Thinking Out Loud Blog

Learn more

‌

Artificial Intelligence as a Social Interface for Neurodivergents

Article by Katherine Eliza

5-10 minutes

It is common knowledge these days that many people are turning to AI chatbots for companionship. Though non-reciprocal, the dopamine-fuelled allure of a judgement-free venting session, for some, is far more appealing than human interaction – especially for those who struggle with it. Autism spectrum disorder is commonly associated with differences in social communication. This may include interpreting emotional cues, maintaining reciprocal conversation and understanding the ‘unwritten rules’ of society. Many autistic individuals resort to ‘masking’, which refers to an autistic person’s attempt to blend in with their neurotypical peers by mirroring behaviour, disguising autistic traits and tightly controlling expression to appear ‘normal’ (National Autistic Society, 2026). Exhausting, right?

Today, researchers are exploring the potential for these understanding AI chatbots as technological assistants for neurodivergent masking. One emerging idea is the use of AI as a form of “social interface”; a real-time guide during conversations – think Iron Man’s assistant Jarvis – or even a reflective listener providing feedback following interactions. Whilst this concept may initially appear speculative, a growing body of research suggests that these technologies are already being explored and may have meaningful potential.

Stanford Report (2025) developed an AI chatbot called ‘Noora’, aiming to coach individuals through empathic conversational responses. Study participants practiced responding empathetically to verbal cues, receiving feedback from the AI tool. After four weeks, strong improvements in real conversations were measured, suggesting the reflective use of AI communication tools was successful. Furthermore, it revealed that short term, AI-assisted intervention resulted in significant improvements in verbal empathy. These types of  conversational practice agents provide a low risk, self-paced, controlled environment to practice interaction without social pressure or the unpredictability and judgement of live interaction (Emergent Mind, 2024). But what about that unpredictability?  Studies suggest these tools could analyse conversational cues and suggest responses, help moderate tone in written communications or provide feedback following social exchanges. Large Language Models (LLMs) can provide live suggestions for empathetic responding in verbal autistic adults and adolescents. (Kogel et. al, 2025). Some researchers argue that AI assistance could potentially outsource the cognitive load of masking, functioning as a technological mediator (Hollis et al., 2026). 

However, along with rising potential comes the rise of significant ethical questions. Critics warn that dependency on these conversational partners may result in further withdrawal from human relationships, potentially reinforcing the isolation it was designed to alleviate (Neuroscience News, 2023). Then comes the glaring question of privacy and data protection. Tools designed to record, store and analyse real conversations inevitably process sensitive information. The tools’ ethical pillars must include robust security, informed consent and strict data protection, which may be tricky to integrate due to the very nature of the tool. Of course, the legal implications of covert audio recording speak for themselves (Bertacchini et al., 2023). 

The concept of an AI- assisted, social interface for autistic people sits on a crossroads between technological advancement, ethical consideration and accessibility. Designed properly, these tools could provide valuable reflection and guidance, diminishing daily anxiety and exhaustion, hence nurturing independence. The greatest consideration for development of these technologies is that they must take care to support, rather than replace human connection. Therefore, the challenges faced may be less technological, but philosophical. Should AI be geared towards moulding neurodivergent brains? Perhaps if society learned to accommodate communication differences, the task wouldn’t be getting outsourced to robots.

References

Pages: 1 2 3