Investigating anthropomorphic cues in AI systems and their impact on AI’s persuasive potential

Description

Artificial Intelligence (AI)-based technology is increasingly exploited to spread misinformation and influence political opinions. Recent research demonstrates that Large Language Models (LLMs) can be equally persuasive as humans, particularly due to their capacity to personalize messages at an unprecedented scale (e.g. Matz et al., 2024). However, the psychological mechanisms of AI’s persuasive potential remain not fully understood. While AI’s persuasive capabilities are often attributed to its logical and analytical reasoning, social cues which play a critical role in human-to-human persuasion may further enhance AI’s effectiveness. To investigate the mechanisms underlying AI-driven persuasion, we conducted a user study manipulating different anthropomorphic cues – including identity, verbal, and non-verbal elements – in an AI agent’s interface and conversational style. We then measured perceived anthropomorphism and other system-related variables. The findings of this research are presented and discussed under the scope of developing empirically validated guidelines for designing AI systems that mitigate manipulative risks and enhance user resilience to balance persuasive efficacy with ethical and social considerations.

Period	11 Apr 2025
Event title	Human Factors and Ergonomics Society Europe Chapter
Event type	Conference
Location	Bologna, ItalyShow on map