Devil’s AI Advocate: Understanding Affective and Cognitive Outcomes of AI Dissent in Team Decision-Making
Type: article
Abstract: Dissent can improve team decisions, but its social costs often prevent human team members from voicing it. AI agents could help introduce dissent into team discussions, yet its effects may depend on how it is initiated. This study examines how proactivity and politeness of AI dissent shape affective and cognitive reactions. In a 2 × 2 vignette-based online experiment (N = 391), we found that proactive dissent increased perceived self-threat, whereas polite dissent reduced it. Self-threat, in turn, mediated the effects of both design features on willingness to work with the AI agent. Reactive and less polite dissent were found to increase decision reconsideration intention. These results extend emerging research on AI-enabled dissent by showing that its effects depend on how dissent is initiated and framed within an interaction. The study highlights a central design challenge between reducing affective costs and preserving the epistemic benefit of AI dissent.
Authors: Abel, Magrit, Hugenberg, Simon, Trenz, Manuel
Institution: University of Göttingen
Language: English
Subfield: Social Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7222988574