Modeling Behavior in a Clinically Diagnostic Sequential Risk-Taking Task.

Abstract

This article models the cognitive processes underlying learning and sequential choice in a risk-taking task for the purposes of understanding how they occur in this moderately complex environment and how behavior in it relates to self-reported real-world risk taking. The best stochastic model assumes that participants incorrectly treat outcome probabilities as stationary, update probabilities in a Bayesian fashion, evaluate choice policies prior to rather than during responding, and maintain constant response sensitivity. The model parameter associated with subjective value of gains correlates well with external risk taking. Both the overall approach, which can be expanded as the basic paradigm is varied, and the specific results provide direction for theories of risky choice and for understanding risk taking as a public health problem.

Age-related characteristics of risky decision-making and progressive expectation formation

“…The basic properties of the BART implies a wide range of strategic processes that could possibly explain the behavior from the basic preconceptions of the decision maker to the extent the action-outcome representations can modulate the learning process [12]. Computational modeling works suggest that those models that take into account expectancy and outcome values (e.g., Bayesian Sequential Risk Taking Model) provide better fit with the behavior data instead of simple reinforcement learning models (e.g., the Target model) that do not rely on the sequential evaluation of gains and losses [12,13].…”

“…Computational modeling works suggest that those models that take into account expectancy and outcome values (e.g., Bayesian Sequential Risk Taking Model) provide better fit with the behavior data instead of simple reinforcement learning models (e.g., the Target model) that do not rely on the sequential evaluation of gains and losses [12,13]. Hence, the context of pumpoutcome pairs (i.e., action-outcome pairs) in the BART should be regarded in the light of extended action sequences and past experiences.…”

“…Each additional pump on the balloon increases the likelihood of balloon burst and losing the accumulated reward but deciding to go on pumping increases the chance to collect more reward. The experience gained through the repeated sequential decision making process ensure the step-by-step acquisition of the reward structure of the task and supports progressive expectation formation about the outcome distribution.The basic properties of the BART implies a wide range of strategic processes that could possibly explain the behavior from the basic preconceptions of the decision maker to the extent the action-outcome representations can modulate the learning process [12]. Computational modeling works suggest that those models that take into account expectancy and outcome values (e.g., Bayesian Sequential Risk Taking Model) provide better fit with the behavior data instead of simple reinforcement learning models (e.g., the Target model) that do not rely on the sequential evaluation of gains and losses [12,13].…”

Decision from models: Generalizing probability information to novel tasks.

“…The utility of a specific monetary reward was modeled as a power function with parameter α >0 (Luce, 2000): $$ufalse(vfalse)=v^{\alpha }$$Pleskac (2008) has pointed out that the utility function and probability weighting function in the task of Wallsten et al (2005) were not simultaneously identifiable. We also found that α and the free parameters in some choice models could not be simultaneously estimated (see Supplemental Appendix C online for proof).…”

“…Our focus on model-based transfer also distinguishes the path lottery task from sampling-based decision tasks that have a similar risk structure, such as the Balloon Analogue Risk Task (BART, Lejuez et al, 2002; Pleskac, 2008; Wallsten, Pleskac, & Lejuez, 2005). In the minefield, every tiny step along the path incurred a fixed probability of triggering a mine and the participant need to make use of this constraint to infer the probability of survival for novel path lengths in the test phase.…”