Explainable AI for Non-Expert Users: A Systematic Mapping Study Using the Petersen et al. SMS Approach

Abstract

Artificial intelligence (AI) has transformed major sectors such as healthcare, marketing, education, and finance, but the opacity of AI decision-making remains a major obstacle for non-expert users. Explainable AI (XAI) aims to close this gap by making AI decisions clear, interpretable, and understandable. Despite substantial growth in XAI research, a critical gap persists: most XAI tools are designed with technical users in mind, leaving non-experts — such as students, patients, auditors, consumers, and public-sector workers — underserved. This study presents a systematic mapping study (SMS), based on Petersen et al.'s five-stage framework, of 412 XAI papers published between 2020 and 2025. After removing 111 duplicates, 301 unique papers were screened and coded using a six-facet classification scheme, and a main analysis set of 104 papers was used for frequency analysis, cross-mapping, and gap analysis. Results show that education (31.7%), general XAI/human-centered AI (HCAI) (29.8%), and healthcare (23.1%) are the most represented domains, while finance, cybersecurity, and public services receive far less attention. Trust and comprehensibility are the most studied HCAI constructs, whereas fairness, human control, and usability are markedly underrepresented. Sixteen research gaps were identified, with key findings pointing to a lack of real-user validation, ambiguous reporting of explanation types, and insufficient natural-language and example-based explanations. This study maps the current XAI landscape for non-expert users and outlines directions for future, human-centered, accessible AI design.