It is commonly assumed that there are qualitatively distinct cognitive strategies that underlie decision making. Because cognitive strategies differ in how information is processed, predecisional information search offers a window onto these strategies. Using a bottom-up approach, we examine whether predecisional information search actually reflects the use of distinct strategies. Specifically, we investigate the extent to which the heterogeneity in people's predecisional information search in a risky choice task reflects qualitatively distinct patterns that should emerge when people use distinct strategies. Our analysis takes into account the distribution of attention across attributes and transitions between attributes. Using cluster analysis, we find just two qualitatively different clusters with low separability: one characterized by balanced attention to all attributes and by transitions occurring mostly within the same option, and one characterized by a focus on outcome information and by frequent attribute-wise transitions. These two clusters were also associated with differences in people's choice behavior. The distribution of these clusters varied considerably across individuals, but less so across choice problems, suggesting that information search is not necessarily guided by features of the choice problem—this result challenges current theories on strategy selection. Our results challenge the common assumption that heterogeneity in predecisional information search is differentiated along clearly distinct information processing policies. Instead, the differentiation seems to fall into just two broad clusters—one resembling rational principles of expectation computation, the other reflecting heuristic principles that neglect probabilities—with considerable variability within each cluster.

The assumption of an inverse S-shaped probability weighting function allows cumulative prospect theory to explain several well-established regularities in risky choice between monetary lotteries. Empirical evidence indicates that in choices between options with nonmonetary outcomes, the shape of the weighting function is strongly influenced by the negative emotions often associated with these outcomes. In its current form, however, cumulative prospect theory is silent with respect to how to formally integrate the influence of affective processes on the shape of the weighting function. Here, we propose an affective probability weighting function in which the two main features of the weighting function, probability sensitivity and elevation, gradually change with the affective value of the nonmonetary outcomes. We test our proposition in a model competition with three data sets. The results show that the affective probability weighting function improves the ability of (cumulative) prospect theory to predict choices between options with nonmonetary outcomes. We observed approximately linear probability weighting for the least affective nonmonetary outcomes and probability neglect for the worst or multiple outcomes. These findings demonstrate that integrating the effect of affective processes in formal decision models is crucial for advancing the understanding of choices between nonmonetary risky options---and thus ensuring the generalizability of the models beyond choices between monetary lotteries.

Nonlinear probability weighting allows cumulative prospect theory (CPT) to account for seminal phenomena in riskychoice (e.g., the certainty effect). The attentional drift diffusion model (aDDM) formalizes that attentional biases canshape preferences as a sequential sampling process. We simulated choices between safe and risky options using the aDDMwith varying attentional biases to safe or risky options and modeled these choices with CPT. Changes in preferences dueto attentional biases were systematically reflected in the parameters of CPT’s weighting function (curvature, elevation).We demonstrate that this also holds empirically, in the sampling paradigm in decision from experience. Hence, nonlinearprobability weighting can arise from option-specific attentional biases in information search. This challenges commoninterpretations of probability-weighting parameters, suggests novel attentional explanations for empirical phenomena as-sociated with characteristic shapes of CPT’s probability-weighting function, and adds to the integration of two prominentcomputational frameworks for decision making.

Decision makers can infer social statistics (e.g., the relative frequency of health risks or consumer preferences in the population) by drawing on samples from their personal social networks. In light of the growing use of the Internet, much of people’s social interactions occur online (e.g., via social media) rather than offline (e.g., via face-to-face contact). Here, we examine to what extent sampling of social network members from memory (social sampling) is affected by whether one usually has online vs. offline contact to a person. In our study, participants judged the popularity of holiday destinations and recalled people in their own online and offline social networks who had vacationed at each destination. Additionally, participants indicated the respective contact mode (offline, online, or mixed) and social category (self, family member, friend, or acquaintance) of each recalled person. We used a hierarchical Bayesian modeling approach to contrast two variants of a cognitive model that assumes sequential and limited search—the social-circle model. The variants assumed the search process underlying social sampling to be guided by either contact mode (online vs. offline) or social category. The model comparison further included an exhaustive sampling strategy and guessing. The majority of participants was best described by a limited rather than an exhaustive search strategy or guessing. Additionally, more than a third of participants were best described by the variant of the social-circle model assuming search to be guided by contact mode. Interestingly, participants who followed this search strategy also relied more strongly on their own experiences than participants who probed their memory by social category. Overall, these results provide the first evidence that contact mode affects social sampling from memory.

Although process data indicate that people often rely on sim-plifying processes when choosing between risky options, cur-rent models of heuristics cannot predict people’s choices veryaccurately. To address this apparent paradox, it has been pro-posed that people might adaptively choose from a toolboxof simple strategies. But which strategies are contained inthis toolbox? And how do people decide when to use whichdecision strategy? Here, we develop a model according towhich the decision maker selects a decision strategy for a givenchoice problem rationally from a toolbox of strategies; the con-tent of the toolbox is estimated for each individual decisionmaker. Using cross-validation on an empirical data set, we findthat this model of strategy selection from a personal adaptivetoolbox predicts people’s choices better than any single strat-egy (even when it is allowed to vary across participants) andbetter than previously proposed toolbox models. Our modelcomparisons show that both inferring the content of the tool-box and rational strategy selection are critical for accuratelypredicting people’s risky choices. Furthermore, our analysisreveals considerable individual differences in the set of strate-gies people are equipped with and how they choose amongthem; these individual differences could partly explain whysome people make better choices than others. These findingsrepresent an important step towards a complete formalizationof the notion that people select their cognitive strategies froma personal adaptive toolbox.