Ethical Challenges in Explainable AI: A Review on Cultural and Social Bias
Nipuna Sankalpa Thalpage, Dulanjaya Epa, Sithara Jayawardhana · Journal of Digital Art & Humanities · 2025
Explainable Artificial Intelligence (XAI) has emerged as a critical domain within AI research, aiming to enhance system transparency and user trust. While technical advancements in XAI have improved algorithmic interpretability, ethical concerns, particularly cultural and social biases, remain underexplored. This study adopts a semi-systematic literature review approach to examine how ethical challenges, particularly cultural and social biases, are addressed in Explainable AI (XAI) research, based on papers published between 2017 and 2025, focusing on how current XAI systems recognize, reinforce, or attempt to address such biases. The findings indicate a predominant focus on Western user populations, minimal engagement with underrepresented communities, and a lack of participatory or culturally responsive design strategies. By analyzing themes across healthcare, education, and decision-support systems, the review highlights the limitations of existing models and the need for inclusive, user-centered approaches. The paper concludes by proposing research directions centered on localized explanation models, participatory design, and expanded evaluation metrics that account for cultural relevance and social equity. These insights contribute to the ongoing effort to align XAI development with ethical principles and ensure equitable AI outcomes across diverse user groups.