Hildebrandt presented on ‘The Methodenstreit in Machine Learning’

Mireille Hildebrandt presented on The New Methodenstreit in Machine Learning at the International Conference on Explaining Machines, at the University of Bielefeld, Germany (the place where famous sociologist Niklas Luhmann worked for most of his professional life). The Conference is part of the TRR 318 Constructing Explainability research project, a collaboration across different regions and disciplines, bringing together computer science, social science and the humanities. The project is led by Elena Esposito and Tobias Matzner. See Elena Esposito’s salient and important article in CRCL, on Transparency versus explanation: The role of ambiguity in legal AI.

Hildebrandt discussed why and how we can learn from the century old ‘Methodenstreit’ that raged in the social sciences ever since the desire to be ‘like’ the natural sciences became mainstream. She traced the different meanings of explanation in terms of reasons and causes, the difference between explanation and justification (highly relevant for law) and finally the key difference between Verstehen (understanding) and Erklaeren (explaining) that defined the difference between causal reasoning and hermeneutical understanding. She proposed to seek ‘a new hermeneutics’ with regard to computational systems, noting that the issue of proxies will be a key point of interest to develop such ‘a new hermeneutics’. For machine learning to be possible, those who develop the research design have to make a whole range of upstream design decisions that have a number of downstream implications and basically determine the affordance of the system. Many of these decisions concern the choice and the construction of proxies, that allow the translation of real world events, states or goals into machine-readable proxies. This ‘translation’ or ‘transformation’ should be the focal point of a new hermeneutics, as it will allow us to highlight how the practice of understanding is transformed into a practice of logical reasoning, preparing for causal explanations. Hildebrandt’s talk paves the way for the monograph on ‘A new hermeneutics for computational law’ that will be one of the key outputs of the COHUBICOL project.

For slides, click here.