Purpose Social value is widely used as a metric for measuring and reporting the performance of social enterprises. However, traditional social return on investment (SROI) frameworks may constrain rigorous long-term assessments of social value. This study aims to examine how explainable artificial intelligence (XAI) techniques using machine learning (ML) measure social value across social enterprises, nongovernmental organizations (NGOs) and the public sector. Design/methodology/approach This study involved a systematic review with meta-regression in accordance with the PRISMA 2020 framework. A total of 2, 183 records were identified, of which 146 were included in the analysis; all records were retrieved from the Scopus database between 2020 and 2025. The XAI–ML model data were analyzed using canonical regression implemented in MATLAB and Python 3. 14. 2. Findings The canonical regression model indicates positive pooled heterogeneity (I2 = 47. 2%, τ2 = 0. 006, I2 = 61. 4%; Qbetween = 9. 82, RMSE = 0. 09, p 0. 001). The XAI–ML model shows that the domains of health and well-being, education and human capital, environment and sustainability and social protection are associated with social value outcomes. Originality/value This study contributes to advancing algorithmic transparency in XAI–ML models, particularly regarding the fairness and accountability of metrics used to measure social value across social enterprises, NGOs and the public sector.
Hanvedes Daovisan (Mon,) studied this question.