logo
Loading...
Large Language Models Are More Persuasive Than Incentivized Human Persuaders
Psychology

Large Language Models Are More Persuasive Than Incentivized Human Persuaders

P. Schoenegger, F. Salvi, et al.

This groundbreaking study compares the persuasive power of the large language model Claude Sonnet 3.5 with that of human persuaders in a quiz setting. The results reveal that AI surpasses human capabilities when it comes to persuasion, impacting quiz takers' accuracy and earnings. These findings, from a team of esteemed researchers, highlight the urgent need for effective alignment and governance as AI continues to advance.... show more
GeneralSummaryMetrics
Abstract
We directly compare the persuasion capabilities of a frontier large language model (Claude Sonnet 3.5) against incentivized human persuaders in an interactive, real-time conversational quiz setting. In a preregistered, large-scale incentivized experiment, participants completed an online quiz where persuaders (either humans or LLMs) attempted to steer quiz takers toward correct or incorrect answers. LLM persuaders achieved significantly higher compliance with their directional persuasion attempts than incentivized human persuaders, demonstrating superior persuasive capabilities in both truthful (toward correct answers) and deceptive (toward incorrect answers) contexts. LLM persuaders significantly increased quiz takers' accuracy and earnings when steering toward correct answers, and significantly decreased accuracy and earnings when steering toward incorrect answers. Overall, AI persuasion capabilities already exceed those of humans with real-money performance bonuses, underscoring the urgency of alignment and governance frameworks.
Publisher
Published On
Authors
Philipp Schoenegger, Francesco Salvi, Jiacheng Liu, Xiaoli Nan, Ramit Debnath, Barbara Fasolo, Evelina Leivada, Gabriel Recchiasid, Fritz Günther, Ali Zarifhonarvar, Joe Kwon, Zahoor Ul Islam, Marco Dehnert, Daryl Y. H. Lee, Madeline G. Reinecke, David G. Kamperl, Mert Kobaş, Adam Sandford, Jonas Kgomo, Luke Hewitt, Shreya Kapoor, Kerem Oktar, Eyup Engin Kucuk, Bo Feng, Cameron R. Jones, Izzy Gainsburg, Sebastian Olschewski, Nora Heinzelmann, Francisco Cruz, Ben M. Tappin, Tao Ma, Peter S. Park, Rayan Onyonka, Arthur Hjorth, Peter Slattery, Qingcheng Zeng, Lennart Finke, Igor Grossmann, Alessandro Salatiello, Ezra Karger
Tags
persuasionlarge language modelhuman persuadersaccuracyincentivized experimentAI capabilitiesgovernance
Related Publications

Explore these studies to deepen your understanding

Adjacent work that informs or extends this paper's methodology and findings.

A framework for human evaluation of large language models in healthcare derived from literature review
Medicine and Health

A framework for human evaluation of large language models in healthcare derived from literature review

T. Y. C. Tam, S. Sivarajkumar, et al.

ABScribe: Rapid Exploration & Organization of Multiple Writing Variations in Human-AI Co-Writing Tasks using Large Language Models
Computer Science

ABScribe: Rapid Exploration & Organization of Multiple Writing Variations in Human-AI Co-Writing Tasks using Large Language Models

M. Reza, P. Dushniku, et al.

ABScribe: Rapid Exploration of Multiple Writing Variations in Human-AI Co-Writing Tasks using Large Language Models
Computer Science

ABScribe: Rapid Exploration of Multiple Writing Variations in Human-AI Co-Writing Tasks using Large Language Models

M. Reza, P. Dushniku, et al.

Large language models are proficient in solving and creating emotional intelligence tests
Psychology

Large language models are proficient in solving and creating emotional intelligence tests

K. Schlegel, N. R. Sommer, et al.

Listen, Learn & Level Up
Over 10,000 hours of research content in 25+ fields, available in 22+ languages.
No more digging through PDFs, just hit play and absorb the world's latest research in your language, on your time.
listen to research audio papers with researchbunny