Survey for Categorising Explainable AI Studies Using Data Analysis Task Frameworks

XAI research lacks clear task descriptions: Many explainable AI studies fail to adequately describe what specific data analysis tasks they're addressing, making it difficult for practitioners to know which findings apply to their own work.

Studies often ignore real-world context: Research frequently tests XAI systems in isolation without considering the actual workplace environment, user background, or organizational constraints where these tools would be deployed, limiting the practical applicability of findings.

User expertise matters but goes unreported: Studies rarely specify whether their test participants were domain experts, AI specialists, or general users, making it impossible to know if results will generalize to different audiences—a critical gap for responsible AI deployment.

Contradictory findings stem from methodological gaps: The field's conflicting recommendations about how to explain AI decisions often result from studies not clearly defining "what" is being explained, "why" it's needed, and "who" needs the explanation, rather than from genuine disagreement about best practices.

Better reporting standards are essential: The authors propose that XAI studies should follow consistent guidelines documenting user expertise levels and task contexts, which would help the field move from scattered findings toward concrete, reliable design recommendations for building trustworthy AI systems.