Evaluating the interpretability of clinical speech AI models: Lessons from two user studies
Computer Speech & Language · 2027-01 · 101 · 101997
DOI: 10.1016/j.csl.2026.101997 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computer Speech & Language, 2027-01. Verfasst von Lingfeng Xu und 2 weiteren Autor:innen. Beteiligte Einrichtung: College of Health Solutions, Arizona State University, 550 N 3rd Street, Phoenix, 85004, Arizona, USA. Die Arbeit nennt 43 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar. Gefördert von: NIDCD NIH HHS, National Institutes of Health, National Institute on Deafness and Other Communication Disorders.
Abstract (Originalsprache)
The deployment of Artificial intelligence (AI) in clinical speech applications has been limited in large part by the lack of interpretability, which is essential for establishing clinician trust and enabling effective decision support. Although methods such as SHapley Additive exPlanations (SHAP) aim to improve transparency in many clinical domains, their applicability to clinical speech-language pathology practice is uncertain. Since these methods rely on data modalities like acoustic signal features and spectrograms, which are unfamiliar to clinicians and misaligned with clinical workflows, the resulting interpretations may introduce additional burden and bias rather than provide clinically meaningful insight. To better understand this challenge, we conducted two consecutive user studies to systematically evaluate a commonly used SHAP-based interpretation design (a bar chart showing the influence of acoustic features on AI decisions) in dysarthria detection. Building on our prior works, eight factors were examined: faithfulness, computational efficiency, cognitive load, human-AI task performance, mental model, user trust, clinical understandability, and decision relevance. The results reveal a previously unrecognized risk in current interpretation practices. The seemingly intuitive bar-chart design frequently misled participating speech-language pathology (SLP) students to interpret feature influence as an indicator of clinical severity. Other findings include difficulty understanding AI mechanisms, discrepancies between human and model reasoning, and the limited ability of interpretations to address clinical questions. Through this work, we highlight the need for interpretation designs that are more closely aligned with clinical reasoning patterns and suggest practical considerations for developing speech-based AI systems that can be meaningfully integrated into clinical practice.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- College of Health Solutions, Arizona State University, 550 N 3rd Street, Phoenix, 85004, Arizona, USA.
- School of Electrical, Computer and Energy Engineering, Arizona State University, 650 E. Tyler Mall, Tempe, 85287, Arizona, USA.
Förderung
- NIDCD NIH HHS
- National Institutes of Health
- National Institute on Deafness and Other Communication Disorders
Zitieren
APA
Xu, L., Berisha, V., & Liss, J. (2027). Evaluating the interpretability of clinical speech AI models: Lessons from two user studies. Computer Speech & Language, 101, 101997. https://doi.org/10.1016/j.csl.2026.101997
Vancouver
Xu L, Berisha V, Liss J. Evaluating the interpretability of clinical speech AI models: Lessons from two user studies. Computer Speech & Language. 2027;101:101997. doi:10.1016/j.csl.2026.101997
BibTeX
@article{Xu2027,
author = {Xu, Lingfeng and Berisha, Visar and Liss, Julie},
title = {Evaluating the interpretability of clinical speech AI models: Lessons from two user studies},
journal = {Computer Speech & Language},
year = {2027},
volume = {101},
pages = {101997},
doi = {10.1016/j.csl.2026.101997}
}
EUVIMED