Data Skeptic Book Ratings and Recommendations
38 snips
Mar 27, 2026 Hannes Rosenbusch, a University of Amsterdam researcher and fiction author, explores how reader differences often outweigh book differences. He talks about why star ratings mislead, how metadata and reviews fall short, and how LLMs and the ISAAC method can help model individual reading preferences. He also contrasts LLMs as research tools versus creative editors.
AI Snips
Chapters
Transcript
Episode notes
Ratings Reflect Readers More Than Books
- Goodreads star ratings mainly reflect reader differences rather than book differences.
- Hannes found that variance in ratings is dominated by which reader rates, so professionally published books cluster in the same score corridor.
Reviews Reveal Reviewers Not Texts
- Written reviews reveal reviewer tendencies more than consistent book-level issues.
- Hannes used review-content analysis and found reviewers mention personal complaints that don't reliably repeat across readers for the same book.
Stable Aggregates Don't Predict Personal Taste
- Large aggregate ratings can stabilize but still poorly predict individual enjoyment.
- Hannes notes books with tens of thousands of ratings can be reliably different at population level yet correlate weakly with a single reader's tastes (his correlation ~0.2).