Artificially intelligent (AI) applications make data-driven predictions that enable personalization on a large scale. As such, recent advances in AI's predictive power might have the potential to create more productive work environments. Using a principal-agent model and understanding AI as...

Should principals explain and justify their evaluations? In this paper the principal's evaluation is private information, but she can provide some justifications by sending a costly message. Indeed, it is optimal for the principal to explain her evaluation to the agent if and only if the...

This paper introduces private sender information into a sender-receiver game of Bayesian persuasion with monotonic sender preferences. I derive properties of increasing differences related to the precision of signals and use these to fully characterize the set of equilibria robust to the...

Consider managers evaluating their employees' performances. Should managers justify their subjective evaluations? Suppose a manager's evaluation is private information. Justifying her evaluation is costly but limits the principal's scope for distorting her evaluation of the employee. I show that...

This paper models two key roles of subjective performance evaluations: their incentive role and their feedback role. The paper shows that the feedback role makes subjective pay feasible even without repeated interaction, as long as there exists some verifiable measure of performance. It also...

Should principals explain and justify their evaluations? Suppose the principal's evaluation is private information, but she can provide justification by sending a costly cheap-talk message. If she does not provide justification, her message space is restricted, but the message is costless. I...

Should principals explain and justify their evaluations? In this paper the principal's evaluation is private information, but she can provide justification by sending a costly cheap-talk message. I show that the principal explains her evaluation to the agent if the evaluation turns out to be...

We study a model of cheap talk with one substantive assumption: The sender's preferences are state independent. Our main observation is that such a sender gains credibility by degrading self-serving information. Using this observation, we examine the sender's benefits from communication, assess...

We show that firms may benefit from allowing some earnings management, because it can make noisy signals more informative. We model a firm that cannot observe a manager's cost of effort, her effort choice, and whether she manipulated a publicly observable signal. An optimal contract links...

Individuals use narratives as rationales or justifications to make their claims more convincing. I provide a general framework for partial verifiability based on narratives. Narratives give many reasons and arguments. The receiver derives the message’s meaning by aggregating these reasons; her...