When people are asked whether they like to take risks, their responses are typically consistent over time and predictive of real-world behavior. Hence, risk attitude can be regarded as a stable psychological trait (Frey et al., 2017). Yet, in behavioral risky choice tasks used in psychological and economic research—such as choices between lotteries, abstractly described in terms of outcomes and probabilities—behavior often varies considerably across measurement time-points and formats of the task (Frey et al., 2017; Pedroni et al., 2017). It seems paradoxical that decisions in these situations—which try to condense the problem of decision making under risk to its essential parts—are rarely an expressions of a person’s stable, latent risk attitude. This dissertation examines why experimental risky choice behavior can be notoriously hard to predict, and how the methodological and theoretical apparatus with which we approach the study of risk preferences shapes the inferences we can make.
In the first chapter I introduce major theoretical perspectives on decision making under risk and the methods their proponents rely on. The notion of constructed preferences (Lichtenstein & Slovic, 2006; Slovic, 1995) is introduced as a general framework for understanding the lack of temporal stability and convergent validity of behavioral measures of risk attitude. According to this framework, behavioral risk preferences may be constructed on the spot, in the light of available cues and processing capacities. Hence, features of the choice environment—which have nothing to do with risk itself—and psychological characteristics of the decision maker—besides dispositional risk attitude—may profoundly shape the process and output of preference construction. In the subsequent chapters I investigate how surface features of stimulus materials, and individual differences in psychological characteristics, as well as their interplay, shape risky choice behavior. I also use different approaches of computational modeling to describe and explain these changes in risky choice and the underlying cognitive processes. In chapter 2 I demonstrate that in choices between a risky and a safe option, apparent age differences in risk attitude crucially depend on whether the options differ in complexity, rather than on age differences in latent risk attitude. In chapter 3 I investigate whether differences in option complexity also shape (age differences in) tasks used to measure framing effects, loss aversion, and delay discounting. This experiment identifies boundary conditions for the effects of option complexity. In chapter 4 I turn from focusing predominantly on behavior and its dependence on the anatomy of the task towards underlying cognitive processes. I demonstrate that risky choice behavior is shaped by differences between younger and older adults in the ability to implement selective attention. In chapter 5 I demonstrate why it may be useful to view risky choice through the lens of different formal theories—both economic and psychological ones—by identifying systematic signatures of attentional biases simulated in the attentional drift diffusion model in the parameters of cumulative prospect theory.
Overall, this dissertation shows why decision making under risk cannot be comprehensively understood in terms of latent risk attitude alone. It identifies specific contextual (option complexity) and psychological (selective attention) determinants of risky choice behavior which need to be taken into account as well, and explains how they affect the underlying process of preference construction, using computational modeling. Moreover, this work underlines the merits of theoretical and methodological pluralism for studying the variable, context-sensitive aspects of risky choice behavior and individual differences therein.