Appendices A-E: Bias in AI Video Interview Evaluations: Effects of Autistic-Associated Behaviors and Debiasing Procedures

Adults with autism spectrum disorder (ASD) experience employment disparities despite evidence that they possess strengths valued in the workplace. One barrier is the interview, where interviewer impressions are often formed from non-verbal behavioral cues. Differences in these behaviors are common in ASD and contribute to negative evaluations. While the use of artificial intelligence (AI) in hiring practices is rapidly increasing, it remains unclear whether these behaviors bring out inherent biases within these systems. The current study examines whether AI Large Language Models (LLM) assign lower evaluative ratings to candidates displaying autistic-associated behaviors and whether a debiasing procedure can reduce these effects. Three AI LLM models rated candidates in mock interview clips across four conditions: control (neurotypical behavior), limited eye contact, monotone speech, and low conversational reciprocity. The AI LLM rated the candidate on interpersonal and cognitive attributes and overall hireability before and after being exposed to pro-neurodiversity research. The LLMs assigned lower ratings in all non-control conditions. After the debiasing procedure, ratings increased across conditions. These findings suggest that AI LLM systems may replicate bias toward neurodivergent applicants but also suggest that these biases can be mitigated.