Cooperator or Competitor? How Human-AI Conflict and AI Capability Shape AI Aversion

Abstract

The growing use of Artificial Intelligence (AI) by a wide range of organizations is accompanied by specific challenges. In particular, AI can be a major source of conflict with its users as it takes on new roles and interferes with conventional processes and activities. Based on the theory of cooperation and competition, cognitive dissonance, and motivated information processing theory, this study delves into the complex interplay between human-AI conflict—specifically, competitive versus cooperative conflicts—and its consequent effects on AI aversion. In addition, by analyzing how different levels of AI capability influence employee responses across various decision-making criticality scenarios, this research also offers fresh insights into the psychological mechanisms driving AI aversion in the workplace. The results of two experimental design studies involving 759 participants reveal five major findings. First, competitive conflict significantly amplifies cognitive dissonance, compared to cooperative interactions. Second, AI capability affects the impact of human-AI conflict on cognitive dissonance, with greater dissonance in competitive conflicts when AI capability is high, and less pronounced differences when AI capability is low. Third, decision criticality affects the impact of AI capability on cognitive dissonance; in critical settings with competitive human-AI conflict, AI's speed of prediction exacerbates dissonance. In non-critical settings, AI's confidence in the accuracy of prediction heightens dissonance. Also, in both critical and non-critical scenarios, both high AI speed and confidence reduce cognitive dissonance in cooperative conflicts. Fourth, cognitive dissonance reduces epistemic and social motivation in processing information. Last, epistemic and social motivation reduce AI aversion. This study's findings underscore the need for tailored AI deployment strategies that consider both the nature of human-AI interactions and the specific decision-making contexts to mitigate adverse psychological impacts and foster more effective human-AI collaborations.