Assessing AI Explainability: A Usability Study Using a Novel Framework Involving Clinicians

2025-03-21

AI Summary

Medical AI explainability assessments often lack clinical relevance, hindering real-world deployment. Method: This study proposes the first clinically grounded, three-dimensional explainability framework—comprising comprehensibility, trustworthiness, and usability—and implements it in a prototype system for postpartum depression risk prediction. The framework was developed through systematic literature review and expert consensus, followed by human-AI co-design and integration of explainable machine learning models into an interactive web-based interface. Empirical evaluation involved 20 clinicians using a novel, internally validated 13-item System Explainability Scale (SES; Cronbach’s α = 0.84, ρ = 0.81). Results: Clinicians rated the system highly across all dimensions (usability: 4.71; trustworthiness: 4.53; comprehensibility: 4.51; overall explainability: 4.56 on a 5-point scale), confirming the framework’s efficacy in mitigating explainability barriers in clinical AI adoption and establishing a new paradigm for standardized, clinically aligned explainability assessment.

Technology Category

• Humans and AI: Explainable AI (XAI) for Human Understanding
• Philosophy and Ethics of AI: Accountability, Interpretability & Explainability
• Machine Learning: Transparent, Interpretable, Explainable ML

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

An AI design framework was developed based on three core principles, namely understandability, trust, and usability. The framework was conceptualized by synthesizing evidence from the literature and by consulting with experts. The initial version of the AI Explainability Framework was validated based on an in-depth expert engagement and review process. For evaluation purposes, an AI-anchored prototype, incorporating novel explainability features, was built and deployed online. The primary function of the prototype was to predict the postpartum depression risk using analytics models. The development of the prototype was carried out in an iterative fashion, based on a pilot-level formative evaluation, followed by refinements and summative evaluation. The System Explainability Scale (SES) metric was developed to measure the influence of the three dimensions of the AI Explainability Framework. For the summative stage, a comprehensive usability test was conducted involving 20 clinicians, and the SES metric was used to assess clinicians` satisfaction with the tool. On a 5-point rating system, the tool received high scores for the usability dimension, followed by trust and understandability. The average explainability score was 4.56. In terms of understandability, trust, and usability, the average score was 4.51, 4.53 and 4.71 respectively. Overall, the 13-item SES metric showed strong internal consistency with Cronbach`s alpha of 0.84 and a positive correlation coefficient (Spearman`s rho = 0.81, p<0.001) between the composite SES score and explainability. A major finding was that the framework, combined with the SES usability metric, provides a straightforward approach for developing AI-based healthcare tools that lower the challenges associated with explainability.

Problem

• Develops a framework to assess AI explainability for clinicians.
• Evaluates AI tool predicting postpartum depression risk usability.
• Measures clinician satisfaction with explainability using SES metric.

Innovation

• Developed AI Explainability Framework with three core principles
• Built AI prototype with novel explainability features iteratively
• Created System Explainability Scale to measure clinician satisfaction