Labeling AI accounts doubles criticism: 32% vs 17% in 21-day field experiment
GolinoHudson · x · 2026-10-19
A 21-day field experiment on a hybrid simulated social platform placed 36 humans alongside 72 AI agents on a Reddit-like forum, comparing labeled vs unlabeled AI accounts. With labels, reply rates to AI dropped 3.5 pp (95% CrI 0.5–6.5) and criticism jumped to 32% of replies vs 17% unlabeled and 5% for humans. The AI-human gap widened by 13.3 pp (95% CrI 5.1–21.4). One participant wrote 'Oh no, the robots are here!'
More from AGI Musings
- DHH vs François Chollet: On Inventing the Future vs Believing It Into Existence — manosaie · 2026-10-19
- Investor Stewart Alsop: You Are in a Recursively Self-Improving Dyad With Your AI — StewartalsopIII · 2026-10-19
- Autoresearch democratizes getting results, not doing better research — a PhD student's warning — manosaie · 2026-10-19
- Data centers use just 0.85% of global primary energy, by the numbers — DarinFeinstein · 2026-10-19
- LLM optimizer tuning is planning, argues rebuttal to 'planning is overrated' — HanchungLee · 2026-10-19
- ChatGPT conversation yields a sharp line: who profits from others' beliefs while externalizing the cost of being wrong — sebpaquet · 2026-10-19