N. Emran

Synthetic Data Quality Evaluation in Generative AI: Current Trends, Challenges, and Future Directions for Social Science Research

The findings indicate that while modern generative models can produce highly realistic and analytically useful datasets, persistent challenges remain, including the lack of standardized benchmarking protocols, utility–privacy trade-offs, privacy leakage risks, bias amplification, limited explainability, and governance concerns.