ChatGPT's inconsistent moral advice influences users' judgment

ChatGPT is not only fun to chat with, but it also searches information, answers questions, and gives advice. With consistent moral advice, it can improve the moral judgment and decisions of users. Unfortunately, ChatGPT's advice is not consistent. Nonetheless, it does influence users' moral judgment, we find in an experiment, even if they know they are advised by a chatting bot, and they underestimate how much they are influenced. Thus, ChatGPT corrupts rather than improves its users' moral judgment. While these findings call for better design of ChatGPT and similar bots, we also propose training to improve users' digital literacy as a remedy. Transparency, however, is not sufficient to enable the responsible use of AI.

Effects of Social Information on Risk Taking and Performance: Understanding Others’ Decisions vs. Comparing Oneself with Others in Short-Term Performance

When a problem leaves decision makers uncertain as to how to approach it, observing others’ decisions can improve one’s own decisions by promoting more accurate judgments and a better insight into the problem. However, observing others’ decisions may also activate motives that prevent this potential from being realized, for instance, ego concerns that prompt excessive risk taking. Our experimental study investigates how two features of the social environment influence the effect of observing others’ decisions on individual risk taking and performance. We manipulated (1) the psychological distance to others whose decisions could be observed (and thereby the tendency to seek self-enhancing social comparison) and (2) the opportunity for interaction (and thereby for a cumulative effect of any such tendency on decisions over time and for an effect on social information itself). Because the two features covary in real-world settings, we designed two treatments corresponding to the two natural combinations. Both treatments provided participants with two other participants’ period decisions in a multiperiod problem under uncertainty. No new objective information about the problem could be inferred from these decisions. We predicted that participants who observed the decisions of distant others (who had solved the same problem earlier) would perform better than participants in a control sample without any information about others’ decisions and that participants who observed the decisions of proximal others (with whom interaction could arise) would take more risk and perform worse than those who observed distant others’ decisions. The data corroborate our predictions. We discuss implications for organizational learning.