Maintain Preference for Autonomous System or Human-Controlled System? The Influence of Trust Conflict and Algorithm Aversion

With the advancing intelligence of machines, human-agent collaboration across varying intelligence levels will become prevalent in future decision-making scenarios. This study focuses on trust conflicts in human-agent collaboration. Study 1(military scenarios, N = 495) investigated initial trust preferences toward different collaborative agents and how subsequent decision aid advice sources/types influence trust decisions, developing a trust conflict quantification model with parameters θ (advice source) and β (initial preference). Study 2 (engineering consulting scenarios, N = 108) extended the initial-preference binary choice to a continuous choice, and further explored initial trust tendencies and advice source impacts. Results show that collaborator preferences vary by scenario type. When decision support advice conflicts with initial trust preferences, individuals adjust their final trust tendencies. Across tasks, trust behaviors are jointly shaped by initial preferences and advice sources. These findings guide the design of adaptive advice strategies in human-machine collaborative systems and provide a theoretical basis for context-specific decision support.