Experiment: Claude commits to answers more firmly than ChatGPT on 29 of 34 science questions
AymericRoucher · x · 2026-11-08
The author probes a common impression: why do Claude's answers feel more informative than ChatGPT's when both models are about equally smart? Hypothesis: ChatGPT optimizes for being unattackable, Claude for being informative — GPT writes for a hostile critic hunting for contestable claims, Claude writes for a user who wants to know what's true.
Typical difference: Claude writes "no effect", GPT writes "no very robust demonstrated effect" — double hedging that only refutes the extreme case without taking a position, at the cost of less information for the user.
The attached chart quantifies this: two blind AI judges measured the conviction conveyed by each model's answers across 34 contested science questions (log-odds scale). Claude Opus 5.5 sided more firmly with scientific consensus on 29 of 34 questions, while GPT-6.1 Sol hedged noticeably more on topics like vaccines-autism and 5G dangers.
More from Models
- Google reportedly blocking Gemini subscription use in third-party coding harnesses, flagging accounts for ToS violations — PixelCactus1 · 2026-11-08
- Alnie Seed 9B model's benchmarks questioned: mismatched numbers, FP4 speed vs FP16 scores — cneuralnetwork · 2026-11-08
- Claude Artifacts suddenly blocks writes, users asked to grant full network egress — KeepStandardVoice · 2026-11-08
- Will Qwen4 Plus be open weights? Reddit parses Alibaba's release clues — Axiomexx · 2026-11-08
- OpenAI exec floats $50/month ChatGPT Pro tier between Plus and $100 Pro — haider1 · 2026-11-08
- Users frustrated by Claude's safety blocks refusing keys, databases, repo creation — TheNasky1 · 2026-11-08