XAI Effect on Laypeople VS Experts Perceptions of AI Outcomes - Preliminary Work

Artificial Intelligence (AI) systems are increasingly integrated into high-stakes domains that affect daily life. Explainable Artificial Intelligence (XAI) methods are essential for understanding how these systems make decisions, particularly when using complex "black box" models. XAI methods are diverse, and we don’t fully understand how different explanations work for different people in different contexts. Hence, this study presents preliminary results of work in process that aims to investigate the effect of different XAI methods (SHAP, LIME, and DiCE) with respect to different stakeholder types (laypeople vs. experts) and system domains (e.g., HR, Legal, Medical, Entertainment). This paper presents the initial results of our first case study, which focuses on an AI-based recruitment system.