Research

Working Papers

• Factors Affecting Perceptions of Random Events

Abstract: The Hot Hand and Gambler's fallacies are both well established biases in probabilistic reasoning, referring to incorrect expectations of positive and negative auto-correlation between outcomes. If both of these opposing patterns of bias arise from the same underlying error in reasoning, what causes one to take precedence over the other? In existing experimental and empirical work, the variety of methods and designs makes it difficult to compare results across studies. In this paper, we report the results of an experiment in which we ask subjects to predict coin tosses after observing previous outcomes, varying both the question used to elicit their predictions and the visual presentation of the past outcomes. We find the same subjects alternating between the Hot Hand and Gambler's fallacies across transparently equivalent treatments. When asked to predict individual outcomes, subjects act in accordance with the Gambler's Fallacy. When asked to state their beliefs as probabilities, they favor the Hot Hand. This presents a challenge to understanding biased reasoning about probability as it implies that people use different subjective models to answer theoretically identical questions.


• Investigating Trust using Natural Language Processing

Abstract: Communication matters in economic decision making, even in games with simple, unique equilibria. However, the exact mechanism through which this effect occurs remains uncertain. Previous explanations have only been able to handle ad hoc hypotheses due to data limitations. In this paper, we employ Natural Language Processing on data from a large-scale principal-agent game run on Amazon’s MTurk system with one-way communication from agents to principals. NLP methods, combined with our data set of over 1000 messages, allow us to approach the data with minimal substantive assumptions to investigate the mechanisms underpinning trust. The structure of our data-set clearly links principals’ choices whether to trust agents to their received messages, allowing us to identify new features that increase the perceived trustworthiness of a message. Additionally, knowing the ultimate decisions of the agents, we also identify features of messages which predict that they are likely not trustworthy. Comparing the features that predict perceived and actual trustworthiness yields a set of features that predict which messages are likely to be both deceptive and effective. These laboratory findings may help identify the types of online interactions where there is a higher likelihood of deception.


• Effort-cost and Communication

Abstract: Cheap talk may come in many different forms that are not equally cheap. Different signals may require different costs to the sender but still carry identical information to the receiver. We report the results of an experiment designed to test whether promises which require higher levels of effort result in greater trust from their recipients. We find that more costly promises lead recipients to trust more frequently. However, there is no corresponding, significant difference in the trustworthiness of their senders. Further, when asked their beliefs explicitly, recipients do not believe that higher cost promises are more likely to be trustworthy. This challenges the usual interpretation of the receiver's choice in trust games with communication, wherein their choice is based solely on their own monetary payoff maximization given their beliefs about the sender's likely action. Their trust is, at least partially, other regarding. We conduct a second experiment to further test this result, using a game in which the sender's message cost has no possible influence on their decision. The results of this second experiment are inconclusive.

auedbr@vt.edu

PhD Graduate
Economics Department
Virginia Tech
Blacksburg, VA