Univ.-Ass. Keketso Kgomosotho, LLB LLM

 Ars Iuris uni:doc

Biography and Research Area

I am an Attorney from Johannesburg, South Africa. I am currently completing my doctoral thesis examining the intersection between machine learning (ML)-based Algorithmic Decision Systems (ADSs) and international non-discrimination law. My research proceeds from fundamental question: Can ML decision systems, based on an exclusively quantitative, empirical, mathematical operational logic ever comply with and operationalise non-discrimination law which is based on substantive equality and other value-rational principles?

I draw on Max Weber's framework of rationality, which distinguishes between formal and substantive rationality, to investigate the epistemic difference between ML’s operational logic and the substantive logic and rationality required by non-discrimination. To that end, I challenge whether ML algorithms are “intelligent,” and epistemically capable of implementing a substantive, value-driven, contextual legal framework designed for human decision-makers with broader rationality capabilities. I also challenge the universal applicability of ADS, especially in consequential decision contexts governed by non-discrimination law. Thus, certain types of decision-making contexts require substantive rationality that, so far, remains beyond the reach of Ai systems, and remains uniquely human. Ultimately, I consider whether the “right to a human decision” might offer an effective substantive and procedural legal safeguard— premised on the finding that

Why this Research matters

Machine learning algorithms increasingly govern critical aspects of our lives—from employment and education to healthcare and criminal justice. However, these systems have a bias and discrimination problem – also known as algorithmic discrimination. Moreover, the research is especially timely as there is currently no international legal framework specifically governing algorithmic discrimination, creating a protection gap in international human rights law.

How we name these computational systems profoundly shapes regulation, deployment, and public trust—making terminology not just semantic but frameworks that define policy and legal protections. AI has been shrouded in misleading labels from the start. Terms like 'intelligent' and 'learning' preceded the actual science, creating persistent misconceptions about what remains, at its core, merely binary computation processing 0s and 1s.

By identifying this fundamental, epistemic incompatibility, my research hopes to provide insights for developing legal regulatory approaches that are research-based and alive to the inherent limitations of algorithmic decision-making.