Empirical Foundations and Computational Models of Unsystematic Errors in Choice Behavior

A large part of judgment and decision-making research aims at explaining behavior that is deviating from the standard economic perception of utility maximization. Following a long tradition, based on Herbert Simon’s notion of the human’s bounded rationality (1956), researchers developed theories to explain and predict systematic deviations from utility maximization (e.g. Kahneman & Tversky, 1979; Loewenstein & Prelec, 1992; Tversky & Kahneman, 1992). The insights gained by understanding decision making as an adaptive selection among different strategies and heuristics (Gigerenzer & Todd, 1999; Payne, Bettman, & Johnson, 1988) or the development of concepts, such as risk aversion (Pratt, 1964), loss aversion (Kahneman & Tversky, 1979), and hyperbolic discounting (Loewenstein & Prelec, 1992), for instance, have been shown to be of major relevance for society. Increasingly, under the term behavioral economics, this line of research has influenced governmental policy making (Madrian, 2014; Shafir, 2013).An important tool in this stream of research is the use of computational models that relate the observable data to mathematically formalized theoretical assumptions. Computational models aim at inferring cognitive processes from data and at quantifying variations of these processes by means of free parameters on continuous scales (Farrell & Lewandowsky, 2010). These inferences build on the assumption that the parameter estimates are valid and reliable. For a large group of models applied in judgment in decision-making research-mainly derivations from the standard utility model-this assumption was recently challenged. Bhatia & Loomes (2017) suggest that the presence of unexplained noise in the data can systematically distort the parameters. This points to the importance of distinguishing systematic errors and biases from unsystematic noise and errors in the cognitive process, which is essential to judgement and decision-making research as well as to cognitive sciences, in general. While systematic errors allow the prediction of behavior and can be applied in practice, unsystematic errors dilute their predictive accuracy and, thus, reduce the effectiveness of an application in practice. For instance, in an attempt to steer choice behavior, inferences on systematic deviations might be used as a guidance for the design of choice environments (e.g. Thaler et al., 2010). This application necessitates that the systematic deviations are reliably distinguished from unsystematic errors. To address this important issue, I propose a research project to study unsystematic errors in decision-making regarding their influence on the inferences of underlying cognitive processes. Using a combination of experimental studies with process tracing and computational modeling, I strive to identify the consequences of errors at different stages of the decision-making process. By model fitting, simulation, and parameter recovery studies, I illustrate their influence on parameter estimates-particularly, if the errors remain unexplained in computational models. Together, these results lend to the development of an error model that covers the most relevant errors in the decision-making process and enable a more reliable distinction between error parameters and other parameters in the model. The present proposal advances fundamental research on decision making by creating a deeper understanding of errors in decision making and provides, at the same time, a theoretically funded tool for application in practice.