Actual Causation: Definitions and Principles

Causal modelling has become ubiquitous in Artificial Intelligence circles, and is gaining popularity in other fields as well. An unsolved problem in this context is how to define actual causation, i.e., when should we say that one event caused another? Although progress has been made over the last decade, not a single definition on offer goes uncontested. The goal of this PhD is to develop new proposals for defining actual causation, and formulate principles which such definitions should satisfy. The literature now contains many definitions of actual causation, each with its own strengths and weaknesses. The fact that the formulations of these different definitions diverge widely, proves a major obstacle for evaluating and comparing them. A second problem is posed by the standard practice of assuming that questions of causation can be separated from their field of application, because this conflicts with recent research that suggests actual causation is a strongly context-dependent concept. In the first part of the PhD we address both issues, by presenting a formal framework for expressing definitions and extensions of actual causation. Using this framework, we construct several definitions of causation, and offer a comparison. Further, we discuss some challenges that these definitions face. Another important problem is that many existing definitions lack proper foundations. Even when a detailed justification is given, it mostly consists of informal guidelines rather than precise formal conditions. By contrast, in the second part of the PhD we aim to make explicit what principles we take to be fundamental to causation, and show their consequences on particular examples. We use these principles to construct a definition of causation which satisfies them, and show how this resolves the challenges from the first part.status: publishe