Combining Human-centred Explainability and Explainable AI

Two separate fields need better integration: The paper identifies that human-centered explainability (focused on how people understand explanations) and explainable AI (focused on technical interpretability methods) have developed largely independently, missing opportunities to make AI systems more genuinely understandable to users.

Technical explanations alone aren't enough: The distinction between these fields suggests that simply making AI's internal logic transparent doesn't automatically make systems understandable to real people—explanations must be designed with human cognition and context in mind.

Design matters for AI transparency: By combining insights from human-centered design with AI explainability techniques, the authors suggest we can create AI systems that are both technically interpretable and actually useful for people trying to understand how decisions are made.

New approaches are emerging: The paper presents preliminary work on an "algebraic machine learning approach" as a concrete example of how these two perspectives can be merged, indicating that practical solutions combining both fields are being developed.

Broader community collaboration is needed: The authors call for continued discussion across the "HCxAI community," suggesting that responsible AI requires ongoing dialogue between AI researchers, designers, and social scientists rather than siloed technical development.