Beyond Transparency: Reframing Ethical Explainability in Human–AI Teams

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Purpose: Explainable artificial intelligence (XAI) is increasingly promoted as a mechanism for improving transparency, trust, and accountability in AI-assisted decision-making. However, its ethical value remains uncertain when AI systems are embedded within human teams and organizational workflows. This systematic literature review examines ethical issues associated with XAI in human–AI team contexts and explores how trust, decision-making, and accountability are associated with the use of XAI systems across organizational domains. Methodology: Guided by the SALSA framework and PRISMA 2020 guidelines, the review synthesized 25 studies drawn from healthcare, business and finance, cyber–physical systems, Earth Observation, and broader human–AI collaboration contexts. Findings: The findings revealed four interrelated themes: the paradox of explainability, in which transparency does not necessarily produce understanding; trust as a dynamic and fragile process shaped by user expertise, task stakes, and workflow integration; distributed accountability and the unresolved ethics of responsibility in human–AI collaboration; and bias, fairness, and the limits of explainability as a standalone ethical solution. Across the reviewed literature, XAI did not function as a self-sufficient safeguard against ethical risk. Rather, its ethical value depended on how explanations were designed, interpreted, coordinated within teams, and governed by organizations. Based on these findings, the study proposes the Sociotechnical Explainability Alignment (SEA) framework, which conceptualizes ethical XAI as requiring alignment across four levels: design, user, team, and organization. The review contributes to XAI scholarship by reframing explainability as a sociotechnical capability rather than a purely technical feature. It also offers practical guidance for organizations seeking to implement AI systems responsibly through human-centered design, trust calibration, role-sensitive explanation practices, and governance mechanisms that support accountability and fairness.