SLPCase SeriesAmerican journal of speech-language pathology2024

Artificial Intelligence-Assisted Speech Therapy for /ɹ/: A Single-Case Experimental Study.

Nina R Benway, Jonathan L Preston

PMID 39173110

WHAT IT FOUND

In five children and young adults with speech sound errors, adding AI-assisted practice to clinician-led therapy improved /ɹ/ production.

Three of five showed clear treatment response on unpracticed words, though results varied by individual.

Key findings

01Three of five participants showed statistically significant improvement on unpracticed words compared to baseline, while two did not.

02The AI classifier's agreement with human clinicians was moderate, falling within the range of disagreement seen between human experts for four of the five participants.

03Participants who were older, had atypical error patterns, or had attention difficulties showed less consistent or smaller treatment effects.

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.

Already have one?

What it does not show

Very small sample size (n=5) limits generalizability. Single-case design lacks a control group, making it difficult to rule out natural maturation or placebo effects. Unstable baselines for two participants complicate the interpretation of treatment effects. The AI system is in pre-alpha development and not available for clinical use. Results may not apply to populations with non-North American English dialects or those with severe apraxia.

Declared interests

Funded by internal grants from Syracuse University and NIH. The principal investigator is associated with the development of the ChainingAI software used in the study.

The easy way to misread this

Do not interpret the average improvement as a guaranteed outcome for all patients. Two of the five participants did not meet the threshold for clinically significant change, and one participant with an atypical error pattern showed no significant difference from baseline during the treatment phase.

Read it on PubMed →