Title:Helping or Homogenizing? GenAI as a Design Partner to Pre-Service SLPs for Just-in-Time Programming of AAC

Abstract:Augmentative and alternative communication (AAC) devices are used by many people around the world who experience difficulties in communicating verbally. One AAC device which is especially useful for minimally verbal autistic children in developing language and communication skills are visual scene displays (VSD). VSDs use images with interactive hotspots embedded in them to directly connect language to real-world contexts which are meaningful to the AAC user. While VSDs can effectively support emergent communicators, their widespread adoption is impacted by how difficult these devices are to configure. We developed a prototype that uses generative AI to automatically suggest initial hotspots on an image to help non-experts efficiently create VSDs. We conducted a within-subjects user study to understand how effective our prototype is in supporting non-expert users, specifically pre-service speech-language pathologists (SLP) who are not familiar with VSDs as an AAC intervention. Pre-service SLPs are actively studying to become clinically certified SLPs and have domain-specific knowledge about language and communication skill development. We evaluated the effectiveness of our prototype based on creation time, quality, and user confidence. We also analyzed the relevance and developmental appropriateness of the automatically generated hotspots and how often users interacted with the generated hotspots. Our results were mixed with SLPs becoming more efficient and confident. However, there were multiple negative impacts as well, including over-reliance and homogenization of communication options. The implications of these findings reach beyond the domain of AAC, especially as generative AI becomes more prevalent across domains, including assistive technology. Future work is needed to further identify and address these risks associated with integrating generative AI into assistive technology.