Eval orgs overindex on benchmarks, neglecting societal impacts

evijit · x · 2026-09-14

The paper author adds that mainstream evaluation orgs pay too little attention to societal impacts — Collective Intelligence is the only one that comes to mind, with the rest scattered across FAccT papers. Overindexing on benchmarks is cited as a major reason for this gap.

Related event: ICML Paper Finds AI Social Impact Evaluations Shrinking(3 posts)→

Original post →

More from Research

Research channel →