【日　時】 2026年2月20日（金）16:00-17:30

【場　所】白金キャンパス高輪校舎2階（15202教室）

【Zoom】https://us02web.zoom.us/j/81473519624
ミーティング ID: 814 7351 9624
パスコード: 892269

【講演者】Dr. Hugo Mercier

フランスCNRS研究者としてJean Nicodt Instituteに所属する認知科学者

https://sites.google.com/site/hugomercier/

【タイトル】

Reliable inferences of knowledge, skill, and belief from a minimal amount of information

【Abstract】
Humans acquire much information from individuals more knowledgeable than them. But, when we are ignorant, how can we tell who is knowledgeable? Past research has focused on indirect and unreliable cues to knowledgeability, from intonation to prestige. Alternatively, people might use the nestedness of knowledge to reliably infer who is knowledgeable. If knowledge is nested, and if someone possesses a rare piece of knowledge (but not a common piece of knowledge), then that individual is likely broadly knowledgeable. This means that, if people can estimate the rarity of different pieces of knowledge, and if knowledge is nested, then they can reliably infer people’s knowledgeability from a minimal cue: the possession of a single piece of knowledge. A series of pre-registered, and replicated experiments (total N>2000, US participants) show that (i) participants can accurately assess the rarity of pieces of knowledge, and (ii) if a participant knows someone possesses one piece of knowledge, they can infer their knowledgeability in the relevant domain. Moreover, this is true even for participants who are themselves ignorant in this domain. We also show that most participants integrate information in line with an optimal Bayesian process. Finally, we extend our findings in two other domains: in arithmetic skills, we show that participants can correctly guess someone’s arithmetic skills from knowing whether they succeeded or failed at performing one operation; in conspiracy theories, we show that participants can correctly guess which conspiracy theories someone would endorse from know that they endorse (or don’t endorse) one given conspiracy theory.