Social prediction: a new research paradigm based on machine learning

Yunsong Chen,Guangye He,Xiaogang Wu,Anning Hu,Guodong Ju

doi:10.1186/s40711-021-00152-z

Abstract

Sociology is a science concerned with both the interpretive understanding of social action and the corresponding causal explanation, process, and result. A causal explanation should be the foundation of prediction. For many years, due to data and computing power constraints, quantitative research in social science has primarily focused on statistical tests to analyze correlations and causality, leaving predictions largely ignored. By sorting out the historical context of "social prediction," this article redefines this concept by introducing why and how machine learning can help prediction in a scientific way. Furthermore, this article summarizes the academic value and governance value of social prediction and suggests that it is a potential breakthrough in the contemporary social research paradigm. We believe that through machine learning, we can witness the advent of an era of a paradigm shift from correlation and causality to social prediction. This shift will provide a rare opportunity for sociology in China to become the international frontier of computational social sciences and accelerate the construction of philosophy and social science with Chinese characteristics.

Highlights

To Weber, sociology is a science that concerns itself with the interpretive understanding of social action as well as the causal explanation of its process and result (Weber 1968[1921]:4)
The disciplinary significance of social prediction Since social prediction can be realized through machine learning, what is its main value for the development of social science, especially sociology as a new research field and research method? Based on our understanding of sociology and some of the latest literature, we classify the disciplinary value of social prediction into three dimensions: academic significance, governance significance, and discourse significance
We argue that in the causal inference of social science, the vast majority of counterfactual constructions can be completed by machine learning methods, and the difference between a counterfactual construction and an actual occurrence can be tested by quantitative methods

Summary

Introduction

To Weber, sociology is a science that concerns itself with the interpretive understanding of social action as well as the causal explanation of its process and result (Weber 1968[1921]:4). The problem is that the discipline has been in a state of under-preparation for prediction for many years, quantitative social science research has developed rather mature methods and models for correlation and causal analysis.

Results

Conclusion