In October 2025, OAR’s Board of Directors authorized funding for eight applied autism research grants. These new research grants, totaling $397,372, bring OAR’s total research funding to $5.8 million since 2002. This article is the first of the previews to be featured in The OARacle this year.
Communication difficulties remain a core diagnostic feature of autism, significantly affecting children’s academic readiness, social development, and long-term well-being. Early interventions can lead to positive outcomes in language skills, verbal and nonverbal communication, and joint attention, especially when the interventions incorporate supportive teaching strategies within natural environments, like home or preschool.
Teaching parents to implement interventions builds rapport between parent and child, creating a foundation for the intervention’s success. However, that success is dependent on parents’ ability to implement the interventions consistently and accurately. Research has found that effective coaching helps parents implement the interventions effectively.
Artificial intelligence opens up opportunities to support parent coaching and makes it more available to more families. Jinjun Xiong, Ph.D., will test the efficacy of PAC.AI, a family-centered, AI-enhanced digital platform designed to train parents of autistic children to deliver naturalistic communication strategies. Rather than replacing human coaching, PAC.AI can supplement professional support and expand access to families who may face geographic, financial, or workforce-related barriers to early intervention services.
The goal of “Study of Parent’s AI Coach for Parent-Implemented Communication Interventions for Children with ASD” is to evaluate PAC.AI’s effectiveness, usability, and acceptability.
Dr. Xiong is an Empire Innovation Professor with the Department of Computer Science & Engineering at the University at Buffalo, New York. Before joining the university in 2021, Dr. Xiong was program director and senior research scientist at IBM’s T.J. Watson Research Center. He also co-founded and co-directed the IBM-Illinois Center for Cognitive Computing Systems Research.
Dr. Xiong and his research team will recruit 30 parents and their autistic children, ages 3 to 6, to participate in the PAC.AI group or a comparison group. A flyer describing the study will be distributed to partners, local parents, and child autism organizations, preschools, and early intervention agencies in Western and Central New York to recruit parents to participate.
All baseline activities will use HIPAA-compliant platforms to ensure accessibility and participants’ comfort. Families will participate from their homes using personal devices such as smartphones or tablets.
The primary intervention will be shared reading via five storybooks provided by the research team. The books will be:
Parents will attend a 30-minute Zoom orientation session led by a research team member. The leader will review study goals, procedures, and timelines. Parents assigned to the PAC.AI group will receive login credentials and a demonstration of the PAC.AI platform. Parents in the comparison group will receive instructions for uploading videos. Both groups will receive digital reference materials, including video tutorials and written guides for technical troubleshooting.
Prior to participation in the study, all participating parents will record and submit three short videos capturing their reading routine to provide baseline data for assessing the parent’s implementation fidelity and the child’s communication behaviors.
A live Zoom training session will include program background, reading prompts, naturalistic communication teaching strategies, high-fidelity guides, video exemplars illustrating the step-by-step implementation, guided reflection, and live Q&A. Parents will also be able to view the training later.
The PAC.AI group will record and upload one three-to-five minute shared reading video weekly. PAC.AI will then prompt the caregiver to complete a brief self-reflection, encouraging them to describe what went well, identify moments of connection, and note any challenges. PAC.AI will analyze each video using AI models trained to identify the naturalistic communication teaching strategy used, timestamp, fidelity score, and child response.
PAC.AI will provide two levels of feedback to parents:
The system uses a friendly interface, incorporating emoji reinforcement, text-based suggestions, and brief video demonstrations to help parents improve their implementation.
Parents in the comparison group will follow the same weekly routine of recording and uploading a weekly video as the PAC-AI group. Within 24 hours of each upload, a trained interventionist will review the video and provide individualized, asynchronous coaching feedback via secure email. This feedback will follow a parallel structure to PAC.AI’s format to ensure comparability between conditions.
Parent intervention outcomes: At the end of the intervention, the research team will conduct qualitative focus group interviews with 10 parents from the PAC.AI group to explore user experience with PAC.AI, including perceived benefits, challenges, emotional support, and alignment with daily routines.
PAC.AI will score parents automatically by detecting the presence and consistency of strategy use and assigning a score. The research team will use those consistency scores to compare against the comparison group, whose videos will be coded by trained research assistants.
A usability scale and a parent survey will give the research team the data to determine if PAC.AI gave parents confidence in their ability to deliver the intervention.
Child communication outcomes: Children’s outcomes will be captured using parent-reported standardized assessments compared between the PAC.AI group and the comparison group.
If proven effective, PAC.AI could broaden access to evidence-based coaching, reducing wait times and travel burdens while empowering families with flexible, high-quality support. It may also help parents correctly and consistently deliver the intervention through its ability to provide immediate feedback, suggestions, and modeling of strategies. By building parents’ confidence incrementally, PAC.AI promotes meaningful behavior change and sustained engagement.
The research team hopes that PAC.AI can help meet the need for clinicians in places where there are shortages. Having PAC.AI to rely on can also free clinicians to concentrate on more complex family needs or increase their caseload without sacrificing quality. If successful, PAC.AI could be deployed as part of a tiered service model, with AI providing weekly support while clinicians check in as needed.
This study has the potential to benefit children, their parents, and clinicians as well as researchers, developers, and policymakers through technology that delivers scalable, innovative, personalized interventions as and where needed.
Sherri Alms is the freelance editor of The OARacle, a role she took on in 2007. She has been a freelance writer and editor for more than 20 years.
Culturally Adapted Approaches to Supporting Arabic-Speaking Individuals with Autism
When we talk about supporting autistic individuals, we often rely on models and strategies developed in cultural contexts very different...
Additional IDEA Funding Available to States
In mid-May, the U.S. Department of Education announced that it will provide an additional $144 million to states and districts to use for Individuals with...
London Marathon Team Reaches New Heights
Windy Tanner intends to fundraise for OAR until she’s raised $50,000—enough to fund a new research study. “In April 2023, my older brother, who was...
