Information Processing in Autism Is Less Impacted By Increasing Decision Complexity

Background: Recent studies have found that people with autism make more consistent and conventionally-rational decisions than do neurotypical (NT) adults (e.g., Farmer et al., 2017). However, little is known about the processes of information sampling and integration that underlie these patterns of behaviour. Studies of NT decision-makers have found that these processes can be illuminated by tracking eye-movements during the decision.

Objectives: To use eye-tracking to test for differences in the information processing that underpins autistic and NT decision-making.

Methods: 35 autistic adults were matched on age, gender, and IQ to 35 NT participants. The experimental task consisted of choosing an apartment to rent from a set presented in table format, where each apartment differed in several attributes (e.g., Rent, Cleanliness, Distance). We varied the number of apartments (N_apartments = 2,3,5,7) and number of attributes (N_attributes = 2,3,5,7) giving 16 combinations in a 4x4 design. The tables were presented on a screen while the participant’s gaze was tracked. We measured several variables including decision time, transition patterns (within apartment or within attribute), search depth (proportion of cells in the table inspected), and proportion of time spent inspecting the ultimately-chosen apartment.

Results: There was a main effect of the number of attributes such that more attributes resulted in slower decision times (F(1.77,120.37)=431.18, p<.01), reduced search depth (F(1.81,123.27)=253.23,p<.01), and an increased proportion of time spent fixating on the ultimately-chosen apartment (F(2.70,183.75)=51.18,p<.01). The same pattern was observed for the number of apartments on decision time (F(1.80,122.63)=453.96,p<.01), and search depth (F(1.78,120.96)=311.73,p<.01). Increasing the number of attributes led participants to transition more frequently from one cell to another within apartments (F(2.21,150.38)=99.27,p<.01), and increasing the number of apartments led participants to transition within attributes more often (F(2.03,137.87)=36.55,p<.01). There were no main effects of autism, and no interactions between autism and the number of apartments on any of the performance measures. However, there were significant interactions between autism and the number of attributes. The autism group’s decision times did not slow as much with increasing numbers of attributes (F(1.77,120.37)=9.61,p<.01). The autism group’s search depth fell more with increasing numbers of attributes (F(1.81,123.27)=5.91, p<.01), and, with increasing attributes, the autism group increased by more the proportion of time spent fixating the ultimately-chosen apartment (F(2.70,183.75)=4.83,p<.01).

Conclusions: This is the first study to use eye-tracking to examine the information processing underpinning multi-attribute decision-making in autism. There is an intriguing pattern suggestive of broadly the same strategy as NT decision-makers, but one that is less impacted by increasing the number of attributes that are available. One interpretation is that people with autism attach great weight to a small number of attributes, either as a simplifying heuristic or because those attributes are stronger, more valid cues to the utility of an option for this population.