Best practices for leveraging usability studies in XAI

Determine effective methodologies for leveraging usability studies within the wider context of explainable artificial intelligence to evaluate and improve the interpretability and accessibility of explanations, including approaches such as standardized question banks and assessments of different query and modality types.

Background

The paper discusses the need to rigorously assess explanation quality and usability, especially for visualization methods and human-centered evaluation in XAI. While EC-based techniques can generate and assess explanations, a key gap remains in how to systematically use usability studies to measure whether users can interpret and trust explanations.

In this context, the authors note that the broader XAI community has proposed directions such as question banks and evaluating query/modality types, and they emphasize that these approaches can be adapted for EC to make explanations more accessible and useful to decision-makers.

References

How best to leverage usability studies within the wider context of XAI is still an open question, with proposals including the creation of question banks, or evaluating different query and modality types.

— Evolutionary Computation and Explainable AI: A Roadmap to Understandable Intelligent Systems  (2406.07811 - Zhou et al., 2024) in Section 3.7 (Assessing Explanations)

DR-LabStack provides method and source context beside structured-data entry forms. Its current information panels offer a place to communicate model context, while their adequacy for clinicians remains an empirical question.

— DR-LabStack: Design and Implementation of a Clinician-Facing Web System for Diabetic Retinopathy Prediction  (2609.10796 - Xu et al., 9 Sep 2026) in Section 2.4, Clinical interfaces and human--AI interaction

In particular, it is unclear whether such proxy metrics align with how stakeholders perceive or use XAI explanations .

— XAI-Arena: Can LLMs Assess the Quality of XAI Explanations?  (2609.09428 - Fleischhauer et al., 8 Sep 2026) in Section 1, Introduction

Indeed, recent work suggests that the distinction between the two paradigms may be less fundamental than it appears: both can instantiate the same geometric object, differing only in how the object is identified. Whether either approach ultimately produces models that people understand better therefore remains an open empirical question which belongs squarely within a model-centric agenda.

— From Interpretability Methods to Interpretable Models  (2609.05399 - Colin et al., 4 Sep 2026) in Section 4, paragraph “Design and post hoc explanation remain uncompared”

Future work should address several open challenges. Alternative uncertainty representations, such as frequency-based formats or hypothetical outcome plots, should be explored to better support intuitive interpretation. Additionally, future work should investigate design and interaction approaches for explaining prediction uncertainty in ways that improve both understandability and actionability. Further research should explore how AI scheduling tools can account for fairness across staff and examine how repeated interaction affects trust calibration and the risk of automation bias over time. Finally, the tension between explanation transparency and cognitive load calls for adaptive and personalized XAI interfaces that tailor explanation depth and salience to individual user needs and expertise levels. Longitudinal and in-situ evaluations would further help determine how these interaction patterns evolve with experience and how AI decision support becomes integrated into everyday work practices.

— Communicating Uncertainty in AI-Supported Nurse Scheduling: Expert Co-Design and Evaluation of an Explainable AI Prototype  (2610.18679 - Rollwagen et al., 13 Oct 2026) in Conclusion, final paragraph