Algorithmic fairness as sociotechnical system

A typology of the information construct

Abstract

Organizations and enterprises search for ways to exploit the vast amount of data that is produced by citizens, sensors, devices and administrative processes. Capitalizing on the produced data should be done responsibly by preventing, mitigating and managing undesired side effects such as violation of rules and regulations, human rights, ethical principles as well as privacy and security requirements. A key challenge in employing data, algorithms and data-driven systems is to adhere to the principle of fairness and justice. In this contribution we focus on the issue of algorithmic fairness, which itself can be framed as a sociotechnical system with interacting social and technical/formal subsystems. Information is a key construct of any sociotechnical system, where information creation and exchange can ease the opacity of interactions between the social and formal subsystems, and of interactions between the subsystems and the environment in which they operate. Based on literature, we categorize the types and flows of the information construct within the sociotechnical systems of algorithmic fairness in 7 categories. As such, the presented insights about the 7 categories of the information construct can form a common mental model whereby social and technical disciplines can inform each other systematically and align their views on algorithmic fairness.