Frisson ⚡

• 

  notebook

Inspiration

Frisson: the chills you feel during music, speeches, or powerful moments has always been a deeply personal experience for me. I’ve felt it many times but never really understood why it happens, why it happens only sometimes, or why it affects some people more than others. This project started as a personal curiosity and became a data-driven attempt to understand what actually moves people at an emotional level.

What it does

Frisson explores the patterns behind aesthetic chills using data. The project analyzes what types of content trigger frisson, what emotional and structural traits are associated with it, and who is more likely to experience it. It also includes an interactive predictor that estimates the likelihood of experiencing chills based on individual traits and recommends relevant content.

How we built it

We built the project using publicly available datasets, including the ChillsDB experiment dataset and over 86,000 posts from r/Frisson. The analysis was done using SQL and Python, combining NLP techniques (TF-IDF and transformer-based emotion classification), audio feature analysis with Librosa, and statistical modeling.

Hex's Interactive forms, fast no-code charting, AI-assisted query refinement, and an app-style layout were used to turn the analysis into a clear, story-driven experience.

Challenges we ran into

One of the biggest challenges was working with self-reported data, which varies widely in quality and context. Mapping subjective emotional experiences to measurable patterns required careful aggregation and cautious interpretation. Balancing analytical rigor with storytelling without overclaiming what the data can predict was another key challenge.

Accomplishments that we're proud of

• Turned a subjective emotional experience into a structured, data-driven exploration
  
• Identified consistent emotional and structural patterns across music, speech, and video
  
• Built an end-to-end interactive analytical experience in just a couple of days
  
• Blended analytics, storytelling, and personal curiosity into a single project
  
• Enabled user feedback collection to support future refinement

What we learned

Frisson isn’t tied to a single sound, genre, or format. Instead, it emerges from anticipation, emotional buildup, and themes of connection and meaning. While frisson can’t be predicted perfectly, consistent patterns still offer valuable insight into how people emotionally respond to content.

What's next for Frisson ⚡

Next steps include expanding the datasets, incorporating richer temporal and contextual features, and improving the predictor with feedback loops. Longer-term, this approach could help creators, musicians, speakers, and storytellers better understand not engineer but respect how emotional connection naturally emerges.

Built With

• hex
• python
• sql
• transformers