University of Twente Student Theses


Urban AI/ML Models as coupled ethical-epistemic tools of optimization

Jansen, Milou (2021) Urban AI/ML Models as coupled ethical-epistemic tools of optimization.

[img] PDF
Abstract:This research explores the underlying epistemic and moral norms of optimization in urban AI/ML models. Optimization is generally perceived as something instrumentally valuable rather than inherently valuable. However, as ML/AI models are never perfect, they are only as optimal as possible. Assumptions lead to uncertainties that are unavoidable, as they are inherent to ML-modelling. Such uncertainties stem from all kinds of potential errors that each come with particular potential consequences. By accepting a particular AI/ML model, one also accepts these structural uncertainties, potential errors and the consequences of these errors. This is referred to as inductive risk, the risk of potentially being wrong based on inference. The ‘optimal performance’ of an AI/ML system is then ultimately bounded by its ‘fit’ with society. Inductive risks are value-laden, with ethical-epistemic judgements that are normative and coupled, as they link moral and epistemic judgements on ML-systems as knowledge-producers in society. Finding the right balance is a collective quest, since the inductive risks of quantified observations are prioritized and interpreted differently among stakeholders. Explicitly acknowledging the value-ladeness of these systems, can support city-governments in finding public governance modes of smartness in the city, in support of their political legitimacy.
Item Type:Essay (Master)
Faculty:BMS: Behavioural, Management and Social Sciences
Subject:02 science and culture in general, 08 philosophy, 10 humanities in general, 70 social sciences in general, 71 sociology, 73 cultural anthropology, 88 social and public administration, 89 political science
Programme:Philosophy of Science, Technology and Society MSc (60024)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page