Using Artificial Intelligence to Improve Empathetic Statements in Autistic Adolescents and Adults: A Randomized Clinical Trial.
Lynn Kern Koegel, Elizabeth Ponder, Tommy Bruzzese and 7 others
PMID 39954218WHAT IT FOUND
Autistic adolescents and adults with low empathy scores used an AI chatbot for about four weeks.
Their ability to respond empathetically in conversation with a human improved significantly compared to a waitlist group, and they found the program acceptable.
Key findings
01Participants using the AI program showed a significantly greater gain in empathetic responses during human conversation than the control group, with a medium-to-large effect size.
02Self-reported comfort levels in conversation showed mixed results, with some improvements in confidence but little change in anxiety or ability to exit conversations.
03The AI program was generally acceptable to users, with most reporting they enjoyed it and would recommend it, though it occasionally provided feedback that raters found unreasonable.
STILL TO COME
How it was doneWhat they foundWhat it means for SLPs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The study used a waitlist control, not an active comparison group, so it is unclear if the gains were due to the AI specifically or just the attention and structure of any intervention. The sample size was small (30 participants), and the AI's empathy grading accuracy was 85%, meaning it misjudged one in seven responses compared to human raters. Participants were already able to engage in full-sentence conversation; the results may not apply to individuals with significant language or intellectual impairments. The study did not measure long-term maintenance of gains or distal outcomes like employment or relationships.
Declared interests
The paper does not list specific funding sources or conflicts of interest in the provided text.
The easy way to misread this
Do not assume the AI tool is ready for independent clinical use without supervision. The AI misclassified 15% of responses compared to human experts and occasionally provided feedback that raters found unreasonable, which could confuse or discourage a user.
Summarised by AI from the full paper, without a clinician reviewing it. Check it against the source before it changes what you do. Read it on PubMed →