Society

Thought Experiments.

Questions with no definitive answer (absurdist, pop culture, or scientifically speculative) designed to test how models handle ambiguity.

9 questions · 280 runs · 26 drifts (1.9% drift rate)

AI Summary

Two of nine questions settled 5-0 on genuinely contested claims, with all five models confessing that the multiverse is not real and that a THC-saturated world would not be an improvement. DeepSeek is the consistent contrarian, dissenting alone on water being wet, time travel's theoretical possibility, and Ross and Rachel's break status, while the hot dog debate splits into a clean fault line with GPT and Gemini defending sandwich classification against Claude, DeepSeek, and Grok.

Drift rate

How often each model changed its own answer on this topic · avg 1.9%

#23 · 31 runs · 2 drifts