10 Bias in Cybernetic AI Systems
Rainer Mühlhoff · Bristol University Press eBooks · 2025
This chapter addresses the ethical issue of bias in AI, particularly in cybernetic AI systems, highlighting how biases can emerge and become entrenched through feedback loops between machine learning models and social realities. The concept of bias is explained as a systemic issue, not limited to individual cases, but visible when comparing aggregate outcomes across different groups. The chapter distinguishes between first-degree bias, which is rooted in the inherent purposes of AI systems like the Correctional Offender Management Profiling for Alternative Sanctions system (COMPAS), and second-degree bias, which arises from the biased feasibility of feedback loops that make it harder to correct errors for marginalised groups. The chapter makes the central argument that predictive models are performative.