SocialWise: LLM-Agentic Conversation Therapy for Individuals with Autism Spectrum Disorder to Enhance Communication Skills

2026-04-20 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Medical Devices & Health Technology · Depth: Advanced, medium

Summary

SocialWise is a browser-based application designed to provide scalable conversation therapy for individuals with Autism Spectrum Disorder (ASD). It addresses the scarcity of affordable and effective support for practicing everyday communication skills, which currently relies on expensive in-person sessions or yields limited improvement from low-cost activities. The system pairs LLM conversational agents with a therapeutic Retrieval Augmented Generation (RAG) knowledge base, allowing users to select scenarios like "ordering food" or "joining a group." Users interact via text or voice, receiving instant, structured feedback on tone, engagement, and alternative phrasing. Implemented with Streamlit, LangChain, and ChromaDB, the prototype runs on any internet-connected computer. Preliminary user data from 34 autistic adults (ages 18-34) showed a perceived helpfulness of 4.15/5 and a 100% recommendation rate, with benefits including improved turn-taking, problem-solving in social conflicts, and reduced anxiety.

Key takeaway

For research scientists developing AI-driven therapeutic tools, SocialWise demonstrates a viable model for delivering scalable, evidence-based communication coaching. You should consider integrating LLM agents with retrieval-augmented generation (RAG) to provide personalized, grounded feedback, and prioritize multimodal interaction and user privacy. This approach can significantly broaden access to specialized therapy, as evidenced by SocialWise's high user satisfaction and perceived helpfulness.

Key insights

SocialWise offers scalable, LLM-agentic conversation therapy for ASD, providing instant, structured feedback via a browser-based platform.

Principles

• Combine LLM agents with RAG for grounded therapeutic feedback.
• Multimodal interaction (text/voice) enhances accessibility.
• Prioritize privacy by keeping data local to the browser.

Method

SocialWise uses a Streamlit web app, GPT-4o-mini for dialogue, Whisper STT, OpenAI TTS, and a LangChain RAG pipeline with ChromaDB to ground feedback in ASD therapy documents.

In practice

• Utilize GPT-4o-mini for conversational agents.
• Implement RAG with ChromaDB for evidence-based grounding.
• Design for browser-based, therapist-free accessibility.

Topics

• Autism Spectrum Disorder
• LLM Conversational Agents
• Conversation Therapy
• Retrieval-Augmented Generation
• Communication Skill Enhancement

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, NLP Engineer

Related on AIssential

• Psychological Digital Twins : Conceptual Models, AI Foundations and Ethical Challenges — Robust Psychological Digital Twins require hybrid AI, combining LLM flexibility with symbolic guardrails for precision mental health.
• TherapyGym: Evaluating and Aligning Clinical Fidelity and Safety in Therapy Chatbots — TherapyGym evaluates and aligns therapy chatbots using clinical fidelity and safety metrics, improving performance through RL with expert-validated feedback.
• An Agentic LLM-Based Framework for Population-Scale Mental Health Screening — An agentic LLM framework enhances mental health screening by ensuring robust, adaptable, and trustworthy AI pipelines.
• PRISM: Prosody-Integrated Multi-Agent Reasoning Framework for Empathetic Spoken Dialogue — PRISM integrates prosody into LLM reasoning via a multi-agent framework for empathetic spoken dialogue.
• GroupMemBench: Benchmarking LLM Agent Memory in Multi-Party Conversations — LLM agent memory systems perform poorly in multi-party conversations, often outperformed by simple baselines.
• Scaling Human Judgment: How Dropbox Uses LLMs to Improve Labeling for RAG Systems — LLMs can significantly augment human labelling to improve data relevance in RAG systems.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.