Using Artificial Intelligence to Support Emotionally Responsive Verbal Communication Among Autistic Workers in a Work Internship Setting

Abstract

Purpose: Some autistic adults experience barriers to employment related to differences in social communication, including responding to others’ emotional cues in workplace interactions. This study extends prior randomized controlled evaluations of Noora by examining feasibility and individual-level change when the intervention is implemented with autistic young adults enrolled in an intensive vocational internship program, using a multiple baseline single-case design with repeated measurement to examine changes in relation to intervention onset within a real-world employment context. Methods: Using a concurrent multiple baseline design across three participants, we evaluated the effects of Noora, an AI-based conversational training tool. Participants completed regular training sessions focused on practicing emotionally responsive verbal communication with real-time AI-generated feedback. The primary dependent variable was performance on standardized, human-delivered conversation probes designed to elicit emotionally responsive communication in workplace-relevant scenarios across baseline, intervention, and retention phases. Results: All participants demonstrated immediate improvements in emotionally responsive verbal communication after starting the Noora program. Gains were observed within Noora training sessions and during generalization and retention conversation probes. These effects were supported by visual analysis and strong effect sizes, indicating a functional relationship between the program and improved emotionally responsive verbal communication. Conclusion: Findings suggest that Noora is a feasible and scalable intervention for supporting emotionally responsive communication in autistic young adults within vocational training contexts.