Knjižnica Filozofskog fakulteta
Sveučilišta u Zagrebu
Faculty of Humanities and Social Sciences Institutional Repository

Probabilistic Causality and Bayesian Networks in Philosophy of Science

Downloads

Downloads per month over past year

Kirš, Martin. (2014). Probabilistic Causality and Bayesian Networks in Philosophy of Science. Diploma Thesis. Filozofski fakultet u Zagrebu, Department of Philosophy. [mentor Lauc, Davor].

[img]
Preview
PDF (Croatian)
Download (2MB) | Preview

Abstract

The topic of the thesis is the probabilistic approach to causation and the Bayesian networks model for representation of causal relations. Central claim of probabilistic causation is that causes increase the probabilities of their effects. In the first part of the thesis I will lay out the main elements of the probabilistic approach to causation and ways to deal with the problems of Humean theory of causation using that approach. I will give special attention to the problem of spurious correlations, non-causal correlations in which neither variable is a cause of the other. The suggested solution to the problem of spurious correlations is Reichenbach’s principle of common cause which makes correlated variables conditionally independent given their common cause. I will also discuss Simpson’s paradox and suggest two solutions using the probabilistic approach to causation. In the second part of the thesis I will present the basics of Bayesian networks and sketch out possible ways of graphically representing probabilistic causation. Two main parts of Bayesian networks are a directed acyclical graph and a probability distribution. I will present main conditions that need to be satisfied in order to build Bayesian networks: Markov condition, minimality condition and faithfulness condition. In the third part of the thesis I will consider the pros and cons of Bayesian networks as appropriate models for the representation of causal relations.

Item Type: Diploma Thesis
Uncontrolled Keywords: causation, probability, Bayesian networks, Markov condition, common cause, conditional probabilities
Subjects: Philosophy
Departments: Department of Philosophy
Supervisor: Lauc, Davor
Date Deposited: 11 Feb 2015 10:12
Last Modified: 12 Feb 2015 09:10
URI: http://darhiv.ffzg.unizg.hr/id/eprint/5093

Actions (login required)

View Item View Item