China Social Science Review

No.4, 2024

 

 

Li Dai

 

The computational social science community has reached consensus on its approach to scientific research, encapsulated by the “prediction-replication” paradigm. By identifying three types of predictions and five uses of machine learning in social science research, we can address questions and critiques from researchers regarding the “inexplicability” and “data-driven” nature of computational methods represented by machine learning. The responsibility of researchers to elucidate their methods is directly proportional to the centrality of these methods in the argumentation. A trade-off exists between prediction and explanation: the more readily research findings can be frequently validated, the more explanation can be postponed. Today, social science research is facing challenges such as unequal distribution of computing resources, data inequality, and difficulties in publication and outcome recognition.