Development and Application of a Generative AI–Human Collaborative Meta-Analysis Procedure: Focusing on Digital Art Therapy Research

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Meta-analysis synthesizes evidence across studies but is time-consuming and susceptible to judgmental variation and data extraction errors. Although generative AI has been applied to individual meta-analytic tasks, few studies have proposed a human–AI collaborative framework with validation procedures for the entire process. To address this gap, this study proposes a six-stage meta-analysis procedure through human– AI collaboration and introduces a human-guided AI convergence process in which AI outputs are iteratively refined according to researchers' judgment criteria. The framework was applied to domestic digital art therapy studies. The findings indicate that generative AI supports rather than replaces human judgment by assisting key meta-analytic tasks, thereby improving procedural consistency. However, the convergence outcomes reflect fidelity indicators derived from cases used in the iterative refinement process, and their generalizability requires validation in independent datasets and other research domains. This study presents a reproducible meta-analysis procedure integrating generative AI with human validation through reusable prompts and standardization rules.