Conceptualizing and Validating the 4R Model of Generative AI-Based Emotional Support Scale: Regulation, Resonance, Reinforcement, and Reflection

This study conceptualizes and validates the 4R Model of Generative AI-Based Emotional Support Scale, encompassing four key dimensions: Regulation (5 items), Resonance (6 items), Reinforcement (5 items), and Reflection (4 items). These dimensions were developed based on expert interviews and are intended to assess the effectiveness of generative AI in providing emotional support. The study sample consisted of 996 participants from a northern Chinese university, with a convenience sampling method used to collect data. The validation process followed multiple steps, including item analysis, exploratory factor analysis, and confirmatory factor analysis, assessing its reliability, content validity, structural validity, convergent validity and discriminant validity. The findings indicate that the scale demonstrates satisfactory reliability and construct validity. However, while overall construct validity is acceptable, the scale exhibited some challenges in discriminant validity. This limitation can be attributed to several factors, including overlapping contextual assumptions, the interwoven nature of different support types, the lack of diversity in the sample, and the overlap of multiple response strategies. These factors likely contributed to the observed difficulties in achieving strong discriminant validity. Key limitations of the study include concerns regarding the representativeness of the sample, the cross-sectional design that precluded the assessment of test-retest reliability, the reliance on a single validation method, and the focus on generative AI technology in general rather than specific AI-based products. Corresponding solutions are proposed, including expanding the sample to ensure greater diversity, employing a longitudinal design to assess test-retest reliability, incorporating multiple validation methods, and focusing on specific AI-based products in future research. In conclusion, while the 4R Model offers a robust framework for evaluating generative AI-based emotional support, further research is necessary to refine the scale, particularly in terms of improving discriminant validity and expanding its applicability across more diverse contexts.