The First Step of Reciprocity: Bidirectional Social Influence Intact in ASD. a Computational Psychiatry Study

Background: Autism originally means « egocentrism ». Despite recent examples of preserved social influence in ASD, the hypothesis that individuals with ASD might be less influenced by others (reduced social influence) and be more influenced by their own ideas when understanding others (strong projection bias) remains appealing

Objectives: In this study, we ask whether adults with ASD (AwASD) learn about and from others' pseudo-personality traits (impatience and prudence) similarly to typically developed adults (TDA). In particular, we want to investigate: (1)whether AwASD's projection bias is stronger than TDA’s, (2)whether learning others’ pseudo- traits is preserved in AwASD, (3)whether AwASD are immune to social influence for such pseudo- traits.

Methods: AwASD and TDA matched for age, IQ and gender took part to the study (n=48). AwASD had been assessed with ADOS-G and met DSM-IV criteria for an ASD. All participants had FSIQ>85. Participants with self-reported depression (Beck depression Inventory score>20) were excluded. In a series of trials, participants were presented with binary choices involving a trade-off between a cost (in terms of delay or risk) and a reward (high vs low). Participants had to successively make their decisions (block 1,3,5) and predict the choices of another (virtual) participants (block 2,4). Note that, in blocks 2 and 4, participants were given feedback to enable them to learn the virtual participant’s pseudo-trait. Based on their choices, we could estimate participants’ personality traits (impatience and prudence) in each decision block. In turn, we could manipulate the similarity of the virtual participant, which was endowed with similar (Same condition) or different (Different condition) pseudo-traits. We measured participants' projection bias by comparing their prediction performance in the Same vs Different conditions (prior to learning), and their sensitivity to social influence by estimating the drift in their trait after having tried to predict the choices of a Different other.

Results: Prediction accuracy at the beginning of the prediction blocks was higher in the Same condition than in the Different condition for both groups (AwASD:p=.003,TDA:p= 1e-4), indicating a significant projection bias that was otherwise not different between groups (p=.6). Second, at the end of the prediction phase, accuracy was good in both conditions and did not differed between groups (AwASD:85%,TDA:89%,p=.2).

Moreover, we found that both participants in both groups increase (resp., decrease) the frequency of prudent or impatient choices they make if they previously tried to predict somebody who was more (resp., less) prudent or impatient than themselves (p=.005). This resulted in an alignment of pseudo-traits that was not different between groups (p=.3).

Conclusions: This study is in line with recent results suggesting that social influence is not affected in ASD. For the first time, it also demonstrates that AwASD and TDA differ neither in terms of their underlying projection bias nor in their ability to learn others' pseudo-traits. Altogether, our results suggest that the first step of reciprocity (i.e. observational Theory of Mind) is intact in ASD. Further work remains to be done to understand how AwASD cope with reciprocal social interactions.