Responsible Use of Artificial Intelligence in Museum Visitor Engagement Technologies

Abstract

Artificial intelligence offers museums new opportunities to personalize interpretation, recommend exhibits, support navigation, analyze visitor behavior, facilitate multilingual access, and adapt digital content to individual interests. These capabilities can strengthen visitor engagement, but they also create ethical challenges when personalization depends on continuous behavioral observation, opaque preference inference, precise indoor location tracking, biometric analysis, extensive data retention, or automated recommendations that narrow rather than broaden cultural discovery. Responsible museum AI therefore requires a broader evaluation framework than engagement maximization alone.

This study develops a simulation-based Responsible-AI model for museum visitor-engagement technologies that integrates relevance, visitor trust, consent comprehension, privacy protection, content diversity, accessibility, explainability, and meaningful human oversight. A synthetic audit of 1,600 museum-visit sessions was created, with 400 cases allocated to each of four engagement architectures: Static Digital, AI-Personalized, Explainable AI, and Responsible AI. No real museum visitor, biometric record, location trace, demographic profile, survey response, or institutional dataset was used. The AI-Personalized configuration produced the highest simulated engagement score of 84.8/100 but also generated the highest privacy-concern score of 72.5/100 and comparatively weak visitor trust. The Responsible-AI architecture retained a similar engagement score of 83.9 while increasing trust to 86.7, consent comprehension to 88.5, and content diversity to 78.6, while  reducing privacy concern to 27.8. Logistic modeling identified third-party behavioral sharing, biometric processing, and opaque profiling as major positive predictors of high ethical-risk engagement events, whereas data minimization, meaningful consent, explanation, and access to museum staff were protective. The findings suggest that responsible museum personalization should not be designed solely to predict what visitors are most likely to select. Instead, museum AI should preserve curiosity, serendipity, interpretive autonomy, accessibility, privacy, and institutional accountability while ensuring that technology remains subordinate to the cultural and educational purposes of the museum.