Cogprints

On causal and constructive modeling of belief change

Ayyadevara, Dr Ravishankar (2006) On causal and constructive modeling of belief change. [Thesis]

Full text not available from this repository.

Abstract

The process of changing beliefs as a result of accepting the new information is often called Belief revision. It occupies a central position in the area of philosophy, theoretical computer science and logic. However, problem of Belief revision in general is how an agent revises her current beliefs when new information obtained from reliable and evidential source contradicts some of the old beliefs, while preserving the core beliefs. One of the key aspects of the problem of changing beliefs is to provide a means of accommodating new information causing minimal change to the beliefs already held by an agent. Therefore, providing an appropriate mechanism for ordering beliefs in the belief revision is important. This dissertation is a contribution to the study of belief revision from both causal and constructive perspective. Modeling of belief revision address the following general question: Given an initial knowledge base and a new piece of information to be incorporated into it, what should be the appropriate ordering of beliefs so that less entrenched beliefs are lost when compared to more entrenched beliefs. In this study we focus on \emph{causal relevance} and propose a new entrenchment ordering, called as causal epistemic entrenchment (CEE). We ground it on a solid semantic foundation by making use of structure, intervention, causal properties, and causal mechanism. The key idea of such an entrenchment ordering is that not all conditional beliefs in a belief set are important and hence not all beliefs would be relevant for the Belief revision. Precisely, the thesis makes the following contributions. 1. Motivated by scientific theory change in the philosophy and history of science in the context of causality, we present an explanatory approach to model belief revision based on the notion causal relevance. However, our formulation of causal relevance is based on semantic considerations such as structure, intervention, causal process, and causal mechanism. 2. Development of a new entrenchment ordering namely the Causal Epistemic Entrenchment (CEE), suitable for prioritizing conditional beliefs in the belief revision process. It works better especially in preserving the core beliefs, and cases arising from iterated belief revision, theory choice, and the causal dependencies of beliefs. 3. We explored the link between causality and the dynamics of beliefs while emphasizing on causal consistency apart from the logicalconsistency.

Item Type:Thesis
Keywords:Belief revision, Epistemic Entrenchment, Ranking of beliefs, Causal beliefs,
Subjects:Philosophy > Logic
Computer Science > Artificial Intelligence
ID Code:6642
Deposited By:Ayyadevara, A Ravishankar
Deposited On:15 Oct 2009 23:53
Last Modified:15 Oct 2009 23:53

Metadata

Repository Staff Only: item control page