Participants Over-Estimate How Helpful They Are in a Two-Player Game Scenario Toward an Artificial Confederate That Discloses a Diagnosis of Autism.

Research on how autistic people are perceived by neurotypical people indicates that disclosing a diagnosis leads to a positive discriminatory bias; however, autobiographical autistic accounts indicate that diagnostic disclosure often results in negative discriminatory behavior. We report on an exploratory study to compare people's self-reported helping behavior with their actual helping behavior toward an assumed autistic collaborator. We led 255 participants to believe that they were interacting online with a real person to play Dyad3D, a maze navigation game where players must work together to open doors, and complete the levels. However, participants were actually playing with an artificial confederate (AC) that is programmed to behave the same way across all interactions. This design enabled us to manipulate the diagnostic status of the AC that participants received prior to collaboration across three conditions: no disclosure, dyslexia-disclosure, and autism-disclosure. We use this method to explore two research questions: (1) is Dyad3D viable in creating a simulated interaction that could deceive participants into believing they were collaborating with another human player online? and (2) what are the effects of disclosing an autism diagnosis on social perception and collaboration? Combined with a post-game questionnaire, we compared differences between diagnostic conditions and differences between self-reported behavior and actual behavior in the game. Our findings show that Dyad3D proved to be an efficient and viable method for creating a believable interaction (deception success rate >96%). Moreover, diagnostic disclosure of autism results in the AC being perceived as more intelligent and useful, but participants also perceived themselves to be more helpful toward the AC than they actually were. We evaluate the strengths and limitations of the current method and provide recommendations for future research. The source code for Dyad3D is freely available (CC-BY-NC 4.0) so that the study is reproducible and open to future adaptation.