Choice Avoidance in Critical Decisions: An Empirical Investigation with Implications for AI Development

Important decisions are often difficult to make. Consequently, people may defer decision-making to other agents when faced with difficult choices, even if the decision has significant implications for their own well-being. However, while sporadic evidence of choice avoidance exists in some contexts, we do not yet fully understand its prevalence and how it varies across different decision problems. To what extent are individuals willing to delegate decision-making authority in difficult decisions, such as determining who to harm in an unavoidable car accident or deciding which employee to promote when multiple candidates qualify? What psychological and contextual factors influence this willingness? These questions require urgent attention today as artificial intelligence (AI) systems rapidly expand into domains with profound policy and ethical implications.

Through a series of pilot-tested survey experiments informed by insights from political philosophy and behavioral sciences, this study aims to measure the prevalence of choice avoidance among Hong Kong citizens and elucidate the underlying factors that drive such behavior. Specifically, we will conduct experiments embedded in a probability-based online survey across four distinct contexts: autonomous driving, human resource management, public policy, and health insurance consumption. Participants will be randomly presented with decision scenarios varying in informational complexity, moral conflicts, and potential consequences of failure. We will then measure their willingness to delegate the decision to another agent, randomly presented as an AI system or a human being. To contextualize the experimental findings within respondents’ lives and explore the influence of institutional and cultural factors, we will conduct in-depth interviews with 60 participants.

This project’s findings will have significant theoretical and practical implications. Theoretically, while extensive research exists on the burdens that difficult decisions place on human beings, these insights remain isolated from studies examining humans’ willingness to delegate decision-making authority to another agent. As a result, existing evidence of humans’ desire for decision-making typically comes from questions that put respondents in a detached, receiving-end perspective, rather than confronting them with the burden of hard decisions. This project will bridge the literatures on decision fatigue and desire for decision and establish the necessary concepts and tools to evaluate choice avoidance. Practically, as AI systems gain increasing capacities, understanding when and how choice avoidance occurs is crucial for formulating policies that ensure the continued relevance of human judgments in our future development.