Explainable Authorship Identification in Cultural Heritage Applications: Analysis of a New Perspective

Explainability is crucial for real-world adoption: The paper demonstrates that authorship identification systems are "hardly useful unless they are supported with suitable explanations," highlighting that high accuracy alone is insufficient for practical deployment in cultural heritage research.

Existing XAI techniques provide only partial solutions: The study tested three types of explanation methods (feature ranking, probing, and counterfactual selection) across three authorship tasks and found that "while these techniques make important first steps towards explainable Authorship Identification, more work remains to be done," indicating current tools are not yet ready for integration into scholarly workflows.

Different explanation approaches serve different purposes: The paper evaluates multiple XAI techniques rather than endorsing a single method, suggesting that scholars in cultural heritage may benefit from having access to varied explanation types depending on their specific research questions about authorship.

Domain expertise matters for AI system design: By specifically focusing on explanations "addressed to scholars working in cultural heritage," the research emphasizes that effective AI systems must be tailored to the needs and expertise of their end users, not just optimized for technical performance metrics.