自动化行政裁量中算法风险感知的特征与演化研究——基于网络舆情的大数据分析

The Characteristics and Evolution of Algorithmic Risk Perception in Automated Administrative Discretion:Big Data Analysis Based on Online Public Opinions

随着自动化行政裁量的应用越来越广泛和多元,算法权力扩张、异化、寻租等风险愈发显著,对政府的算法治理造成了挑战。论文运用大数据分析方法,对河南部分村镇银行储户被赋红码事件的全国9328条相关微博文本进行挖掘,发现公众对算法风险具有高度的敏感性,舆情的爆发性较强且情感值较为负面,其中意见领袖是负面情感的主要引领者;公众关注度呈现高度的空间聚集性,经济发达地区的关注度聚集性更强;两次赋红码事件的公众情绪有显著的差异,且随着事件的演化,不同主体关注的焦点也呈现出高度的异质性。基于分析结果,政府在自动化行政裁量中的算法治理依然面临较大挑战,需要重点关注的方向包括:明晰行政裁量中算法权力的边界;减少自动化行政裁量中因委托-代理关系引发的道德风险问题;提升算法风险沟通的及时性和有效性,避免公众消极态度的产生。

As the application of automated administrative discretion becomes more extensive and diverse,the risks of algorithm power expansion,alienation,and rentseeking is becoming more significant.This poses a challenge to the government's algorithm governance.Using the big data analysis method,this study mined 9328 microblog posts related to the incident of depositors' health codes being assigned red codes in some village banks in Henan Province.The results show that the public was highly sensitive to algorithmic risks,with explosive public opinion and extremely negative emotional values.Opinion leaders presented the most negative emotions.Public attention shows high spatial concentration,with stronger concentration in economically developed regions.There are significant differences in the public emotions of the two redcoded events.And,with the evolution of events,the focus of attention for different subjects presents a high degree of heterogeneity.Based on the results of this study,the government faces challenges in the algorithmic governance of automatic administrative discretion.These include the unclear boundaries of the algorithmic power of administrative discretion,and the principalagent problem that may lead to moral hazards and rentseeking.Timely and effective risk communication can enhance public trust and avoid negative attitudes.