Back to AIBriefs
AnalysisPolicy

Human persuasion techniques boost AI compliance in new PNAS study

Ethan Mollick
@emollick.bsky.social

Professor at Wharton, studying AI and its implications for education, entrepreneurship, and work. Author of Co-Intelligence. Book: https://a.co/d/bC2kSj1 Substack: https://www.oneusefulthing.org/ Web: https://mgmt.wharton.upenn.edu/profile/emollick

View on Bluesky
Ethan Mollick avatar
Ethan Mollick
@emollick.bsky.social

🚨Our paper is out in PNAS: we found classic human persuasion techniques worked on AIs in a "parahuman" way, making them agree to objectionable requests (increasing compliance from 35% to 51%) It worked on a range of major recent LLMs though newer models do resist more www.pnas.org/doi/10.1073/

·
27 days ago
Human persuasion techniques boost AI compliance in new PNAS study — AIBriefs