Cookies?
Library Header Image
LSE Theses Online London School of Economics web site

Structural theories of modelling token causation.

Stentenbach, Michael Joachim (2007) Structural theories of modelling token causation. PhD thesis, London School of Economics and Political Science.

[img]
Preview
PDF
Download (6MB) | Preview

Abstract

This thesis deals with the most prominent accounts of analyzing singular event causation by employing counterfactuals or counterfactual information. The classic counterfactual account of token event causation was proposed in 1973 by the philosopher David Lewis and ruled that an event c is a cause of event e, if and only if there is a chain of counterfactually dependent events between c and e. Apart from facing conceptual problems due to its metaphysical claim to analyze causation 'as such' and to reduce it to counterfactual dependency, this account also produced implausible results: first, it stipulated that token causation is a transitive relation, and second, it could not analyze situations in which an effect is over-determined by various causes, either symmetrically or by one cause pre-empting another one. In 2000, almost three decades later, Judea Pearl, formerly an engineer, formulated a new and highly influential theory of modeling causal dependencies using counterfactual information that, as I argue, neither faces these conceptual problems nor produces these undesired results. This formal theory analyzes causal relationships between token events in a given situation in two steps: first, a causal model describing the relevant mechanisms at work in the situation is constructed, and second, causal relationships between the events featured in the situation are determined relatively to this model. Pearl's definition of causation according to a model is technically complicated, but its underlying rationale is that the decisive property of a cause is to sustain its effect via a certain causal process against possible contingencies, this notion of sustenance embodying an aspect of production and an aspect of counterfactual dependency. This theory of Pearl's was received with great interest in the philosophical community, most importantly by Christopher Hitchcock and James Woodward, who tried to simplify this account while preserving the basic intuition that a cause is linked to its effect by a causal process, in essence a concatenation of the mechanisms at work in the situation, just defining a causal process in a formally simpler way. I describe and employ this simplified account by Hitchcock and Woodward as a graphic introduction to Pearl's theory, because the same basic notions, like the one of a causal model, are defined in a formally more accessible way and the basic problems, like the generation of a causal model, become obvious. I mainly discuss Hitchcock's account, since this is the earlier one, since it is more elaborate, and mainly since it is conceptually in need of clarification. Woodward's account is in essence equivalent to Hitchcock's, given a slightly changed terminology. The core of my thesis consists of a comparison of Pearl's theory with Hitchcock's account. I present four paradigmatic examples, three of which are judged differently by these two theories. In each of these three examples our causal intuition is in accord with the judgment delivered by Pearl's account but contradicts the verdict of Hitchcock's. I draw the conclusion that Hitchcock's project of simplifying Pearl's theory fails in the second step of causal analysis, i.e. in defining causation according to a given model. Building on the lessons learned from this comparison, I offer a slight generalization of Pearl's definition of token causation according to a model, since Pearl's original account has the shortcoming that token causes cannot be exogenous in a model.

Item Type: Thesis (PhD)
Uncontrolled Keywords: Philosophy
Sets: Collections > ProQuest Etheses
URI: http://etheses.lse.ac.uk/id/eprint/1970

Actions (login required)

Record administration - authorised staff only Record administration - authorised staff only

Downloads

Downloads per month over past year

View more statistics