ASD-Chat: An Innovative Dialogue Intervention System for Children with Autism based on LLM and VB-MAPP

TL;DR

Behavioral performance and physiological signal analysis results indicate that the ASD-Chat system achieves competitive intervention effects comparable to those of professional interventionists, indicating that the dialogue paradigm and prototype system designed have the potential for long-term intervention in the future.

Abstract

Early diagnosis and professional intervention can help children with autism spectrum disorder (ASD) return to normal life. However, the scarcity and imbalance of professional medical resources currently prevent many autistic children from receiving the necessary diagnosis and intervention. Therefore, numerous paradigms have been proposed that use computer technology to assist or independently conduct ASD interventions, with the aim of alleviating the aforementioned problem. However, these paradigms often lack a foundation in clinical intervention methods and suffer from a lack of personalization. Addressing these concerns, we propose ASD-Chat, a social intervention system based on VB-MAPP (Verbal Behavior Milestones Assessment and Placement Program) and powered by ChatGPT as the backbone for dialogue generation. Specifically, we designed intervention paradigms and prompts based on the clinical intervention method VB-MAPP and utilized ChatGPT's generative capabilities to facilitate social dialogue interventions. Experimental results demonstrate that our proposed system achieves competitive intervention effects to those of professional interventionists, making it a promising tool for long-term interventions in real healthcare scenario in the future.

Figures (6)

• Figure 1: The Workflow of ASD-Chat. The complete flow of the conversation is divided into three stages: Preparation, Intervention, and Collection. In the Preparation stage, the child's basic information and preferences are inputted, and after selecting a virtual avatar, as a kind male lion is chosen and shown here, the Intervention stage begins. There will be five topic dialogue scenarios based on the VB-MAPP in the Intervention stage. ASD-Chat greets the child and initiates the formal dialogue intervention. Based on the child's responses, ASD-Chat provides encouragement or praise and then leading to a new round of dialogue. When a time-out occurs, ASD-Chat bids farewell to the child, marking the end of a dialogue topic. For Collection, data is collected synchronously during the conversation and saved sequentially when conversations finish.
• Figure 2: Template of system prompt used by ASD-Chat. This prompt is inputted into ASD-Chat at the beginning of the conversation, allowing it to understand the basic information about the child, the current topic context, and its basic conversational restrictions. $Prompt_{info}$ indicates the child's preference such as food and color. $Prompt_{topic}$ indicates the detail description of current topic.
• Figure 3: The effectiveness of the ASD-Chat system in social intervention scenarios is evaluated from two aspects: Engagement and Quality. Engagement is measured by the average length of pure text and speaking duration per dialogue round, while Quality is assessed based on the semantic coherence in each dialogue round.
• Figure 4: The interpolation of the subject-averaged HbO amplitude for each channel across all subjects. Both scenarios have strong brain activations in different channels, but the magnitude of this difference is around 0.1, meaning that the two scenarios have similar brain activation effects.
• Figure 5: The average amplitude of all channels in each topic conversation for both scenarios of Subject 9. The ASD-Chat scenario has stronger brain activations for most topics (food, animal, family, and color). Although the toy topic of the interventionist scenario evokes larger amplitude changes, the interpolation between the two is small, only 0.31, which means that the activation strength is similar.