nonphysical. Insofar as I believe in physicalism, I must deny that we have such concepts. We can certainly talk about such concepts, but only in the abstract, since we cannot possess them.

These quasiphenomenal properties should be distinguished from the kinds of phenomenal properties we zombies seem quite able to discuss (though we do not, on my view, instantiate either). To help keep them straight, while I call the phenomenal properties I’ve been discussing so far Q-properties, I’ll call hypothetical phenomenal properties that are inconceivable to zombies X-properties.

Ex hypothesi, as far as my view is concerned, we have no concept of Xhood, though perhaps we have a concept of having a concept of Xhood. If even that is closed to us, then we might at least employ a Ramseyfied version of the property, speaking of properties that are like phenomenal properties but with the added constraint of being unintelligible to non-angels. The realist will then attempt to assert ‘∀Q∃X(Q=X)’, while I will insist that the attempt is a failure and will negate structurally similar Ramsey sentences.

At this point, having clarified to what extent such a dispute is even expressible from our respective perspectives, a stalemate will seem to have been reached, and the question of whether phenomenal properties are also X-properties will likely need to be bracketed so that other points for and against realism can be contested.

(Footnote: Previously, I noted that phenomenal seeming is the sort of epistemic status that one can only be in if one actually instantiates the phenomenal properties one thereby seems to have. By definition, denying phenomenal seeming requires one to deny the occurrence of phenomenal states, and vice versa; what we instead possess is a pseudophenomenal seeming, a quasiperceptual, introspective nonveridical awareness of having phenomenal properties—and a corresponding nonveridical awareness of phenomenally seeming to have them.

One interesting way to further cash out phenomenal seeming, and to further incorporate the fundamentalist’s intuitions into the theory, is to say that phenomenal seeming, if we instantiated it, would involve quasiperceptually seeming to instantiate X-properties. The pseudophenomenal seeming we actually undergo, on the other hand, involves quasiperceptually seeming to instantiate Q-properties. The pseudo-presence of Q-properties would then be built into the illusion of phenomenal being, while the pseudo-representedness of X-properties would be built into the illusion of phenomenal seeming. This isn’t necessary for distinguishing the two kinds of seeming, but it might be an added theoretical benefit of the distinction, if the X-state intuition seems sufficiently strong and basic.)

O17: First-person zombiehood. It is widely accepted that we can conceive of someone’s being a zombie, in the third person. But one cannot conceive of oneself’s being a zombie, in the first person. In O7 and O10, it was claimed that my zombiehood is not a live epistemic possibility. Here, it is instead claimed that my zombiehood is in some sense inconceivable. I cannot even entertain the idea. I can imagine someone who looks and acts like me being a zombie, but I cannot imagine actually being one, for the same reason I cannot imagine being a rock: by assumption, there is nothing it’s like to be a zombie.

Nagel’s ‘what it’s like’ locution is standardly taken to pick out phenomenal properties. But in that case the anti-realist will simply respond that there’s nothing it’s ‘like’ to be anything. We have no actual experience at all with ‘what it’s like’ even to be ourselves, in that sense.

Possibly this sounds excessively strange; possibly ‘what it’s like’ shouldn’t be treated as theory-laden or illusion-prone to the same extent as ‘phenomenal property’. Perhaps, for instance, we should reinterpret ‘what it’s like’ to pick out our core epistemic situation, or our perception-like states, or everything to which we have cognitive or epistemic access. In the latter senses (on my view) functional duplicates cannot vary in ‘what it’s like to be them’—whereas phenomenal what-it’s-like-ness can vary independently of function, but we have only an illusory acquaintance with what it’s phenomenally like to be anything. We should say that this is ‘what it’s like’ to be a zombie, to the extent that ‘what it’s like’ is not a fundamentally mistaken idea.

Once we extricate phenomenality from our more metaphysically unassuming self-concepts, we may find that we have much more immediately important subjective content and structure in common with zombies than we had initially thought. Even fundamentalists should expect most things about the subjectivity of some organism that are likely to be put into words to be functionally specifiable—the perspectival nature of experience, its privacy, its ownership, its perceived self/other divisions, its cognitive and epistemic immediacy, its internal structure, its nonphenomenal representational content, its transparency, its rich network of causal and functional-similarity relationships to other states, and many (if not all) of its behaviors concerning the phenomenal (Chalmers 1996, Tye 1995). For the same reason that phenomenality seemed so difficult to reduce and so irrelevant to our world’s causal order, phenomenal eliminativism can retain the rest of our folk-psychological notions essentially intact. This may help alleviate the sense of alienness we ordinarily attribute to the zombie’s existence.

O18: Veridical quasiperception. In O17, the concern was that I can’t imagine what it
would like to be a zombie. Here, the concern is that I can’t imagine what it would like to be a physical near-duplicate of myself that lacks any illusion of phenomenality. If the apparent phenomenality of experience is nonveridical, then, Ludwig argues, it should be possible to represent e.g. visual scenes veridically, i.e., nonphenomenally. It should be possible to strip away the delusive ‘paint’ and perceive things as they really are, clearly and correctly. But when we imagine methodically subtracting the phenomenal from experience, we don’t seem to be left at the end of this process with an accurate experience. Rather, we seem to be left with no experience at all.

It is not surprising that we can’t imagine what it would be like for our imagination itself to be free of a persistent illusion. The more interesting concern here seems to be with hashing out, at least in the abstract if not in fleshed-out imaginative exercises, how a being like us would differ if her phenomenology lacked any indicators of the phenomenal.

Presumably she would still have introspective access to her internal states, but if asked whether these states are so rich in subjectively evident content that they appear partly ineffable and irreducible, the veridical introspecter either wouldn’t understand the question, wouldn’t have any idea as to the answer, or would know that the answer is ‘No’.

(If the second possibility seems incredible, note that there are innumerable natural questions about our phenomenology that introspection alone has a difficult or impossible time answering (Schwitzgebel 2011). Our experience does not consist of cohesive fields of instantiated properties, all specified in richly interlocking detail and ripe for immediate introspective verification (Dennett 1991, 344-356; Noë 2002, 3-8). If our visual representations can underdetermine what is in our blind spot, then I see no reason why a being’s metarepresentations could not underdetermine whether its own mental life consists of functional-seeming states. Our actual experience, however, is not with such underdetermination—phenomenal properties readily reveal themselves upon reflection to be irreducible plena, and this appearance remains consistent across intuitions and thought experiments.)

One assumption that seems to be lurking in the back of this objection, and of a number of others, is that our experience or cognition is wholly phenomenal. But Chalmers’ project of differentiating functional psychological concepts from phenomenal ones, I think, largely dispenses with that mistake (Chalmers 1996, 11-12). The only times when it seems as though our psychology is reducible to our phenomenal state is when we’ve already assumed at the outset that phenomenality is a necessary criterion for our cognition, perception, etc.

The objector may also be assuming that since the phenomenal is dispensable to our mental life (because it isn’t there!), the pseudophenomenal illusion is equally dispensable to our mental life. But the illusion itself is a real brain process. It plays a major causal role at the very least in our philosophizing, and possibly in more common behavioral and perceptual events as well. We should not be surprised if its functioning turns out in practice to be inextricably linked to other neural processes, even if in principle we could construct a cognizer (probably with a substantially different high-level neural architecture) that lacked this illusion.

O19: Evolution. Succumbing to a pervasive quasiperceptual illusion seems maladaptive.
It is presumably computationally expensive to go to the trouble of simulating innumerable
properties that are simply not present; so illusionless subpopulations ought to have outcompeted the deluded long ago. How could so complex and pervasive an illusion ever evolve?

Coming from a fundamentalist, this objection is very vulnerable to a tu quoque response;
the problem of explaining how we could evolve to have phenomenal consciousness of the sort we seem to is, if anything, even more difficult. Still, this is an important concern, and I don’t have a satisfying answer yet. Of particular relevance is Chalmers’ discussion of why we should expect complex self-modeling reasoning computers to think of the variations between their perceptual states as “brute facts”.

When we ask how it knows that it sees the red bicycle, an efficiently designed system would say, “I just see it!” If such a system were reflective, it might start wondering about how it is that things look red, and about why it is that red just is a particular way, and blue another. (Chalmers 1996, 185)

This is compelling, but I do not think it tells the whole story. We could build a machine that matches the above description and that in effect has super-blindsight. Such a being might very well come to wonder about the nature and origin of its brute visual-environment intuitions, and might very well come to false conclusions as a result, but it wouldn’t do so in the manner that we do.

Chalmers specifies that the errors of the reflective machine he has in mind arise from its perceptual states, dismissing the super-blindseer’s cognition as “curiously indirect”. But we can also imagine a normally sighted phenomenally unconscious automaton that can’t help spontaneously forming judgments about the irreducibility of perception, but whose perception on immediate introspection does not warrant such judgments. (At most, it merely causes them.)

On my view, the key missing element in both of these cases is that the illusion of
phenomenality is perception-like. The ‘zombic hunch’ is not a hunch. It is not a free-floating intuitive judgment, not even an irresistibly compelling one. I seem rather to have been immediately acquainted with a phenomenal manifold from an early age, long before I started reflecting on or theorizing about my consciousness. I seem to have concrete epistemic and metaphysical access to my mental life, in a way analogous to—but stronger than—my epistemic and metaphysical access to physical objects in my perceptual environment. And this seeming itself is grounded (and perhaps consists) in my quasiperceptual modeling of my mental states as phenomenal states. It is unclear to me that such a sophisticated illusion falls out of Chalmers’ self-modeling automaton so naturally.

(The above is perhaps why fundamentalists find fundamentalism intellectually satisfying. Their hypothesis seems to amount to an explanation of their observations. In acquiescing to realism, it feels as though one is acknowledging a natural phenomenon of some sort; it does not feel as though one is merely giving in to a vague ‘hunch’.)

We do not know how computationally intensive or structurally complex the phenomenal illusion’s mechanism is. As far as we know, pseudophenomenality is more efficient than the alternative, if not from a top-down engineer’s perspective then at least from within the constraint of building a reasoner via incremental cumulative random mutations and selection.

Our quasiperceptual awareness of our own experiential states in many ways mirrors our perceptual awareness of our environment; possibly the former system is modeled after the latter, and the illusion of the phenomenal is a vestige of some mechanism that was adaptive in the case of perception but is not adaptive in its higher-order incarnation. I find it extremely likely that much or all of the phenomenal illusion stems from random genetic drift or is a spandrel.  Still, some explanation is indeed owed for the origin of the raw material of which the illusion is a vestige or side-effect.

Adaptive explanations of all or part of the illusion’s underpinnings are difficult to make, because many of the properties we intuitively associate with phenomenality also have functional counterparts. For instance, our experience could have appeared vivid, manifest, virtual, and private without appearing so content-rich as to outstrip its functional roles.

Still, in the interest of stimulating further research, I’ll suggest a very tentative hypothesis: Marking our experience in introspection as having a manifestly functionless dimension encourages humans to overextend their theory of mind in a strictly nonveridical but evolutionarily advantageous way.

Evolutionary theorists have hypothesized that we possess a Hyperactive Agent Detection Device, driving us to err on the side of detecting predators and other agentive threats very readily (Barrett 2000). Perhaps our experiences’ apparent phenomenality helped further exaggerate this device’s effect, in a fashion that had further selective advantages. Perhaps, for instance, our ancestors had an easier time forming fruitful empirical hypotheses about apparent agents than about mindless phenomena. If protoscientific hypotheses like ‘find shelter when the sky spirits become angry’ lent themselves to our progenitors’ reasoning and motivational drives more readily than did their nonmentalist counterparts, then a mechanism for encouraging such hypotheses may have been adaptive. A readiness to anthropomorphize external phenomena may also have been advantageous for persuading or cooperating with other agents, e.g., in religious contexts. Hence — still indulging here in wild historical speculation — it might have been useful to suffer from the illusion that one’s mental states abound with fully autonomous property instantiations, if the result was that one’s theory of mind could more readily generalize to entities that are, structurally and behaviorally, thoroughly inhuman. The conceivability of zombies and inverts, the epistemic gap, etc. would then be side-effects of a more general perception-like mechanism imbuing our mental states with flexibility as to their nonmental correlates.

This is all assuming, perhaps implausibly, that a simple and intuitive functional explanation is likely to hold here. By directing attention to this particular problem, my hope is that readers may devise more sophisticated accounts that likewise avoid conflating the different explananda.

Non-interactionist fundamentalists can join me in theorizing about possible explanations along these lines, since they will agree that a completely functional explanation can be given of the appearance of phenomenality inasmuch as this differentially affects our cognition and behavior. The only disagreement will, once again, concern whether explaining this appearance explains everything.

O20: Ethics. A final objection is to the moral implications of my view. It is natural to suppose that organisms are moral patients in large part because they are conscious (Harris 2010, 32-33; cf. Joyce 2007/2009, §5). In evaluating whether it is wrong to torture a robot, what matters is not whether it superficially behaves as though it is in pain, but whether it actually experiences pain. By eliminating phenomenal consciousness, your view leads to the unacceptable consequence that we must also eliminate morality.

One might argue theoretically here, from the truth of moral realism to the truth of phenomenal realism. Or one may argue at the practical level, claiming that our moral conduct would be hindered by a widespread belief in phenomenal anti-realism. In the former case, I see no incoherence in a phenomenally anti-realist moral realism. I concede that the idea is counterintuitive, but this is a drop in the bucket relative to phenomenal anti-realism’s internal bizarrery! Insofar as we do have reason to believe in moral facts that depend on phenomenal facts, I think this reason largely derives from our prior reasons to believe that we are phenomenally conscious, and not the other way around.

I am more sympathetic to the practical concern, and I’d like to spend more time thinking about the social and psychological ramifications of eliminativism someday. It can be bracketed for the moment, as the potential harmfulness of eliminativism is independent of its truth, and this philosophy paper is probably dry and dense enough to bear little risk of Internet virality.

The importance we ascribe to the phenomenal in our own lives, and our difficulty in disentangling phenomenal and nonphenomenal mental concepts, explains why we appeal to the phenomenal-nonphenomenal distinction to provide a clear and intuitive line dividing the proper objects of our concern from Cartesian automata. But if we did not have a prior reason to associate ourselves—and as a consequence our prudential and moral concerns—with a phenomenal dimension, then I do not think that anything basic to morality itself would have forced this association upon us. The notions of fairness, obligation, and harm make no explicit appeal to phenomenal consciousness, and to the extent that the two are implicitly associated it is not obvious that they could not come to be extricated once phenomenal realism itself came to be doubted.

I take the task of extricating moral and prudential norms from the phenomenal aspect of affect and volition to parallel the task of extricating epistemic norms from the phenomenal aspect of belief and perception. In both cases I believe that the original concepts are either inclusive enough to encompass zombies, or malleable enough to allow us to unproblematically expand their scope if the anti-realist thesis becomes sufficiently credible. It is not as though phenomenal anti-realism, whether morally realist or not, has the consequence that one stops preferring some things over others. As long as we in fact have values and concerns and interests, we will continue to have norms and value systems in practice.

Some of the above objections to eliminativism, I believe, are reasonable enough to give fundamentalists grounds to be wary about abandoning their foundational assumptions. But I think that my responses, though cursory, show that phenomenal eliminativism deserves much more consideration than it has received to date—particularly by those who do not consider O7 (self-evidence) conclusive.

I suspect that a number of my responses will not feel very satisfying to those who found the objections initially plausible. In a number cases I do little more than say, ‘Your premise, if true, would refute my view; but I can consistently affirm that your premise only seems true.’ Such arguments by their very nature can’t get eliminativism off the hook. They neither dispense with the semblance nor explain it—not in robust cognitive detail, at any rate.

My goal in such cases is not to completely defeat every objection to eliminativism; it is only to show that they are not fatal to eliminativism, and that the sorts of claims a theoretical eliminativist must assent to are interesting and novel enough to warrant much more careful investigation in the literature.

I have at least given moderate fundamentalists reason to take phenomenal anti-realism seriously, and not to dismiss it as manifestly absurd. For my main audience, committed physicalists, combining this account with my earlier argument against phenomenal reductionism indicates that eliminativist views are at present their best hope for fending off fundamentalism.

Fully explicating a plausible anti-realist account of human cognition is a large project, and I do not think that the arguments for physicalism are so strong, Chalmersian epistemology so bizarre, or the internal problems for eliminativism so minor that it is manifestly unreasonable to reject eliminativism; but I do think that the balance of evidence currently favors it.

On my anti-realist account, the intuitions that make the hard problem so hard can be explained quite straightforwardly.

Zombies and inverts seem possible because they are possible—but this poses no problem for physicalism if the angelic properties that zombies lack and that inverts interchange are not instantiated in our own world (Beisecker 2010).

Bats’ minds seem ineffable to us because we mistakenly think that they, like we, are phenomenal angels; once this mistake is corrected in our own case, we lose any reason to impute unphysical properties to other minds.

Mary’s understanding of color seems incomplete to us for the same reason—either Mary is a zombie and we have misunderstood her epistemic situation for the same reason we misunderstand our own, or Mary is an angel and the failure of physicalism in her world does not endanger physicalism in our own.

In all of these cases, the intuitive force, epistemic basicness, and experiential pervasiveness of this anti-physicalist evidence is best explained by a representational defect in a perception-like self-modeling faculty.

Fundamentalists, having more or less succeeded in vanquishing reductionism, should move on to the more difficult and foundational task of accounting and arguing for the special epistemic status of the phenomenal, so as to defend realism itself. At the same time, physicalists who appreciate the severity of the hard problem should shift their focus from defending reductionism to constructing eliminative theories.

I have only begun to spell out one such theory above, but it’s already forced me to reconsider my position on a number of foundational disputes in a variety of philosophical fields. Until more of phenomenal anti-realism’s consequences for philosophy as a whole have been drawn out, the overall merits of the theory will be difficult to estimate. Still, it’s already easy to see that the most radical break with philosophical tradition that is called for by the anti-realist is a rejection of the phenomenological tradition, following Descartes, of taking assorted experiential appearances as self-evident and self-verifying. Although the anti-realist is free to admit phenomenology as a legitimate, and frequently reliable, source of evidence, it will be very difficult ever to recapture a gem of true certitude from experience, once a phenomenon as epistemically basic as phenomenality itself has relinquished this status.

I’d expect those who find this step an easy one to take to be those who already felt disposed to grant that the reach of moderate skepticism is, if not deep, at least exceptionally wide.

Bibliography

• Augustine of Hippo (c. 417/2002). On the Trinity. Cambridge University Press.
• Barrett, Justin (2000). “Exploring the Natural Foundations of Religion”. Trends on Cognitive Science 4:29-34.
• Beisecker, David (2010). “Zombies, Phenomenal Concepts, and the Paradox of Phenomenal Judgment”. Journal of Consciousness Studies 17(3-4):28-46.
• Block, Ned (1995). “On a confusion about the function of consciousness”. Behavioral and Brain Sciences 18:227-47.
• Chalmers, David J. (1996). The Conscious Mind: In Search of a Fundamental Theory. Oxford University Press.
• ————— (2002/2010). “Consciousness and its Place in Nature”. In D. Chalmers (ed.), Philosophy of Mind: Classical and Contemporary Readings (pp. ??-??). Oxford University Press. Reprinted in D. Chalmers, The Character of Consciousness (pp. 103- 39). Oxford University Press, Inc.
• ————— (2006/2010b). “Phenomenal Concepts and the Explanatory Gap”. In T. Alter & S. Walter (eds.), Phenomenal Concepts and Phenomenal Knowledge: New Essays on Consciousness and Physicalism (pp. 167-194). Oxford University Press. Reprinted in D. Chalmers, The Character of Consciousness (pp. 305-336). Oxford University Press.
• Dennett, Daniel C. (1979). “On the absence of phenomenology”. In D.F. Gustafson & B.L. Tapscott (eds.), Body, Mind, and Method. Kluwer.
• ————— (1988). “Quining Qualia”. In A.J. Marcel & E. Bislach (eds.), Consciousness in Contemporary Science (pp. 42-77). Oxford University Press.
• ————— (1991). Consciousness Explained. Boston: Little Brown.
• ————— (2001). “The zombic hunch: Extinction of an intuition?” In A. O’Hear (ed.), Philosophy at the New Millennium. Cambridge University Press.
• Hall, Richard J. (2007). “Phenomenal Properties as Dummy Properties”. Philosophical Studies 135:199-223.
• Harris, Sam (2010). The Moral Landscape: How Science Can Determine Human Values. Free Press.
• Jackson, Frank (1986). “What Mary Didn’t Know”. The Journal of Philosophy 83(5):291-295.
• Joyce, Richard (2007/2009). “Moral Anti-Realism”. In E.N. Zalda (ed.), The Stanford Encyclopedia of Philosophy (Summer 2009 Edition).
• Levine, Joseph (1983). “Materialism and Qualia: The Explanatory Gap”. Pacific Philosophical Quarterly 64:354-61.
• Nagel, Thomas (1974). “What Is It Like to Be a Bat?” The Philosophical Review 83(4):435-450.
• Nida-Rümelin, Martine (1996). “What Mary Couldn’t Know: Belief about Phenomenal States”. In T. Metzinger (ed.), Conscious Experience (pp. 219-241). Allen Press.
• ————— (2002/2010). “Qualia: The Knowledge Argument”. In E.N. Zalta (ed.), The Stanford Encyclopedia of Philosophy (Summer 2010 Edition).
• Nisbett, Richard E. & Wilson, Timothy DeCamp (1977). “Telling More Than We Can Know: Verbal Reports on Mental Processes”. Psychological Review 84(3):231-259.
• Noë, Alva (2002). “Is the Visual World a Grand Illusion?” Journal of Consciousness Studies 9(5-6):1-12.
• Pryor, James (2004). “What’s Wrong with Moore’s Argument?” Philosophical Issues 14(1):349- 378.
• Schwitzgebel, Eric (2011). Perplexities of Consciousness. MIT Press.
• Shepard, Roger N. (1990). Mind Sights: Original Visual Illusions, Ambiguities, and Other Anomalies. WH Freeman and Company.
• Smithies, Declan (2012). “A Simple Theory of Introspection”. In D. Smithies & D. Stoljar (eds.), Introspection and Consciousness (pp. 259-294). Oxford University Press.
• ————— (forthcoming). “The Phenomenal Basis of Epistemic Justification”.
• Tye, Michael (1996). Ten Problems of Consciousness: A Representational Theory of the Phenomenal Mind. MIT Press.
• Wallace, Meg (2007). “Mental Fictionalism”. Unpublished draft.
• Wittgenstein, Ludwig (1922/1999). Tractatus Logico-Philosophicus. Dover Publications, Inc.

My co-worker Eliezer Yudkowsky recently made the case on social media that Trump’s candidacy raises national security concerns qualitatively more serious than the sort you see in a normal US presidential race:

Every election, the Chicken Littles of both parties make a big deal out of how this year’s election opponent is the Worst Ever and Literally Hitler, and take every single thing their opponents do and try to make it sound as terrible as possible, and so on.

Okay, but here’s the thing. What does not happen every election cycle — and this happened months ago, not in the wake of the current bandwagon — is the entire Republican national security establishment going holy shit and repudiating the Republican candidate en masse.

Maybe I shouldn’t take it for granted that everyone already looks at the world and sees (a) a level of politics that’s theater and (b) a level of politics that’s deadly, deadly serious.

Maybe you heard that Trump said maybe we shouldn’t defend NATO countries if Russia invades. And you interpreted that as Trump expressing fed-up-ness with American military spending and our trying to defend everything in the world without getting much in return. Somebody in the newspapers seemed to be making a big deal out of it, just like they make a big deal out of Clinton emails. Clutching at their pearl necklaces and fainting about how terribly important it is that America honor its commitments to other countries, or something.

No.

The people in the national security bureaucracy — hell, even me, even though I’m not a national security bureaucrat and have only read a handful of military history books — heard that and thought:

“Holy shit.”

Members of the Washington, DC establishment who privately laugh about sexual assault and insider trading, heard that and thought: “You don’t do that.”

And I want to try to spell out why this was so terrible. But I’m not sure it will do any good for me to try to talk about that, until I can slice off and distinguish that discussion from the “that was oh such a terrible bad idea” of the pundits that you’ve already filtered out by now.

Which is a difference that might be difficult to convey. Because the news media and the pundits and certainly all the politicians pretend that, why, of course it’s all deadly serious. It’s hard to convey without me sounding like I myself think that taking bribes and grabbing genitals are not “deadly serious”. And to be clear, it’s not that all discussion of national security is on this separate serious level, because there’s lots of pearl-clutching and professional-wrestling about national security too, in the media.

In a previous thread on my Facebook wall, about the existence of expertise, Brent Dill observed that from the perspective of somebody sufficiently ignorant, maybe there doesn’t seem to be any higher expertise in the world. “Then who builds the spaceships, dammit?” I asked, and Brent Dill replied, “They just saw a TV special on how the moon landings were faked.” From your perspective and my perspective, there are these sorts of entertaining TV shows about the moon landing being faked, and then above that is Real Science, where some things are pretty darned solid despite all the frothy arguments that go on about the replication crisis.

Okay, but what if all you see is the arguments over whether the moon landing was real, and as far as you know, that’s all there is to see? You understand that the TV shows are entertainment, but you don’t realize that there’s a non-entertainment part. Or maybe you think that “did aliens build the pyramids” is real serious science and that it doesn’t get any more serious than that. The Discovery Channel isn’t going to tell you about it. If you’re sufficiently immersed in that world, maybe it’s the only world there is.

Maybe I’m assuming too much when I assume that everybody knows that politics is theater. Maybe the reason the political theater works is that people honestly don’t realize it’s all professional wrestling. That hypothesis doesn’t seem to quite fit with people’s behavior, but maybe this is one of those things I’m not likely to understand that well?

It does occur to me, though, that I might be presuming too much in supposing that other people realize that there’s a Level B in politics as well. Maybe if you grow up with the modern media, it’s easy to think that the Level A is all that exists and there is no deadly serious politics, that people clutching at their pearls and fainting is as serious as anything ever gets.

At this point the analogy to science breaks down, because in science, the Level B above the Discovery Channel is a virtuous place where you find the real pursuit of truth. The Level B in politics is not in the same way the repository of true concern for truly important things. But the Level B in Washington, DC, the issues that people take seriously unlike insider trading, is also not just sociopaths reacting to disasters that are so bad that their own personal hometown might get a nuclear missile. The Level B does contain more stuff than that. The Level B is also not the upper ranks of the Illuminati where they discuss how to keep power and worry about things so bad that they might affect their personal stock prices, because there aren’t any Illuminati and Washington, DC doesn’t work like that either. It’s not the level at which people are just trying to do their jobs, because nobody in Washington, DC is just trying to do their jobs.

But it is the level where you worry about things like the stability of the Europe-Russia border, not because a journalist is going to clutch their pearls in offense because you don’t seem concerned enough, but because you actually care about the stability of the Europe-Russia border. Yes, there are people in Washington, DC like that.

I said this in a comment elsewhere on Facebook, but I’m going to repeat it here, in case there’s people on my Facebook wall who haven’t seen it before:

My reading of history books is admittedly biased by having read about historically interesting cases. This does tend to be cases where things went very right, or more usually, very wrong. American revolution, French revolution, World War I, World War II.

Perhaps there are dozens of other cases where a country elected an impulsive, chaotic, populist leader and nothing whatsoever went wrong.

But when I think of Trump, I think of Hitler, and not in the generic sense of “Hitler” meaning “bad”. I think of the British diplomats who sent Hitler a sternly worded note on the eve of Hitler’s invasion of Poland, warning that Britain would defend Poland even though they hadn’t defended Czechoslovakia. According to The Rise and Fall of the Third Reich, Hitler read the note himself instead of having his diplomatic corps explain it to him, and interpreted the standard diplomatic politesse as conciliatory and a go-ahead to invade Poland.

I once played a four-hour live simulation/game called the National Security Decision-Making Game, which was run by various people who were ex-whatevers. There were around 80 of us simulating just 3 different countries, with myself trying to play the Secretary of Defense of the US.

Thinking myself probably above-average intelligence for the room, I’d originally asked for a position that involved intrigue; I was given the title for Director of National Intelligence. But somebody who’d played the game before said he really wanted to be DNI, so I traded it for his Secretary of Defense position. Which I’m glad happened, because my ambitions rapidly went from world optimization to “Understand what is happening immediately around the Department of Defense.”

By the end of NSDM, I left with a suddenly increased respect for any administration that gets to the end of 4 years without nuclear weapons being used. We did not do that well in our NSDM session. I left with a greatly increased appreciation of the real skill and competence possessed by the high-level bureaucrats like the Secretary of Defense who keep everything from toppling over, and who understand what the sternly worded diplomatic notes mean.

I think that a lot of the real function of government is to keep things from toppling over like they did in our NSDM session, and that this depends on the functionaries including the President staying inside certain bounds of behavior — people who understand how the game is supposed to be played. It’s not always a good game and you may be tempted to call for blowing it up rather than letting it continue as usual. Avoid this temptation. Randomly blowing it up will not end well. It can be so, so much worse than it already is.

The system isn’t as stable as it might look when you’re just strolling along your non-melted streets year after year, without any missiles ever falling on your own hometown. I don’t even know how much work it really takes to prevent everything from falling over.

If I were to try summarize very briefly why Trump’s remarks on NATO crossed a holy shit line, it’d be along the lines of: “If you read the history books, you realize that it is really really bad to have any ambiguity about which minor powers the major powers will defend; that is how World War I and World War II both started.”

And: “In the wake of the second World War that started from that kind of ambiguity, the senior leaders in both the East and the West, enemies though they may have been, decided to learn the lesson and henceforth be more clear about which countries they’d defend. Not only did Trump blow through that, he did so in a way that indicates he has no idea of how World War I started and why this is one of the things you absolutely don’t do. He doesn’t listen to advisors. He doesn’t have advisors! God knows what other guardrails he’s going to blow through!”

Trump didn’t realize he was blowing through one of the deadly serious guardrails. And Trump is not actually stupid, he does not actually have an IQ below 100, he took economics at Wharton. So it’s fine, it’s okay, it does not make you a bad person, if you also don’t know why that was so much more terrible than everything else the media is making a fuss about. Not every citizen of America needs to read The Guns of August and The Rise and Fall of the Third Reich, like I did, or study a vastly larger amount of real military history, like the Republican national security bureaucracy did.

However. If you want to be in the national government, you are supposed to know that this is one of those places where you talk to one of the senior bureaucrats before deciding that it is okay to mouth off about NATO commitments, or before deciding to go ahead and invade Poland.

Even people in Washington, DC who haven’t read any history books understand that part, because most people in Washington, DC do know that politics has a deadly serious level as well as a media theater level.

What Trump said wasn’t a gaffe, it was not one of those things that you’d have to be an idiot to say in front of journalists, it was a world-threatening misstep in the real-life version of the National Security Decision-Making Game.

And now Xi Jinping is thinking about the part where Donald Trump said “Why do we have all these nukes if we can’t use them?” and wondering whether China can take for granted America’s possession of nuclear weapons given that America’s electoral system seems to allow for a certain kind of President. That has already happened and cannot be undone. Even though Donald Trump doesn’t seem to give a fuck about the NSDM, the NSDM gives a fuck about him.

Like it or not, there is in Washington, DC a perceived difference between “committed sexual assault” and “violated the system guardrails that prevent World War III”. Some people in Washington, DC think sexual assault is a big joke, and other people honestly believe it is quite bad and would be just as swift to fire any abuser whether or not the journalists knew about it. But both of those kinds of people understand that the current culture in Washington, DC dictates a difference.

And I’m glad that cultural rule exists, because the Level A culture where everybody clutches their pearls and every gaffe indicates a life-threatening level of incompetence and everything is oh so terribly serious, is not a culture where policy-making can take place. The fact that there are quiet backroom talks with no journalists present, in which at least some people are actually concerned about the Europe-Russia border, is why the Earth hasn’t already blown up.

Again, it’s not that all discussion of national security takes place on that level, there’s lots of theater about that too, the entire Transportation Security Administration is well-known to be pure theater. But there’s also cases where somebody blows through the real actual guardrails, which is when the senior bureaucrats in your own party repudiate you.

That didn’t happen to former Secretary of State Hillary Clinton when she made a move about suggesting a no-fly zone over Syria. Because that’s a kind of move people make sometimes in the NSDM. Some people disagreed with that as a Level B move, and maybe some of those people wrote Level A articles about how terribly wrong and serious it was; but nobody who understood that Level B existed went holy shit over it. From their standpoint the difference is clear-cut, even if it so happens that you can’t immediately perceive that difference and what Clinton did sounds maybe arguably just as bad as what Trump did. The holy shit reaction to Trump from senior national security bureaucrats, and from a lot of smart people you know who never seemed that worked up about Mitt Romney in 2012, is one cue as to what just happened. Even though you can’t tell the difference yourself and maybe shouldn’t expect to be able to tell the difference yourself. Certainly the media isn’t going to tell you about the difference, because everyone in the Level A theater has to pretend that all the professional wrestling is terribly, terribly serious.

Maybe you wish that Washington, DC culture would take sexual assault more seriously, as something deadly serious in its own right, as serious at it is possible to be — instead of some people laughing it off, some people being frankly offended, and everyone in Washington, DC tacitly understanding that this is not one of the issues that everyone has agreed to take deadly seriously even when no journalists are looking.

Maybe you look at that, and conclude that this ‘deadly serious level of politics’ thingy does not respect your own values and priorities. Maybe you conclude that the kind of political issues people are fighting over theatrically in the newspapers are, yes, every bit as vital to you as that so-called ‘deadly serious’ stuff, even if a lot of other people are treating them as entertainment.

I think you’re making a dreadful mistake. Scope is real. If you ever have to choose between voting a convicted serial abuser of children into the Presidential office — but this person otherwise seems stable and collected — versus a Presidential candidate who seems easy to provoke and who has ‘bad days’ and doesn’t listen to advisors and once said “Why do we have all these nukes if we can’t use them?”, it is deadly important that you vote for the pedophile. It isn’t physically possible to abuse enough children per day over 4 years to do as much damage as you can do with one wrong move in the National Security Decision-Making Game.

An evil but sane NSDM player is far, far less dangerous than an impulsive one who doesn’t care all that much about what the rules of NSDM are supposed to be.

That’s one reason why people at my level of national security expertise and above–and there’s a hell of a lot of headway above me on that one–went into holy shit mode over Donald Trump, on both sides of the aisle.

Programmer Michael Keenan details the circumstances under which ambiguous alliances sparked WW1 and WW2, as well as the Korean War and the Iraq-Kuwait War.

Maritime security author John-Clark Levin’s “How Trump Could Realistically Start a Nuclear War” picks one example of how Trump’s temperament is likely to increase the probability of international catastrophes:

[…] Trump has pulled off a stunning victory. Transition team Chairman Chris Christie’s first task is assembling a solid national security staff, but most of the top officials who would normally serve in a Republican administration refuse to serve under Trump (many have already stated this publicly).

So instead of the smartest and ablest leaders, Trump is forced to fill essential defense and intelligence posts with hacks  —  the B-level talent whose ambition overcomes their objections, and the C-level talent whose loyalty wins them undeserved appointments. Some good people do go to work in the Trump White House, judging they’d rather be in the Situation Room than the alt-righters who’d otherwise get their jobs. Maybe, they think, they can influence Trump for the better.

By the end of Trump’s first 100 days, though, it’s clear that his lifelong leadership style will not change. First, he surrounds himself with people who flatter him and tell him whatever he wants to hear, because he is a “great loyalty freak.” Second, he can’t take criticism or dissent, and sees these as disloyalty that must be punished. Third, he has a profound insecurity that cannot tolerate advice from big minds and strong personalities. Instead, he famously said, “Always be around unsuccessful people because everybody will respect you.” Over Trump’s first years in office, these traits force more good people out of government, replaced by those willing to follow Trump blindly.

Meanwhile, just as Trump boasted during the campaign that he knew more about ISIS than the generals, he now demoralizes top military leaders, publicly denigrating their competence even as he thwarts them with meddling and micromanagement. Trump continues to alienate the CIA, just as during the campaign he ignored consensus assessments by U.S. intelligence agencies in their briefings to him, instead bizarrely insisting that the DNC hacks weren’t the work of Russia, but perhaps a “400-pound man sitting on his bed.” Under these conditions, many of the best people in the Pentagon leave, and Trump makes good on his campaign statements by firing others.

On an October afternoon in 2019, terror strikes again  —  this time, a truck bomb in Jersey City kills almost a hundred people. Trump’s first instinct  —  as it has been throughout his long and well-documented life  —  is immediate and overwhelming retaliation. Despite contemplating nuclear counterattacks during the campaign (“Somebody hits us within ISIS  —  you wouldn’t fight back with a nuke?”), Trump now orders conventional punitive airstrikes on civilian areas in Raqqa, Syria. “We have no choice, people,” Trump says. Trump’s Justice Department insists that the strikes are legal, but as images of maimed children fill the airwaves, officers of conscience start retiring or resigning rather than participate in what they see as war crimes.