Public Administration and Policy Review ›› 2023, Vol. 12 ›› Issue (6): 77-.

Previous Articles     Next Articles

How Task Objectivity and Emotional Similarity AffectPerceived Fairness and Acceptance of Algorithmic Decision-making?—An Empirical Analysis based on Survey Experiment

  

  • Online:2023-11-17 Published:2023-11-17

任务客观性、情感相似度何以影响算法决策感知公平与接受度?——基于调查实验的实证分析

  

Abstract:

With the in-depth development of the new generation of information technology,algorithmic decision-making (ADM) has greatly improved the scientificity and accuracy of governmentdecision-making. At the same time, algorithmic discrimination, decision-making “black box”,lack of accountability and lack of interpretability and other dilemmas brought by ADM have led the government to face new challenges of decision-making fairness. Based on the simplification theory and the theory of computer as an actor, this study constructs 2 (task objectivity: subjective task vs objective task) × 2 (emotional similarity: high vs low) two-factor inter-test experiment from the perspective of algorithmic simplification and algorithmic anthropomorphism. Taking 1356 Shanghai residents as subjects, this study analyzes the causal relationship between task objectivity and citizens?? acceptance of ADM. We also examine the moderating role of emotional similarity and the mediating role of perceived fairness in that process. This study finds that citizens have higher perceptions of procedural fairness, distribution fairness, and acceptance of ADM when performing subjective tasks. Compared with high emotional similarity, the situation with low emotional similarity further strengthens citizens?? perceptions of procedural fairness, distribution fairness and acceptance of ADM. And when the emotional similarity is low, perceived procedural fairness and distributive fairness play moderated mediating effects between the task objectivity of ADM and ADM acceptance.This study provides insights into the contextual variation of simplification properties and anthropomorphic boundaries of ADM, and puts forward suggestions on how to introduce algorithms to assist decision-making and improve algorithm design in public sector.

Key words: Algorithmic Decision Making, Emotional Similarity, Perceived Fairness, BehavioralPublic Administration, Survey Experiment

摘要:

伴随着新一代信息技术的纵深发展,算法决策极大提升了政府决策的科学性和精准性。与此同时,算法决策带来的算法歧视、决策“暗箱”、责任性缺失和可解释性不足等困境,导致政府面临着新的决策公平性挑战。本研究基于简化理论和计算机作为行动者理论,分别从人工智能算法简化性视角下的任务客观性和拟人化视角下的情感相似度出发,构建了2 (算法任务客观性:主观任务vs客观任务)×2 (情感相似度:高vs低)的两因素被试间实验,以1356位上海市居民为被试,分析了任务客观性对公民算法决策接受度的因果关系,检验了情感相似度在这一过程中的调节作用以及感知公平的中介作用。研究发现,相较于执行主观任务,公民对算法决策执行客观任务时的感知程序公平、分配公平和接受度更高。相较于高情感相似度,低情感相似度的情境进一步强化了公民对算法决策的感知程序公平、分配公平和接受度。且当情感相似度较低时,感知程序公平、分配公平在算法决策客观性影响算法决策接受度之间发挥有调节的中介效应。本研究对于算法决策简化属性的情境变化和拟人化边界做了思考,并对公共部门如何引入算法进行辅助决策和改进算法设计提出了对策建议。

关键词: 算法决策, 情感相似度, 感知公平, 行为公共管理, 调查实验