RESEARCH

Abstract. In expert services, such as medical services, clients often lack specialized skills to diagnose their problems. However, it is common for them to participate in decision making through self-diagnosis, potentially leading to noncompliance against experts' advice. This study investigates how clients' active participation impacts experts' investment of effort in the precision of diagnosis. Through a theory-driven lab experiment, passive and active clients, who vary in whether they have an option to go against experts, consult experts for diagnosing unknown problems, and experts exert effort to improve the diagnostic precision. The results show that giving clients an option to go against experts' advice reduces experts' investment in improving the accuracy of diagnosis. This effect is particularly large among experts who prioritize clients' well-being. Additionally, providing clients with more information about experts, such as through online ratings, mitigates the negative effects of client participation. This study sheds light on the potential adverse effect of client involvement in expert services.

Abstract: Virtual assistants powered by artificial intelligence are present in virtually every aspect of daily life. Although they are computer algorithms, most are represented with humanized personal characteristics. We study whether assigning them a gender affects the propensity to delegate a search in two online experiments and compare it to human counterparts of identical characteristics. Virtual assistants generally receive higher delegation than humans. Gender has differential effects in delegation rates impacting the user’s welfare. The results are entirely driven by female subjects. We find mild spillover effects, primarily decreasing the selection of male humans after interacting with low-quality male virtual assistants.

Abstract: We exploit a unique longitudinal dataset of university students to study the stability of risk preferences over five years. We find that overall, subjects' risk tolerance measured by incentivized lottery choice increases over time, while decreases if measured by a self-assessed survey question. Moreover, we examine the impact of negative experiences and emotions on subjects' temporal change of risk preferences. We demonstrate that for the same group of respondents, the risk tolerance elicited by the incentivized measure is more robust while the survey measure is more sensitive in the face of negative shocks. Our findings contribute to a deeper understanding of how risk preferences evolve over time and underscore the significance of appropriate measurement methods when studying risk attitudes.

Abstract: Previous studies have shown that punishment opportunities can reduce free riding effectively in public goods production and that negative emotions toward free riders play an important role in precipitating punishment. By varying the timing of punishment in a public good game, we develop a novel punishment rule, the "Pre-Punishment" rule, which is designed to involve a lower level of emotional arousal compared to the classical "Post-Punishment" rule. We employed biometric measures (eye trackers and skin conductance response) in a lab experiment to capture the psychological responses, which will shed light on the mechanism mediating punishment behavior, the response to punishment, and the impact on cooperative behavior. Our results show that this new punishment rule works equally well in increasing contribution compared to the Post-Punishment rule. However, the biometric finding indicates that the effectiveness of Pre-Punishment rule does not rely on subjects' emotional arousal. This study provides useful suggestions for policymakers and managers for designing proper penalty rules to increase cooperation, and will also contribute to the public good game literature by uncovering the psychological processes underlying the effectiveness of punishment institutions.

We study the interaction between competition and social proximity on altruism, trust, and reciprocity, with an emphasis on gender differences. We decompose the behavioral channels by utilizing variants of both the Trust Game and the Dictator Game in a design that systematically controls the transmission of relevant information regarding the outcome of a previous competition and the social proximity between the competitors. On an aggregate level, our results suggest that knowledge of a victory over a socially proximate opponent leads to a large and significant increase in participants' investment in the Trust Game, but not in the Dictator Game. However, we also found substantial gender heterogeneity across treatments: men tend to show higher levels of altruism upon learning that their competitor shares similar characteristics with them. Moreover, this information yields a significant increase in men's trust when they are also told that they won the competition. In contrast, women's decisions about trust and altruism are not responsive to any of our treatment conditions. Our results provide helpful insights into the structure of incentives within institutions and companies, which are known to affect performance.