Evidence for a Risk Averse Decision-Making Style in Autism Spectrum Disorders

Background: Decision-making tasks allow researchers to examine two inter-related processes that are essential to our daily lives: learning and motivation. These tasks assess these processes in the context of non-social stimuli, with reference both to positive and negative outcomes. The Iowa Gambling Task (IGT) has been used to study decision-making in a number of clinical groups, including those with Autism Spectrum Disorder (ASD). Johnson et al. (2006) and Mussey et al. (2008) reported impaired decision-making performance in high functioning individuals with ASD. Given the complexity of the IGT, formal cognitive models have been developed to separate complex behaviour into several component constructs or processes (i.e., new variables) that cannot be derived based on observable behaviour. In our previous study (Johnson et al., 2006) we utilized the Expectancy-Valence Learning model (EVL, Busemeyer et al., 2002) to examine IGT performance and found high attention to loss in the ASD group relative to controls. Additional decision-making tasks, also amenable to formal cognitive modeling, have been used in other clinical groups. Specifically, the Balloon Analog Risk Task (BART, Lejuez et al., 2002) and Game of Dice Task (GDT, Brand et al., 2004) may provide further insight into decision-making processes in ASD.
Objectives: To determine if previously reported differences on decision-making tasks in ASD are task-specific (IGT only) or task-general, we examined the consistency of decision-making results across three tasks.
Methods: Current analyses represent data from 14 youths with ASD (10M/4F) and 16 age- and IQ-matched controls (10M/6F), aged 9-17 years; data collection is ongoing. All participants completed a battery of decision-making tasks including the IGT, BART, and GDT. In addition to traditional group comparisons of task performance, we utilized formal cognitive models to assess IGT and BART performance.
Results: There was a significant difference between groups on the BART (t (27) = 2.54, p = .017), with the ASD group making fewer risky choices than controls. Mathematical modeling results for the BART indicated group differences on two parameters: 1) the choice consistency parameter, with the ASD group exhibiting more consistent responses than the TD group; and 2) the sensitivity to rewards parameter, with the ASD group showing lower sensitivity to rewards relative to TD participants. Together, these findings suggest that ASD participants were less motivated by reward and demonstrated less exploratory, or risk-averse, response patterns. Consistent with BART findings, the ASD group made fewer risky choices on the GDT than controls, t (28) = -1.95, p = .06. In contrast to previous findings, there were no group differences on IGT performance, largely because of an unexpectedly flat learning slope for the TD group. There were no group differences on the EVL model of the IGT.
Conclusions: Overall, these findings suggest that children and adolescents with ASD demonstrate a risk-averse decision-making pattern across tasks. Results will be discussed in the context of the preference for sameness and aversion to change often associated with the autism spectrum.