Machine learning in public administration research: review and prospects
系统回顾2003至2023年机器学习在公共管理文献中的应用，发现主要集中在文本分析、事件预测和因果推断三大领域，其中文本分析占75.3%，并讨论了算法偏差和黑箱问题等挑战。
Although machine learning (ML) has developed rapidly and begun to reshape social science generally, its influence on the field of public administration research remains underexplored. Therefore, this article provides a systematic review of the application of ML in the public administration literature from 2003 to 2023, finding that it falls into three main areas: text analysis, event prediction, and causal inference. Text analysis, including topic mining, text classification, and sentiment analysis, is by far the most popular application of ML, with about 75.3% of the literatures employing these methods. About 13.3% of articles employed ML for event prediction, including predictor identification and social outcome prediction. The remaining 11.3% of articles used ML for causal inference, including confounder identification and counterfactual prediction. Although facing challenges such as algorithmic bias and the black box problem, ML is expected to continue to advance and expand the available approaches to public administration research.