Bayesian Updating in the EEG : Differentiation between Automatic and Controlled Processes of Human Economic Decision Making

Abstract

Research has shown that economic decision makers often do not behave according to the prescriptions of rationality, but instead show systematic deviations from rational behavior (e.g., Starmer, 2000). One approach to explain these deviations is taking a dual-process perspective (see Evans, 2008; Sanfey & Chang, 2008; Weber & Johnson, 2009) in which a distinction is made between deliberate, resource-consuming controlled processes and fast, effortless automatic processes. In many cases, deviations from rationality such as violations of Bayesian updating (e.g., Fujikawa & Oda, 2005; Ouwersloot, Nijkamp, & Rietveld, 1998; Zizzo, Stolarz-Fantino, Wen, & Fantino, 2000) may be ascribed to rather automatic, intuitive processes. Conflicts between such intuitive strategies and Bayesian updating can be assumed to affect decision outcomes and response times, as well as electrocortical activity. Moreover, it is suggested that personality characteristics such as faith in intuition (Epstein, Pacini, Denes-Raj, & Heier, 1996) and event-related potentials of the EEG can explain much of the variance in decision behavior across participants.In order to shed light on these issues, the present dissertation investigated decision behavior and corresponding electrocortical activity in posterior probability tasks in which participants had to update prior probabilities on the basis of new evidence to maximize payoff, and in which an intuitive strategy sometimes conflicted with Bayesian calculations.In Study 1a the reinforcement heuristic (Charness & Levin, 2005) led to a high rate of decision errors, especially for people who trusted their intuitive feelings. Participants‟ proneness to this rather automatic heuristic was significantly positively associated with the amplitude of the FRN. When the reinforcement heuristic was not available (Study 1b), participants were better able to apply the controlled process of Bayesian updating. Interindividual heterogeneity was again reflected in the amplitude of the FRN. In Study 2, conflicts between the representativeness heuristic (Grether, 1980, 1992) and Bayesian calculations led to a high error rate and long response times and were associated with a strongly pronounced N2 amplitude. Individual N2 amplitudes reflected the extent to which participants were able to detect this conflict and consequently to suppress the automatic response. Further, the conservatism heuristic (Edwards, 1968) was positively associated with participants‟ LRP amplitude and with faith in intuition. On the whole, results of the studies support the relevance of the dual-process perspective to economic decision making. Moreover, findings strongly suggest that interindividual differences are highly relevant within this context, and prove that neuroscientific methods can provide a better understanding of the physiological basis of (boundely) rational decision making.