Overview of basic design recommendations for user-centered explanation interfaces for AI-based clinical decision...

Area of Science:

• Healthcare Informatics
• Human-Computer Interaction
• Artificial Intelligence

Background:

• Limited adoption of AI-based clinical decision support systems (CDSS) in healthcare.
• End-user difficulty understanding opaque AI model outputs hinders trust and implementation.
• Lack of clear guidelines for designing effective explanation user interfaces (XUIs) for AI in healthcare.

Purpose of the Study:

• To identify and synthesize recommendations for the user-centered design of XUIs from scientific literature.
• To provide an overview of best practices for presenting AI explanations to end-users in healthcare.
• To address the gap in knowledge regarding optimal XUI design for AI-driven healthcare tools.

Main Methods:

• Conducted a scoping review of articles published between 2017-2022.
• Included original research and literature reviews focusing on XUIs for end-users or domain experts.
• Extracted and organized recommendations using a mind map, summarizing 240 initial recommendations into 64.

Main Results:

• Extracted 240 user-centered design recommendations for XUIs from 47 included articles.
• Summarized and organized these recommendations into a mind map, resulting in 64 key recommendations.
• Identified a need for basic, awareness-raising recommendations rather than specific, implementable designs for AI-based CDSS.

Conclusions:

• The review offers a synopsis of fundamental recommendations for user-centered XUI design in healthcare.
• Highlights that specific, directly implementable design guidelines for AI-based CDSS are not yet established.
• Emphasizes the importance of raising awareness about user-centered design principles for XUIs in AI healthcare applications.