Dynamics of information processing in risky choices : Making thorough use of behavioral data

Abstract

This work examines the dynamics of information use and integration of decisions under risk, in particular repeated monetary gambles, thereby advocating the value of a joint analysis of choice proportions and response times. Based on findings in perceptual decision making research that choice behavior is subject to systematic changes with response time (RT), and studies from judgment and decision making research suggesting that people apply different decision strategies across trials, this work aimed to investigate the extent to which changes in the application of decision strategies are systematically related to changes in RT. Furthermore, this work examined how different types of feedback (Research Paper I), the manipulation of the information format (Research Paper II), and the presentation order of choice options (Research Paper III) affect the pattern of information integration and risky choice behavior in general. Finally, as part of the latter research paper, both the applicability and the limits of a simple drift diffusion model to the analysis of risky choices are tested. The present thesis establishes that risky choice behavior is not always stable over time, but subject to systematic variations with RT, supporting the idea that people use different strategies across trials. In this regard, simpler decision rules and dominant choice preferences (e.g., for high probabilities lotteries or sure outcomes) are particularly pronounced for fast responses. Slower responses are characterized by either more integrative decision strategies (e.g., when conflicting information is presented), or by maintaining the initial strategy/heuristic (e.g., if the information format is difficult to process). In addition, a simple diffusion model can successfully account for a large portion of risky choice behavior, and parameter analyses provide useful insights in the underlying cognitive processes (e.g., a response bias for sure outcomes over lotteries in the gain domain). However, a limit is reached for cases in which decision strategies change with RT, for which more complex models are required. The results are reviewed in the context (and limits) of established theoretical frameworks, and an alternative explanation of the underlying mechanism is discussed.