Human self-determination within algorithmic sociotechnical systems

Abstract

In order to investigate the protection of human self-determination within algorithmic sociotechnical systems, we study the relationships between the concepts of mutability, bias, feedback loops, and power dynamics. We focus on the interactions between people and algorithmic systems in the case of Recommender Systems (RS) and provide novel theoretical analysis informed by human-in-the-loop system design and Supervisory Control, in order to question the dynamics in our interactions with RSs. We explore what meaningful reliability monitoring means in the context of RSs and elaborate on the need for metrics that encompass human-algorithmic interaction. We derive a metric we call a barrier-to-exit which is a proxy to the amount of effort a user needs to expend in order for the system to recognize their change in preference. Our goal is to highlight the assumptions and limitations of RSs and introduce a human-centered method of combating deterministic design.