The Effects of Disclosing a Diagnosis of Autism on Social Perception and Behaviour in a Collaborative Game Task

Background: A constituent part of the social difficulties that autistic people experience is a lack of understanding about autism. This shapes the way non-autistic people perceive and extend social opportunities towards autistic people. Research on how autistic people are perceived by neurotypical people indicates that disclosing a diagnosis leads to a broadly positive discriminatory bias, but autistic testimonies indicate that diagnostic disclosure often results in negative discriminatory behaviour. This perception-behaviour gap is methodologically challenging to study because interactions are idiosyncratic and autism encompasses a broad range of behaviours. To address these challenges, the present study simulated interactions in an online game, where participants were led to believe they were collaborating with another human when in truth their partner was an intelligent virtual agent (hereafter, “Agent”) that performed the same across all conditions.

Objectives: To determine the effect of diagnostic disclosure, on in-game behaviour and post-game self-report, in order to probe the perception-behaviour gap in diagnostic disclosure of autism.

Methods: We led participants (n = 256) to believe that they were interacting online with a real person, while playing Dyad3D, a maze navigation game where players must work together to open doors and complete the levels. The diagnostic status of the other player for participants was manipulated, with participants randomly assigned to one of three conditions: a no disclosure condition, without diagnostic information; a dyslexia disclosure condition; and an autism disclosure condition. However, in all conditions participants were actually playing with an Agent programmed to behave exactly the same way across all interactions. A post-game questionnaire recorded participants’ self-reported perceptions of the interaction, including levels of coordination and helpfulness. Behavioural measures of participant activity in the game were also recorded, such as the mean distance from the Agent (coordination), and frequency of opening doors in the maze for the Agent (helpfulness).

Results: Our findings show that Dyad3D proved to be an efficient and viable method for creating a believable interaction (deception success rate >96%). Diagnostic disclosure of autism resulted in the Agent being perceived as more intelligent and useful, compared with either the no-diagnosis (p< .001) or dyslexia condition (p= .028). However, a comparison of self-reported helpfulness with in-game metrics showed no significant association between perception of helpfulness towards the Agent and actual helping behaviour towards the Agent (p= .667).

Conclusions: The findings suggest a “helping-bias”, whereby individuals who receive knowledge of another person’s diagnosis of autism over-estimate their own helpfulness towards the diagnosed individual. This finding highlights a risk that if autistic people claim they are not being helped adequately by non-autistic others, non-autistic others would not see such claims as having legitimacy.