Algorithmic Normativity

Abstract

This study focuses on the normativity of algorithms and employs multidisciplinary theories and methods to analyze its manifestations and impacts at the technological, sociotechnical, and behavioral levels. Through engineering practice cases and experiments on recommendation systems, the normativity of technological evolution, the integration of engineers’ values, and the behavioral characteristics of learning machines within the algorithmic system are revealed. The experiments demonstrate that technological normativity enhances the click-through rate and conversion rate of recommendation systems; sociotechnical normativity improves the fairness and satisfaction of recommendations; and behavioral normativity promotes the expansion of users’ interests, with user attributes playing a moderating role. The research findings contribute to understanding the role of algorithmic systems in engineering and social processes, provide a theoretical framework for interdisciplinary research, contribute to the study of human-machine relationships and the social impacts of algorithms, and offer references for algorithm governance. Finally, research limitations and future directions are enumerated, including incorporating geographical factors, examining cross-cultural effects, exploring emerging fields, and constructing algorithm governance mechanisms.