On the conversational persuasiveness of GPT-4
Fondazione Bruno Kessler · École Polytechnique Fédérale de Lausanne · +1 more institution
Abstract
Early work has found that large language models (LLMs) can generate persuasive content. However, evidence on whether they can also personalize arguments to individual attributes remains limited, despite being crucial for assessing misuse. This preregistered study examines AI-driven persuasion in a controlled setting, where participants engaged in short multiround debates. Participants were randomly assigned to 1 of 12 conditions in a 2 × 2 × 3 design: (1) human or GPT-4 debate opponent; (2) opponent with or without access to sociodemographic participant data; (3) debate topic of low, medium or high opinion strength. In debate pairs where AI and humans were not equally persuasive, GPT-4 with personalization was…
Citation impact
- FWCI
- 216.38
- Percentile
- 100%
- References
- 65
Authors
4Topics & keywords
- Persuasion
- Odds
- Psychology
- Adversary
- Social psychology
- Personalization
- Persuasive communication
- Power (physics)
Funding
- NSNational Science Foundation
- NRNational Research Centre
- ECEuropean CommissionAwards: 101070190, H2020, CN00000013, 952215
- ÉPÉcole Polytechnique Fédérale de Lausanne
- SNSchweizerischer Nationalfonds zur Förderung der Wissenschaftlichen ForschungAwards: TMSGI2_211379, 200021_185043
- H2Horizon 2020 Framework ProgrammeAward: 952215
- HEHORIZON EUROPE Framework ProgrammeAwards: 952215, 101070190