Data-Driven AIGC Exhibition Evaluation Framework Research: Model Construction and Empirical Analysis

Abstract

Artificial Intelligence Generated Content (AIGC) has been used more and more in art exhibitions in recent years. At the same time, evaluating AIGC-based exhibitions is still difficult because existing methods often rely on limited indicators or subjective judgments. To deal with this problem, this paper develops a data-based evaluation approach by combining different kinds of information. The study brings together design features generated by AIGC, audience behavior recorded during exhibitions, and questionnaire results. These different types of data are analyzed together. Partial Least Squares Structural Equation Modeling (PLS-SEM) is used to examine the relationships among technological characteristics, audience behavior, and exhibition perception. XGBoost is then applied to predict exhibition experience and to identify the factors that have relatively larger effects on the evaluation results. Based on the above analysis, the study further combines information from technology, exhibition space, and audience responses to form an evaluation process. The results indicate that this method can partly reduce the subjectivity that exists in traditional exhibition evaluation. In addition, the analysis shows that AIGC-generated visual diversity has a positive relatio nship with visitors’ narrative immersion. The prediction model also shows better consistency than some traditional evaluation approaches. These findings suggest that quantitative analysis can provide another way to understand AIGC exhibitions, especially in terms of content design, curatorial choices, and future evaluation practices.