College
College of Nursing
Mentor Information
Byeol Kim
Description
Early identification of autism spectrum disorder (ASD) remains challenging because many clinically relevant social, emotional, and communication behaviors are difficult to capture during brief clinical encounters. Young children may have limited attention spans, communicate differently with unfamiliar adults, struggle to express thoughts and emotions, or behave differently in clinical settings than in everyday environments. Consequently, clinicians often rely on brief observations and caregiver reports, which may not fully represent children’s typical communication behaviors. This study introduces an AI-powered conversational application for smartphones that supports longitudinal collection of naturalistic communication data from children. The application engages children in age-appropriate, play-based conversations using AI-generated speech designed to encourage sustained engagement and reduce the formality of clinician-led interviews. Conversations are personalized using child-specific demographic and behavioral profiles, allowing dialogue to adapt to developmental level, communication abilities, and clinical concerns. Following each interaction, the application generates structured summaries of conversational content, communication characteristics, behavioral observations, and potential developmental concerns for clinician review. Unlike clinic-based assessments, the proposed approach enables repeated conversational interactions in familiar environments, allowing communication patterns to be observed over time rather than during one appointment. This longitudinal approach may complement existing autism screening and diagnostic workflows by providing contextual information about a child’s language use, social reciprocity, emotional expression, and conversational behaviors across multiple sessions. Future work will focus on optimizing AI conversation strategies for autism assessment, including developmentally appropriate prompts, adaptive questioning techniques, and conversational scenarios that elicit clinically meaningful language, social communication, and behavioral characteristics relevant to ASD diagnosis.
A Mobile AI Conversational Platform for Supporting Autism Assessment in Children
Early identification of autism spectrum disorder (ASD) remains challenging because many clinically relevant social, emotional, and communication behaviors are difficult to capture during brief clinical encounters. Young children may have limited attention spans, communicate differently with unfamiliar adults, struggle to express thoughts and emotions, or behave differently in clinical settings than in everyday environments. Consequently, clinicians often rely on brief observations and caregiver reports, which may not fully represent children’s typical communication behaviors. This study introduces an AI-powered conversational application for smartphones that supports longitudinal collection of naturalistic communication data from children. The application engages children in age-appropriate, play-based conversations using AI-generated speech designed to encourage sustained engagement and reduce the formality of clinician-led interviews. Conversations are personalized using child-specific demographic and behavioral profiles, allowing dialogue to adapt to developmental level, communication abilities, and clinical concerns. Following each interaction, the application generates structured summaries of conversational content, communication characteristics, behavioral observations, and potential developmental concerns for clinician review. Unlike clinic-based assessments, the proposed approach enables repeated conversational interactions in familiar environments, allowing communication patterns to be observed over time rather than during one appointment. This longitudinal approach may complement existing autism screening and diagnostic workflows by providing contextual information about a child’s language use, social reciprocity, emotional expression, and conversational behaviors across multiple sessions. Future work will focus on optimizing AI conversation strategies for autism assessment, including developmentally appropriate prompts, adaptive questioning techniques, and conversational scenarios that elicit clinically meaningful language, social communication, and behavioral characteristics relevant to ASD diagnosis.
