Participatory practices in AI design for cultural heritage: a scoping review on methods, inclusion, and epistemic justice

Purpose This study aims to examine how users and communities are integrated into the design and evaluation of semantic or neurosymbolic (NeSy) information systems in digital libraries. Through a structured scoping review and content analysis of academic databases, it identifies and describes multidisciplinary participative practices in the development of these systems; maps methodological trends, gaps and practical recommendations for collaboration in epistemically just designs to support inclusion in digital libraries. Design/methodology/approach The study combines a search, appraisal, synthesis, analysis-based scoping review with bibliometric mapping and qualitative content analysis. It is useful for identifying multidisciplinary participative practices and for observing their methodologies in analytical variables: 1-Timing and process-driven patterns, 2-Methods, 3-Engagement and levels of user participation and 4-Ownership, inclusion, representational gaps and epistemic-justice considerations. Findings Analysis reveals that participative practices are integrated and documented in all 22 analyzable cases, with uneven reporting across phases: early user integration occurs in seven practices (32%), while users are positioned in the late evaluation phase or in narrowly scoped continuous contribution in other cases. Participatory methods cluster around co-design workshops, usability testing and crowd-science. The participation levels concentrate in consultative and collaborative stages, with some empowerment actions. Epistemic-justice considerations are under-documented. Bias-aware modeling and provenance tracking are found in specific cases. Evaluation practices tend to rely on usability and information-retrieval metrics, leaving participation impacts, trust-building and inclusion outcomes under-documented. Research limitations/implications The analysis is constrained by uneven documentation across practices, a temporal focus on the past decade and a corpus dominated by English-language and Global North sources, which limits generalizability. Geographic and linguistic coverage remain uneven: cases cluster in Europe and the Global North and several source papers do not report site geography explicitly. Hybrid initiatives often resist neat categorization, which underscores the need for more transparent reporting of participatory processes and evaluation methods. Future research should combine longitudinal and comparative designs, broaden geographic and linguistic coverage and refine tools for assessing participation, trust and epistemic justice in human-centered AI systems for digital libraries. Originality/value By mapping who is involved, the timing, the methodology, the engagement and the ownership, the study offers a structured lens on the inclusiveness, trustworthiness and accountability of human-centered semantic and AI practices in digital libraries. To authors’ knowledge, no prior synthesis integrates participatory design, citizen science, semantic/NeSy and epistemic-justice literatures in a single cross case map of existing documentation of participative practices on digital library and cultural heritage information systems. The five-lens framework and four-dimension codebook serve as analytical tools; corpus findings and recommendations serve as a pragmatic guide for practitioners seeking bias-aware participation in digital library infrastructures.