Political Participation and Policy Feedback in Algorithmic Society

• In Event: Artificial Intelligence and Challenges in Policymaking

Abstract

Algorithmic decision-making systems, rooted in artificial intelligence (abbreviated as AI), has been widely used across private and public sectors, gradually forming algorithmic society in recent years. Despite algorithms to offer effectiveness and efficiency gains, stakeholders remain concerned about potential undesirable risks, questioning whether an algorithmic society would truly serve public interests. This concern leads to a surge of scholarly interest in the social impacts and possible regulatory policy responses against such transformation.
Among them, one of the least studied important issues is how algorithms impact the political attitudes and behaviors of citizens, especially their participation in policy process. On one hand, it is considered that the complexity and opaqueness of algorithms as well as its implementation process would impede the policy participation of citizens and further decrease their motivation, leading them to only caring about the output legitimacy rather than throughput or input legitimacy. On the other hand, frequently mentioned transparency and accountability principles in AI governance initiatives and laws globally still proves that stakeholders care about how the algorithms are designed, implemented and regulated, e.g. the policy process under algorithmic society.
Our paper contributes to these disputes by categorizing algorithms into distinct types and considering their heterogeneous impacts on citizen’s political participation. We contend that the ascent of algorithmic society does not exhibit uniform characteristics and impacts when algorithms are designed and implemented in diverse environments. The disputes in extant literature could be partially attributed to the heterogeneity of algorithms and its application environments. We categorized algorithms into two dimensions: the technical dimension, concerning whether they require implementation feedback, and the value dimension, assessing their connection to basic individual rights like privacy and justice. Using a survey experiment approach, we examine how individuals respond when presented with questions about political participation in different algorithmic scenarios. Given the complexity of AI algorithms, our survey targeted tech elites with basic knowledge in algorithm design. We collaborated with the China Association for Science and Technology and collected 2285 valid questionnaires.
Our paper shows that even among tech elites, exposure to the algorithmic society scenario leads to a decrease in their trust in the application of algorithms. Compared to the control group that was exposed to nothing in the questionnaires, the group exposed to algorithms demanding implementation feedback and linked to individual rights demonstrated a significantly increase in political participation. Conversely, the group exposed to algorithms not demanding implementation feedback and not related to individual rights showed a substantial decrease in voting motivation. This implies that individual’s political participations are strengthened in a highly interactive scenario between algorithmic and society. Furthermore, when individuals were asked about their willingness to provide suggestions to AI regulatory policies, the group exposed to either the algorithms demanding implementation feedback or linked to individual rights showed a significantly increase in their willingness to engage in regulation. The advent of algorithmic society will inevitably trigger widespread attention from individuals to governance. Our paper contributes to the understanding of political behaviors as the public enters the era of algorithmic society.

Authors

• Kai Jia, Shanghai Jiaotong University

• Jing Zhao