Heterogeneity of strategy use in the Iowa gambling task: A comparison of win-stay/lose-shift and reinforcement learning models

Abstract

The Iowa gambling task (IGT) has been used in numerous studies, often to examine decision-making performance in different clinical populations. Reinforcement learning (RL) models such as the expectancy valence (EV) model have often been used to characterize choice behavior in this work, and accordingly, parameter differences from these models have been used to examine differences in decision-making processes between different populations. These RL models assume a strategy whereby participants incrementally update the expected rewards for each option and probabilistically select options with higher expected rewards. Here we show that a formal model that assumes a win-stay/lose-shift (WSLS) strategy-which is sensitive only to the outcome of the previous choice-provides the best fit to IGT data from about half of our sample of healthy young adults, and that a prospect valence learning (PVL) model that utilizes a decay reinforcement learning rule provides the best fit to the other half of the data. Further analyses suggested that the better fits of the WSLS model to many participants' data were not due to an enhanced ability of the WSLS model to mimic the RL strategy assumed by the PVL and EV models. These results suggest that WSLS is a common strategy in the IGT and that both heuristic-based and RL-based models should be used to inform decision-making behavior in the IGT and similar choice tasks.

Adaptable history biases in human perceptual decisions

“…Choice-history biases have been observed when subjects had to make judgments about physical weights (9), auditory stimuli (12,16,18), or visual stimuli (4,8,10,11,14,18). As in our experiment, subjects show a diversity of biases ranging from switching strategies (4,9,12,14,16), to staying strategies (10,11,18), to success-stay/fail-switch strategies (5, 6, 8, 18).…”

“…These biases generally manifest as switch-after-failure A B or as stay-after-success strategies (8,10,11,18). Both types of biases were disadvantageous in that they reduced sensitivity for most of the subjects.…”

“…The optimal behavior, instead, is to respond randomly, because any strategy can be exploited by other players (7). People also resort to the suboptimal success-stay/fail-switch strategy in situations that would require more complex decisionmaking strategies (8). Thus, despite our ability in some situations to adapt to choice history statistics, in other contexts, we are unable to appropriately adjust behavior to account for choice history statistics.…”

Understanding Optimal Decision-Making in Wargaming

“…Although most studies of IGT have concentrated on its ability to distinguish between healthy and clinical populations, recent studies have focused on the variability in decision performance found among healthy participants (Steingroever et al, 2013;Worthy et al, 2012). Steingroever et al conducted a meta-analysis to determine if healthy participants consistently (1) select good decks to a greater extent than bad decks, and (2) go through a clear explo- Figure 2.7: Screen shot of the original Iowa Gambling Task by Bechara et al (2005).…”

Entrepreneurs’ Exploratory Perseverance in Learning Settings

“…Most IGT studies that explicitly provide such information (e.g., Bechara et al, 1999;Lehto & Elorinne, 2003;Worthy, Hawthorne, & Otto, 2013) identify a significant sub-group of healthy control participants (e.g., 23% of controls in Bechara et al, 1999;7.5% in Lehto & Elorinne, 2003) who exhibit impaired performance in the IGT, often similar and sometimes even worse than the performance of the patients (e.g., Bechara et al, 1999;Lehto & Elorinne, 2003). Although the behavior of these healthy underperformers resembled that of the patients, they differed in terms of the physiological results (i.e., SCR), which reflect emotional reactions.…”