Summary

• Traditional outlets are dying. Loss of print advertising revenue has contributed to poor quality content.

• Other models are trying to fill the gap. They cobble together distribution and monetisation solutions but scale is limited.

• Distribution, monetisation, cost reduction and content production are the four key opportunities for technology, and specifically AI.

Traditional local media is dying

Revenue: Squeezed by Facebook and Google

Major traditional local news has relied on advertising. Globally Facebook and Google have taken 80% ad market share, squeezing out print.

• ‘Yet hundreds of local newspaper titles have closed in the past two decades. News publishers’ traditional print revenues have collapsed as people increasingly read news online. The smaller audiences for local news make shifting to digital business models based on online advertising or subscriptions particularly difficult.’ Source: House of Commons Committee report on Sustainability of Local Journalism, January 2023

• Digital readers are worth one eighth of a print reader. Traditional outlets use traffic-based programmatic advertising through Google and Meta, which local media can’t win (Source: https://manchestermill.co.uk/p/i-worked-for-the-north-wests-local

• Subscriptions are a winner-take-most market.

• A minimal staff of six costs £250,000 PA to run. To cover this a publication would need to convert 10% of the local population.

Content: Increasingly irrelevant

Focused on neighbour complaints, crime, housing developments, swamped with ads. Not compelling for a modern audience.

• See https://www.gazette-news.co.uk/ as an example.

Consolidation:

• 295 (20%) of the UK’s local publications closed between 2005 and 2021. Source: Media Reform Coalition

• After Newsquest bought Archant they now control 30% of the overall market.

• 50% of local authorities are covered by only one outlet.

The Coronavirus pandemic has accelerated the long-term decline in both circulation and

advertising revenues in the local press industry [...] since 2005 publishers have closed as many as 295 local newspapers—shrinking the UK’s local press by a fifth. [...] The largest local publishers have [cut] costs by consolidating distinct daily and weekly print titles into regional ‘hubs’ or rebranded online-only services. The result has been a growth in the scale of news ‘deserts’ around the UK

Source: https://www.mediareform.org.uk/wp-content/uploads/2021/03/Who-Owns-the-UK-Media_final2.pdf

Other models are trying to fill the gap

Content: Higher-quality

New local media often covers a broader spectrum of topics than traditional print media.

• The Manchester Mill aims for longer-form content. Charges £7 a month through Substack

Distribution and revenue: Struggling

But their distribution channels are either print (which is fine for some) or inconvenient platforms – for example a PDF on Tumblr.

• Peckham Peculiar: Just a tumblr site with a PDF. Need crowdfunding since advertising from local businesses has been ‘a bumpy ride’. Source: Peckham Peculiar crowdfunding campaign.

See others:

• El Mundo Latino ad-heavy but print + PDF https://www.facebook.com/MundoLatinoNewsUK/

• Glaven Valley Newsletter: Older, audience in coastal Norfolk. But in financial trouble Recently ran an appeal that asked for donations via cheque or bank transfer. https://glavenvalleynewsletter.weebly.com/

• New Cross Gate Newsletter https://www.nxgtrust.org.uk/

Digitisation: Can be slow

The Manchester Mill is digitally native.

Many local publications are slow to adopt digital solutions:

• ‘[S]uccessfully navigating and engaging with the online advertising market, setting up a subscription service, or devising other forms of business model innovation requires a level of technical and financial resource that smaller local news publishers can often lack.’ Source: House of Commons Committee report on Sustainability of Local Journalism, January 2023

Four key opportunities

Solution space: an opportunity from AI?

References

• https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2021/local-news-unbundled-where-audience-value-still-lies

• https://localnewsinitiative.northwestern.edu/posts/2022/08/16/local-news-business-models/

• https://stratechery.com/2017/the-local-news-business-model/

• https://journalift.org/overview-of-the-most-popular-media-business-models/

• https://www.pewresearch.org/journalism/2012/03/05/search-new-business-model/

• https://publications.parliament.uk/pa/cm5803/cmselect/cmcumeds/153/summary.html

• https://manchestermill.co.uk/p/i-worked-for-the-north-wests-local

• https://manchestermill.co.uk/p/the-case-for-a-new-newspaper-in-manchester

Pocket links I’ve saved over the years:

• https://www.bloomberg.com/news/articles/2018-09-05/local-news-is-dying-and-it-s-taking-small-town-america-with-it?utm_source=pocket_reader&leadSource=uverify%20wall

• https://www.niemanlab.org/2018/11/can-this-network-of-lit-to-be-local-newsletters-unlock-younger-civic-engagement/?utm_source=pocket_reader

• https://www.niemanlab.org/2019/02/inside-insides-new-local-newsletters-and-its-plans-to-keep-scaling-with-750000-active-subscribers-on-board/

• https://techcrunch.com/2020/01/28/flipboard-expands-into-local-news/?utm_source=pocket_reader

• https://www.niemanlab.org/2020/09/as-local-news-outlets-shift-to-subscription-they-wonder-what-should-facebooks-role-be/?utm_source=pocket_reader

• https://www.theguardian.com/australia-news/2021/jan/04/inside-the-hyper-local-world-of-facebooks-buy-nothing-groups?utm_source=pocket_reader

• https://www.cjr.org/the_media_today/complicating-the-save-local-news-mantra.php?utm_source=pocket_reader

• https://www.cjr.org/local_news/local_reporters_decline_coverage_density.php?mc_cid=dd3b535065&mc_eid=17799bc251&utm_source=pocket_reader

• https://www.mediareform.org.uk/blog/newsquests-acquisition-of-archant-more-bad-news-for-local-journalism?utm_source=pocket_reader

• https://theconversation.com/local-elections-are-about-the-most-important-issues-affecting-our-daily-lives-so-why-do-they-always-become-a-referendum-on-the-prime-minister-182552

• https://www.niemanlab.org/2022/06/i-follow-all-these-pages-and-i-still-dont-get-answers-this-new-report-is-a-great-example-of-centering-consumers-in-local-news-research/?utm_source=pocket_reader

Problem: It’s difficult for News Media Outlets to be profitable

Background

Print advertising has collapsed as a revenue stream for News Media Outlets (NMOs), and although online advertising has grown, it doesn’t generate the same level of income.

We ask how NMOs can generate additional revenue without adding a lot of additional work? We found that merchandise is one potential avenue for profit.

We formulated this problem with the help of an expert in local journalism.

Decline in advertising revenue

Advertising revenue has declined massively. According to 2018 research for the Department of Digital, Culture, Media and Sport, newspaper print advertising brought in £4.62 billion in 2007 and £1.43 billion in 2017.

During that time digital advertising income went from negligible to £0.48 billion — nowhere near enough to cover the losses from the decline in print ads.

Specifically at a local level, total advertising income dropped from £2.74 billion in 2007 to £1.27 billion in 2017. And this situation has only worsened since then.

Stakeholder groups:

Businesses

• Large NMOs
• Local NMOs
• Individuals (YouTubers, Twitters, etc)

Customers

• with disposable income
• without disposable income

1. Large NMOs

Risks: High costs at the outset. Maybe nobody buys it. Brand damage is possible if merchandise is conflicts with the brand’s stated values (produced in sweat shops, etc)

Benefits: Generates revenue. Market segmentation associated with branding leads to higher profit

2. Local NMOs

Risks: High costs at the outset. Maybe nobody buys it. Even if people are buying it, stock takes up space and is costly to store and send. Distraction from the main business of doing good journalism. Customers potentially getting more value from the act of supporting local business than the utility of the product

Benefits: Creates awareness/acts as advertising. Generate revenue.

3. Individual NMOs

Risks: High costs, storage, risk that nobody will buy it. Brand damage is possible if merchandise is conflicts with the brand’s stated values (produced in sweat shops, etc)

Benefits: Create a sense of community among the customer base

4. Customers:

Risks: Cost stratification of the merchandise can lead to customer bases also being stratified by income level, so that more wealthy readers will gravitate towards certain NMOs and the type of news that a person reads will be more influenced by their income level.

Benefit: Can create a sense of belonging to a group through recognizing others with the same merchandise. Creates meaning for the individual by supporting journalism.

AI angles

AI-generated merchandise, for example, a personalised news article which is based on the customer and generated by AI in the style of the paper. Limited upfront costs, high branding value, no physical stock to store, limited staff costs.

Editor’s note: A team at Hack the Press June 2023 built this idea; find the demo here: https://htparticle.fly.dev/spectacular-storyteller and the slides here Spectacular StoryTeller.

Customer data to recommend what kind of merchandise the NMO’s specific customer base is most likely to buy. Doing the market research so that the business can make a decent profit and limit the risks associated with creating products

Problem

Misinformation is harder to spot on TikTok because of the speed at which it goes viral.

Example: Trial of Johnny Depp and Amber Heard

• Mass of anti-Amber Heard content led to the hounding of a real person, and likely influenced the decision of a jury.
• Jury members admitted to watching content about the trial online.
• At the height of the trial, Amber Heard became one of the most abused people on the planet, despite domestic violence charities worldwide expressing concern about the public’s response to the trial.

• The nature of TikTok’s recommendation algorithm means that sensational content, including falsified or exaggerated facts, conspiracy theories, and disinformation spread on a scale quicker than other social media for high profile or much-talked about cases.
• For example the trial of Johnny Depp and Amber Heard and major geopolitical events like the war in Ukraine or elections.

Future direction

• Limitation of the reach of harmful misinformation or inflammatory content without infringing on freedom of expression principles or ruining content creators’ ability to use comedy.
• Perhaps an app that can log into TikTok or an extension that uses AI to flag content on a user’s feed that is likely false information - similar to Twitter’s community notes feature but might not be as diligent.

Summary

• People need credible information to make decisions that rely on information from their environment.
• But it’s getting harder to know which information to trust.
• We use mechanisms like consensus, raw evidence and professionals. Sometimes this fails – we get conned and overwhelmed.
• We should explore how people who share information and build credibility and how consumers can verify it.

Why do I rely on credibility?

I need to make decisions, and I need good information to make those decisions.

Therefore, I need to know what my local, national and global environment looks like.

• “I need to decide what movie I’m going to see with my friends”
• “I need to decide the best way to spend my money on consumer goods, charity, etc”
• “I need to decide which politicians I vote for”

Key reason: I lack credible information

We have plenty of information with which we can make decisions, but not all of it is good. Therefore the problem we’re trying to solve is having credible information.

• Anyone can make content, and so much already exists.
• We can’t verify most of it and can’t even tell if it’s real!
• This problem will be amplified due to generative AI.

How do I determine what’s credible?

Three ways to determine what’s credible

Credible information appears in at least three forms. But each one has its issues.

Consensus

• Form: “common knowledge”, voting blocs, communities.
• Issue: Artificial consensus created by echo chambers.

Evidence

• Form: photo, video, textual, biometric data, metadata?
• Issue: How do we prevent it from being corrupted or misrepresented?

Professional truth seekers

• Form: Journalists and Data professionals
• Issue: How do we make sure they do their jobs?

I can be an individual or part of an organisation

Organisations

• When the BBC wants to report the weather, they trust meteorologists.
• Government organisations, e.g. Statistical Bureaus, Police.
• Independent orgs that produce data (housing data, market research)

How do individuals decide?

• There’s a degree of spoon-feeding
• Sometimes through independent research of varying quality. Can we support user research and curiosity?

What does the problem of credibility look like in practice, when credibility fails or isn’t there?

People can make decisions against their own interests.

• Voting based on false promises and false information.
• People trusting bad actors (con men, unlicensed physicians).
• Making plans that get subverted by reality.
• People failing to make decisions at all due to uncertainty.

Aspects of credibility to explore

Personal credibility

• Contextual credibility: could someone’s authority/credibility be more visibly nuanced when they speak on things?
• Buildable credibility: Is it possible to build credibility over time?  How, and what would be necessary for this to happen?
• Could both of these be applied to individuals and organisations, respectively?

Professional credibility

• How could we leverage networks to establish trust and build personal credibility of journalists/content contributors/sources?
• How could this credibility be presented in a way that helps people feel comfortable that the way they interact with, enjoy, understand, act on insights is appropriate for the level of trust they have, and where it fits on a spectrum of art/fiction/true story/fact.
• Can contributors indicate/suggest/verify the credibility they feel that should be placed on their own content and could it be buildable, contextual, topic-related, altered according to the forum for conversations etc?

Tiny problem statement

Some news sources are perceived to be more harmful than others. How do we know whether the content that is created does more good than harm? One way to measure this might be to look at the user needs that articles address.

This is a small hack that you could iterate on...

Tiny exploration of the problem

The smartco user needs model describes a range of user needs for journalism.

What if you tagged and categorised e.g. a teen magazine - which needs are they addressing? Where are the gaps?

For digital news sources, what could you learn from the share rates of articles addressing different needs? Which are the anomalies? Can you use this sort of data to reason about news avoidance?

Tiny suggestion

Build a categoriser for news article needs. This may require you to develop some rules that generative AI can use to determine which needs an article is attempting to meet.

The internet archive has a newspaper category, from which you can download sources

British Library newspaper archive is free to search, contains many types of news

• Obituaries
• Wedding notices
• Letters to the editor
• etc.

You could use the above sources as training data.

Problem space

People want to know about:

• Local crime
• Restaurants
• Weather
• Traffic
• Local events
• Planning and housing
• Local services (council, schools, handymen)
• Local interest

The most expensive to produce are:

• Investigations: These are expensive because of the higher time investment required and often higher count of staff involved due to breadth of skills required

Journalists need to:

• Accumulate information from different sources
• Verify information in these sources, establish what is important and irrelevant
• Document all their actions taken throughout the process
• Evaluate their findings and construct a story
• Expensive and potentially extensive legal cross-checks
• Create content to publish

The expensive part of investigations is the time spent by a journalist to understand information from different sources and verify the information

• It isn’t going to be viable to just ‘apply AI’ here, but are there recurring aspects of investigations where it could help?

The highest revenue generating are:

• Assuming an ad-based model, the highest revenue generating are likely to be clickbait
• Or could be broad and general - i.e. exactly what local news can’t be

This is a problem because:

• The revenue for each article is too low to sustain the journalistic effort to produce it
• This reduces the incentives for media outlets to create quality content, reducing access to local news

Solution space

Key opportunities:

• Automatically generate articles/video.
• Planning data
• AI could potentially improve the quality, breadth and relevance of local news stories
• Suggest articles based on trends in/new data
• Can AI comb through financial records to find wrongdoing?
• Can journalists gather information from sources in a lower-cost way?
• Can journalists gather expert opinions in an automated way?
• Identify leading voices or subject matter experts in a particular subject, gather their contact details together, and submit requests for comment

Summary

• Home-made news - abuse of the forwarding function on WhatsApp / home-made news content generation on social media causes loss of trust in platforms and news
• Loss of trust by users - who hope to be informed of local / global events in an impartial way.
• Hurts platforms - trust is a soft selling point for user adoption

Key Problem

Modern devices and the internet allow users and news audiences the ability to

• (i) Consume articles and media from various online traditional news outlets and social media platforms (e.g. Facebook, YouTube, News Websites)
• (ii) Create, upload and distribute their own media and information (Twitter, Facebook etc); and crucially
• (iii) Share “organic content” privately or semi-privately on their own social media platforms or on private end-to-end messaging functions

Various parts of the population experience a lack of trust in traditional journalism or social media platforms, or simply face information overload, engage in avoidance of news articles. They may rely on their local communities and friends to fact-check and determine the relevance of news articles to suit their various needs - and share them through links / copy and paste methods / liking statuses.

• Sometimes this is useful - live tweeting an active ongoing event.
• Sometimes this is just benign - e.g. sharing the location of your favourite doughnut shop
• Sometimes this is problematic - spread of fake news (e.g. forwarding of articles with no provenance on WhatsApp, copy/pasted/shared), up to problems as dramatic as a lack of digital literacy caused Incitement of violence against the Rohingya, pandemic fake news.

No profit incentive for managers of private end-to-end communications to regulate fake news - WhatsApp, or resourcing problems for social media moderators

• Revenue from business API integration rather than consumer problems
• No meaningful way of imposing moderation requirements without imposing disproportionate costs on businesses

How AI can exacerbate this problem:

• Increasing the volume of the problem. Creating more content which can overwhelm consumers.
• Black box / AI generated content as a problem of provenance - different kettle of fish.

Key opportunities/Solution ideas?

Content: Technology-based solutions to:

• identify unsourced content / AI-generated content,
• promote critical thinking by explaining interests behind a story,
• cross-reference live user content with other similar sources,
• explain context, if necessary.

The Self Administered Interview, recently approved for rollout by the College of Policing, captures structured information from witnesses of large events.

Designed by the Goldsmiths Forensic Psychology Unit it’s intended to help eliminate or at least reduce personal bias by asking a series of structured questions. It’s designed to create sufficient information to identify key witnesses, and populations of people who had similar or different experiences of an event.

In 2017, participants at Hack the Police 2 built a couple of digital prototypes to illustrate ways to capture information at the scene of an event from witnesses before they dispersed.

Can something similar apply to news? How do you represent and capture structured information about world events?

The wayback machine has a copy of the BBC News Labs Manifesto for Structured Journalism published in 2015.

“There is a wealth of knowledge created during the ‘gathering and assessing’ phases of reporting that most publishing systems ignore. In the work we’re pursuing this summer, BBC News Labs is taking a look at “structured journalism” - can we empower journalists with better ways of working with information, beyond the headline and text of a story? How can this affect the quality of our reporting? Can it promote a greater public understanding of current affairs and issues - what the BBC’s Royal Charter describes as “sustaining citizenship and civil society?”

The BBC R&D blog describes the Mythology Engine - an application of this for fiction, which might just as well serve factual news.

Can we help journalists collect and maintain a structured database of facts and news, that can then be used to improve audience experiences?