Artificial Intelligence and Contingency: A Practice-Theoretical Perspective

The paper analyzes how subsymbolic Artificial Intelligence (AI) works in sociotechnical practices. In a sociological context, it is not yet clear how to understand and evaluate the consequences of technologies based on deep learning and artificial neural networks for social contexts. This paper proposes to understand the social effects of interactive AI as the generation of contingency in practice contexts. Three forms of contingencies are distinguished: agential contingency, epistemic contingency, and formative contingency. Against this background, new perspectives emerge on the embedding and reflection of AI in usage practice.