Paper asks whether LLMs are rational: coherence norms published in AI journal
peterbhase · x · 2026-11-05
peterbhase announced that his paper on beliefs in LLMs — specifically whether coherence norms apply to them — has been published in the journal Artificial Intelligence, about two years after submission.
The CS+Philosophy paper tackles a foundational question for interpretability theory: are language models rational? It examines whether the convenient explanatory equation "behavior = beliefs + desires" can legitimately be applied to LLMs, i.e., whether the rationality assumption underpinning belief-desire explanations holds for language models.
The work carries direct implications for how we conceptualize model interpretability and whether treating models as rational agents is theoretically sound.
More from AGI Musings
- OpenAI staffer: data from paid work is the data model that will last — shyamalanadkat · 2026-11-05
- MLOps engineer: narrow specialization will hurt your career in the AI era — kmeanskaran · 2026-11-05
- Founders warn 'Pause AI' policies would force startups to ration frontier-model inference — fkasummer · 2026-11-05
- AI Moats Are Talent and Distribution, Not Models, Argues Investor — gethackteam · 2026-11-05
- Alpha School graduate on record: he barely read books, contradicting 'books are central' claim — benjaminjriley · 2026-11-05
- Stanford's James Zou on AI agents entering the wet lab, in Nature — james_y_zou · 2026-11-05