A Proactive Multi-Agent Dialogue Framework for Assessing Social Language Disorder Traits in Autism

2026-05-21

AI Summary

This study addresses the challenge of eliciting social language difficulties (SLD) in individuals with autism spectrum disorder (ASD) during spontaneous conversation, which often remain latent without targeted questioning. To overcome the lack of systematicity in traditional clinical assessment prompts, the authors propose TPA (Think, Plan, Ask), a multi-agent dialogue framework that introduces active inference–driven questioning strategies into AI-assisted ASD language evaluation for the first time. Built upon large language models and grounded in real ADOS-2 Module 4 data, the framework simulates collaborative dialogues between patient and clinician agents to systematically elicit otherwise hidden SLD features. Experimental results across 35 participants and 484 dialogue turns demonstrate an 82.1% coverage rate of SLD features—16.6% higher than clinical replay—and significantly improved diagnostic efficiency (AUCC: 0.628 vs. 0.458).

Abstract

Characteristic linguistic behaviors associated with Social Language Disorder (SLD) in autism spectrum disorder, including echoic repetition, pronoun displacement, and stereotyped media quoting, are largely absent from spontaneous conversation and only emerge under specific conversational conditions. In structured clinical assessments, this latency means that questioning strategy selection is a critical yet underappreciated determinant of how much diagnostic information a conversation yields. Whether large language models (LLMs) can be guided to proactively select questioning strategies that systematically surface these latent traits remains largely unexplored. Here we present TPA (Think, Plan, Ask), a proactive multi-agent dialogue framework applied to the language assessment component of the Autism Diagnostic Observation Schedule Module 4 (ADOS-2), in which a doctor agent explicitly reasons about which traits remain unobserved before selecting a clinically grounded strategy and generating a targeted question. A patient agent grounded in real ADOS-2 clinical data enables reproducible evaluation without real patient participation, validated across three independent experiments confirming adequate fidelity to real patient language. Evaluated on 484 episodes from 35 patients, TPA outperforms six competitive dialogue planning baselines across all primary metrics, achieving 82.1% SLD trait coverage, 16.6% higher than automated replay of real clinical dialogues conducted by trained clinicians (65.5%), with substantially greater per-turn diagnostic efficiency (AUCC: 0.628 vs. 0.458, absolute gain +0.170). These results demonstrate that proactive questioning strategy selection substantially improves the efficiency of automated SLD trait assessment, with direct implications for scalable AI-assisted clinical screening.

Problem

• Social Language Disorder
• Autism Spectrum Disorder
• Proactive Questioning
• Diagnostic Efficiency
• Latent Linguistic Traits

Innovation

• proactive dialogue planning
• multi-agent framework
• social language disorder
• autism assessment
• large language models