Excuse-seeking behavior that facilitates replacing altruistic choices with selfinterested ones has been documented in several domains. In a laboratory study, we replicate three leading papers on this topic: Dana et al. (2007), and the use of information avoidance; Exley (2015), and the use of differential risk preferences; and Di Tella et al. (2015), and the use of motivated beliefs. The replications were conducted as part of a graduate course, attempting to embed one answer to the growing call for experimental replications within the pedagogic process. We fully replicate the simpler Dana et al. paper, and broadly replicate the core findings for the other two projects, though with reduced effect sizes and a failure to replicate on some secondary measures. Finally, we attempt to connect behaviors to facilitate the understanding of how each fit within the broader literature. However, we find no connections across domains.

We compare three populations commonly used in experiments by economists and other social scientists: undergraduate students at a physical location (lab), Amazon's Mechanical Turk (MTurk), and Prolific. The comparison is made along three dimensions: the noise in the data due to inattention, the cost per observation, and the elasticity of response. We draw samples from each population, examining decisions in four one-shot games with varying tensions between the individual and socially efficient choices. When there is no tension, where individual and pro-social incentives coincide, noisy behavior accounts for 60% of the observations on MTurk, 19% on Prolific, and 14% for the lab. Taking costs into account, if noisy data is the only concern Prolific dominates from an inferential power point of view, combining relatively low noise with a cost per observation one fifth of the lab's. However, because the lab population is more sensitive to treatment, across our main PD game comparison the lab still outperforms both Prolific and MTurk.

In repeated-game applications, where both the collusive and noncollusive outcomes can be supported as equilibria, it is crucial for researchers to understand when each equilibria is more likely to result. Controlled experiments have provided a selection criterion for the two-player repeated prisoner's dilemma. The theoretical selection criterion uses game primitives to measure the set of beliefs such that an agent prefers to defect in all rounds relative to cooperating conditionally. The set of such beliefs is large when uncertainty with respect to what others will do--strategic uncertainty--is large, and the prediction is that a noncooperative outcome would emerge. In this paper, we experimentally test this model of selection and its underlying mechanism of strategic uncertainty by manipulating the total number of players. Our results affirm the model as a tool for predicting when tacit collusion is likely/unlikely to be successful. Extending the analysis, we corroborate the findings and mechanism of the model, in alternative settings, and for the decisions of non-human AI agents.

We demonstrate potential pitfalls when extrapolating behavioral findings identified from one-shot choices to repeated settings. As a case study, we examine the use of "regret lotteries" as a behavioral tool to boost motivation. Based on findings from one-shot settings, presenting counterfactual information generates the potential for regret, which can be used to increase a lottery's value. This result has motivated the increasing use of regret lotteries in the field to incentivize recurrent decisions like exercise and compliance with company directives. Using a controlled experiment we show that while regret lotteries are the superior motivational tool in one-shot decisions, for repeated decisions the effect is entirely reversed. These findings have implications for incentive and policy design, highlighting the scope for error when extrapolating one-shot findings to inherently repeated settings.

I develop a novel model of group-based communication in which group members communicate with one another. Communication is costly in the sense that group members who choose to send or listen to messages incur costs. Equilibrium strategies have an intuitive characterization - those with the best information send, those with the worst information receive. Free-riding leads to less information exchange than is optimal, but a simple system of transfers and subsidies can correct this. Examining the model's predictions with an experiment I find that subjects over-communicate when costs are high, but fail to benefit from this as much as they should. Additionally, I find that listening costs are more harmful to fare, in contrast with the theory which indicates sending costs.

We outline a mechanism for eliciting probabilities using two uniform random numbers that is equivalent to the binarized scoring rule (BSR). Though our implementation is simple to describe and has a non-mathematical explanation, it retains the desirable theoretical features of the BSR. Moreover, we show that a discretized version with evenly-spaced reporting intervals can be implemented in the field with no more equipment than a pair of dice.