AI-Simulated Survey Matches Real Human Rankings: Spearman r=0.886 in Head-to-Head Match-ups
soumitrashukla9 · x · 2026-11-13
Responding to "this is just AI's opinion — what about real humans?", economist John Horton ran a human-vs-AI comparison study with Prolific: AI simulations (n=18) and real respondents (n=5) judged the same options in head-to-head match-ups.
Results showed strong agreement: win rates track closely (Photo F: 100% vs 100%; Photo A: 80% vs 78%), and Bradley-Terry scores correlate at Spearman r=0.886 (p=0.0188). Tiny samples make this preliminary, but it's a reproducible paradigm for testing whether AI can simulate human survey responses.
More from Fun
- One prompt spawns a 1M-nanobot GPU swarm that morphs into any shape at 60fps — bennash · 2026-11-13
- "Nvidia means envy in Latin, founder's downfall was Lust" — the internet's favorite finale plot — menhguin · 2026-11-13
- "It's 2027 and the biggest person in the news is a man named Jensen?" — zetalyrae · 2026-11-13
- User vents: GPT-6.1 Terra 'always finds a way to make me extremely angry' — Angaisb_ · 2026-11-13
- DeepSeek V4 Pro agent sets up a full Hive lab in VMware in about half an hour — huangyun_122 · 2026-11-13
- Group chat meme: "the face that launched a thousand chips" — NathanpmYoung · 2026-11-13