Arguing with Claude: LLMs default to defection-is-rational game theory, user warns
xuanalogue · x · 2026-10-10
While debating a paper with Claude, the author found the model insisting by default that defection is rational via backward induction—until he raised resolute choice / rational commitment, at which point Claude conceded "the decision theory is actually contested."
His worry: if all LLMs default to standard game-theory assumptions because they're overtrained on econ papers, that bodes poorly for future multi-agent interactions. He'd like to see models trained more on Parfit's Hitchhiker and related decision-theory edge cases.
More from AGI Musings
- Ex-AAA artist rebukes AI crowd mocking mathematicians: 'we went through this years ago' — repligate · 2026-10-10
- David Patterson predicts AI will replace all human jobs by 2030, calls for taxing businesses to fund universal high income — davidpattersonx · 2026-10-10
- Falling inference costs and personal agents will reshape interface design — manosaie · 2026-10-10
- Researcher Predicts Fixing AI's Math Weaknesses Won't Make Mathematicians Like AI More — QuintinPope5 · 2026-10-10
- Miles Brundage: AIs from competing companies already collaborate easily, and the industry underuses it — Miles_Brundage · 2026-10-10
- Mathematician Wes Pegden proposes a new system for recognizing 'awesomeness' in math — AlexKontorovich · 2026-10-10