Beyond labels: Realism and participatory design as keys to effective AI-based neurodiversity training in the workplace

This study investigates the efficacy of AI-generated, gamified interventions in improving interactions between autistic and non-autistic adults. Grounded in the “double empathy problem,” the research tested whether brief cognitive interventions, such as diagnostic disclosure or behavioural descriptions, could shift first impressions and increase the willingness of participants (N = 86) to engage with autistic peers. Our study reveals that realism is a key component in AI-generated scenario-based interventions for cross-neurotype interactions, and provides practical strategies to be implemented when designing a similar intervention for workplace training. The findings suggest that digital inclusion training must move beyond simple information-sharing. Organisations should prioritise participatory design, involving neurodivergent individuals in co-creating training materials to ensure authenticity. To be effective, technology-driven training must bridge the “uncanny valley” by focusing on high-fidelity, emotionally authentic simulations that encourage neurotypical peers to adapt their own communication styles, rather than expecting autistic individuals to conform to neurotypical norms.