Effects of conversational AI-enhanced peer-mediated intervention by peers with intellectual disabilities on conversational skills in children with ASD.
Adding AI whispers to peer buddies sparks more child-to-child talk than peers alone.
01Research in Context
What this study did
The team paired kids who have autism with classmates who have mild intellectual disability.
Each peer wore an earbud linked to a phone app.
The app whispered conversation starters like "Ask what he built" when the child stayed quiet for five seconds.
Baseline, then the AI help, then peers alone, measured talk turns and new words.
What they found
With the AI prompts, kids with autism started twice as many topics and answered 40 % more often.
They also used a bigger variety of words.
When the AI stopped, most children kept the new skills and even used them with brand-new partners.
How this fits with other research
Lowe et al. (1995) and Syriopoulou-Delli et al. (2012) already showed that trained peers alone lift social talk.
This study adds a tech booster and still sees gains, so it extends rather than replaces those older methods.
Doherty et al. (2018) used PECS pictures for peer exchanges and got similar jumps in responses.
The new AI tool reaches the same goal without picture cards, giving teachers another option.
Why it matters
You can keep your peer programs and simply plug in low-cost AI coaching.
No extra staff, no big materials—just an earbud and a free app.
Try it during recess or center time and watch conversation bloom.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Put one earbud on a trained peer, open the AI chat app, and prompt three conversation starters at recess.
02At a glance
03Original abstract
BACKGROUND/AIMS: Children with autism spectrum disorder (ASD) often experience conversational difficulties that hinder social interactions and peer relationships. This study examined whether a conversational AI-enhanced peer-mediated intervention delivered by peers with intellectual disabilities (CAI-PMI-ID) yielded superior outcomes in improving conversational skills in children with ASD compared with PMI-ID-only. METHOD: This study employed a multiple-probe across participants design and an alternating treatments design to compare the effects of CAI-PMI-ID and PMI-ID-only. Conversational outcomes included quantitative measures (frequency of appropriate initiations and responses) and qualitative measures (mean length of utterance in morphemes, MLU-M; number of different words, NDW; and total number of words, TNW). RESULTS: CAI-PMI-ID produced greater improvements in the frequency of appropriate conversational initiations and responses than PMI-ID-only. It also yielded superior outcomes in lexical diversity and productivity, as reflected in NDW and TNW. The two intervention conditions resulted in limited and variable effects on MLU-M. CAI-PMI-ID facilitated the successful generalization of conversational skills to interact with different peers. CONCLUSIONS/IMPLICATIONS: These preliminary findings support the feasibility and added value of CAI-PMI-ID for enhancing core conversational behaviors and lexical productivity among children with ASD. Future research should integrate evidence-based syntactic supports to promote MLU-M, leverage multimodal and scenario-based AI in natural settings, strengthen AI-peer collaboration, and improve AI speech recognition to enhance intervention effects.
Research in developmental disabilities, 2026 · doi:10.1016/j.ridd.2026.105249