Abstract
As conversational AI tools become primary resources for information, understanding the psychological drivers of user reliance is critical. This study investigates how individual decision-making "coping patterns"—specifically Vigilance, Buckpassing, and Hypervigilance—influence how users interact with AI advice. Through an online experiment with 810 participants, researchers found that "buckpassers" (those who defer decisions to avoid responsibility) are more likely to seek AI help and report higher reliance, yet they spend significantly less time scrutinizing AI explanations. This behavior makes them uniquely vulnerable to AI misinformation and hallucinations.
Introduction
Most research on AI trust focuses on system-level attributes like accuracy or transparency. However, this paper shifts the lens toward the user's internal psychological state. By examining how individuals handle the stress of making choices, the authors explore whether AI is being used as a tool for informed decision-making or simply as a way to escape the burden of responsibility.
Decision-Making Patterns
The study utilizes the Melbourne Decision-Making Questionnaire (MDMQ) to categorize participants into four primary coping patterns. These patterns dictate how an individual processes information and reaches a conclusion.
- Vigilance: Characterized by thorough information gathering and careful evaluation of alternatives.
- Buckpassing: A tendency to leave decisions to others to avoid the stress of responsibility.
- Hypervigilance: A frantic, rushed approach to decision-making aimed at immediate stress relief.
- Procrastination: Putting off decisions (though not the primary focus of this study).
Experimental Methodology
To test these patterns in an AI context, researchers conducted an experiment on the LabintheWild platform. Participants were asked to evaluate 10 nutrition-related statements (e.g., "The digestion process begins in the mouth") to determine if they were facts or myths.
Crucially, AI assistance was optional. Participants could choose to see the "AI Decision" and then further choose to see an "AI Explanation." The AI (GPT-4) was programmed to provide a mix of accurate and inaccurate suggestions to observe how users reacted to potential "hallucinations."
Key Insights
The Buckpassing Paradox
Individuals with high buckpassing scores were the most likely to seek out AI suggestions and reported the highest subjective reliance. However, they spent the least amount of time reading the AI's explanations. For these users, the AI serves as a "get out of jail free" card for decision-making stress rather than a source of knowledge.
Vigilance as a Scrutiny Filter
In contrast, "vigilant" individuals did not necessarily seek AI help more often. However, when they did, they spent significantly more time scrutinizing the explanations. For every one-point increase in a user's vigilance score, they spent an average of 0.53 seconds more reading AI explanations. They use AI as a data point to be cross-referenced, not a final authority.
Vulnerability to Misinformation
Because buckpassers seek answers but skip the reasoning, they are significantly more susceptible to AI errors. The study found that participants agreed with the AI's decision approximately 83% of the time, even though the AI was intentionally wrong in 4 out of 10 cases. This blind agreement suggests that buckpassing behavior leads to a dangerous over-reliance on automated systems.
Conclusion and Future Work
The findings suggest that AI reliance is not just a matter of system accuracy, but a reflection of the user's psychological profile. Younger participants and those with lower education levels tended to score higher in buckpassing, indicating they may be more at risk. The authors propose designing "adaptive" AI interfaces that detect a user's decision-making style and provide "cognitive nudges"—design features that encourage critical thinking—to those prone to blind reliance.
Summary updated