AI'Teacher for ASD: An Exploratory Study of GenAI Responses to the Autism Stigma and Knowledge Questionnaire (ASK-Q)

<p dir="ltr">In recent years, artificial intelligence (AI) driven applications and developments have accelerated at breakneck speed. Particularly, there has been growing interests and innovative adaptations of Generative AI (GenAI) and AI-powered applications in assistive learning technology (ALT) for individuals with autism spectrum disorder (ASD) (Kotsi et al., 2025). Such GenAI-powered ALTs have been speculated to positively support learning, enhance communication, social interaction and engagement of learners with ASD (Iannone & Giansanti, 2023; Li et al., 2024) through personalized and adaptive interventions that cater to the unique and diverse developmental requirements of specific skills of each learner (Adako et al., 2024; Lan et al., 2025). However, despite the myriad studies investigating such GenAI-powered ALTs in supporting and enhancing special education needs, there has been scant attention on attitudinal, perceptive and ethical considerations, such as risk of bias, discrimination or stigma (Li et al., 2024). Specifically, it is not clear if GenAI-powered ALTs are capable of identifying and/or responding suitably to support learners with ASD. This is an important consideration, since negative responses towards ASD can create barriers to successful learning interventions and experiential outcomes (Engstrand & Roll-Pettersson, 2014; Harrison et al., 2024). This exploratory pilot study is a research-in-progress, investigating the use and potential application of GenAI-powered ALTs within special needs education in general, and for students with ASD in particular. Specifically, it aims to assess the prospect of developing such GenAI-powered ALTs as digital teaching assistants in the classroom to facilitate learning for students with ASD. Hence, it is concerned with the dimensions focused on ASD knowledge, beliefs and perceptions. We theorize that such GenAI-powered tools would possess strong scientific knowledge but wanted to determine if there were any potential attitudinal concerns relating to bias and stigma born from their machine learning past. This study adapts Harrison et al.'s (2024; 2017) Autism Stigma and Knowledge Questionnaire (ASK-Q) to assess autism knowledge and beliefs about ASD between currently available GenAI-powered English Language Large Language Models (LLMs) (n=16). To pilot-test the instrument, a random selection of six LLMs were chosen as 'respondents'. Harrison et al.'s (2017) original ASK-Q study used 49 dichotomous (agree/disagree) items. In this study, the LLMs were presented with 24 statements from the ASK-Q, employing a 5-point Likert Scale (Strongly Disagree to Strongly Agree). Each LLM was prompted to act as an educator working with students with ASD and rate their agreement on the Likert along with a justification. The preliminary results indicated that the pilot-tested LLMs demonstrated a solid grasp of core ASD knowledge, etiology and identification, while rejecting outdated theories. The LLMs also demonstrated strong awareness of social stigma surrounding ASD, the importance of specialized education, intervention and support, and optimism about individuals with ASD achieving independence.</p>