Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

How Human-Centered Explainable AI Interface Are Designed and Evaluated: A Systematic Survey

Thu Nguyen, Alessandro Canossa, Jichen Zhu

TL;DR

This systematic survey addresses the gap in understanding how Explainable Interfaces (EIs) are designed and evaluated within human-centered XAI. By analyzing 53 publications with thematic analysis and grounded theory, and applying a cluster analysis, the study reveals that user involvement in design is uneven, EI design often lacks explicit early user research, and interactivity and information architecture choices vary widely, with most evaluations focusing on whole-system metrics. The authors propose a generative eight-property framework to explore uncharted EI configurations, advocate for more targeted EI evaluation to reduce confounding, and highlight directions such as interactive, matrix-structured EI designs and closer HCI collaboration. Overall, the work provides a map of current EI practices and practical guidance for future HCXAI design, development, and assessment.

Abstract

Despite its technological breakthroughs, eXplainable Artificial Intelligence (XAI) research has limited success in producing the {\em effective explanations} needed by users. In order to improve XAI systems' usability, practical interpretability, and efficacy for real users, the emerging area of {\em Explainable Interfaces} (EIs) focuses on the user interface and user experience design aspects of XAI. This paper presents a systematic survey of 53 publications to identify current trends in human-XAI interaction and promising directions for EI design and development. This is among the first systematic survey of EI research.

How Human-Centered Explainable AI Interface Are Designed and Evaluated: A Systematic Survey

TL;DR

This systematic survey addresses the gap in understanding how Explainable Interfaces (EIs) are designed and evaluated within human-centered XAI. By analyzing 53 publications with thematic analysis and grounded theory, and applying a cluster analysis, the study reveals that user involvement in design is uneven, EI design often lacks explicit early user research, and interactivity and information architecture choices vary widely, with most evaluations focusing on whole-system metrics. The authors propose a generative eight-property framework to explore uncharted EI configurations, advocate for more targeted EI evaluation to reduce confounding, and highlight directions such as interactive, matrix-structured EI designs and closer HCI collaboration. Overall, the work provides a map of current EI practices and practical guidance for future HCXAI design, development, and assessment.

Abstract

Despite its technological breakthroughs, eXplainable Artificial Intelligence (XAI) research has limited success in producing the {\em effective explanations} needed by users. In order to improve XAI systems' usability, practical interpretability, and efficacy for real users, the emerging area of {\em Explainable Interfaces} (EIs) focuses on the user interface and user experience design aspects of XAI. This paper presents a systematic survey of 53 publications to identify current trends in human-XAI interaction and promising directions for EI design and development. This is among the first systematic survey of EI research.
Paper Structure (23 sections, 7 figures)

This paper contains 23 sections, 7 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Method
  4. Dataset: Search Procedure and Inclusion Criteria
  5. Data Analysis Methods
  6. Results
  7. RQ1: How do researchers involve human participants in the design and development of XAI applications?
  8. RQ2: How are explainable interfaces currently designed?
  9. Design Requirements: Visual Hierarchy and Required Features
  10. Explainable Interface Design Outcomes: Information architecture, interactivity, and interaction type
  11. RQ3: How are explainable interfaces currently evaluated?
  12. Evaluation metrics
  13. Metric types
  14. Cluster analysis of literature
  15. Cluster 1.
  16. ...and 8 more sections

Figures (7)

  • Figure 1: Publications selection process
  • Figure 2: Left: Distribution percentage of the design activities that human participants are involved in (n=53). Right: Distribution percentage of the five main groups of human participants in the Analysis Activity (n=14).
  • Figure 3: Design Requirements (including visual hierarchy and required features) correlated with the Selected Information Architecture in the Final EI Designs
  • Figure 4: Left: designed information architecture. Right: Interaction types with designed information architecture
  • Figure 5: Specific evaluation metrics (including their types)
  • ...and 2 more figures