TUTORIAL: Belief Change: Introduction and Overview

(updated 16/3/2024)

ECAI 2024, the 27th European Conference on Artificial Intelligence


-- The 1985 paper by Carlos Alchourrón (1931–1996), Peter Gärdenfors, and David Makinson (AGM), “On the Logic of Theory Change: Partial Meet Contraction and Revision Functions” was the starting-point of a large and rapidly growing literature that employs formal models in the investigation of changes in belief states and databases.

-- In this tutorial, the quite 40 years of the logic of theory change (belief change or belief revision) are summarized.

-- The topics covered include equivalent characterizations of AGM operations, extended representations of the belief states, change operators not included in the original framework, iterated change, applications of the model, its connections with other formal frameworks, and criticism of the model.


-- PDF format. (CURRENT FROM 2018 - TO BE UPDATED)

A detailed outline of the tutorial

1 Introduction

2 Equivalent characterizations

2.1 AGM briefly summarized
2.2 Safe and kernel contraction
2.3 Epistemic Entrenchment
2.4 Grove’s spheres
2.5 Distance models
2.6 Specified meet contraction

3 Criticism of the model

3.1 The recovery postulate
3.2 The success postulates
3.3 Are belief sets too large?
3.4 Lack of information in the belief set

4 Extended representations of belief states

4.1 Belief bases
4.2 Probability and plausibility
4.3 Ranking models
4.4 Extensions of the language
4.5 Change in norms, preferences, goals, and desires

5 Iterated change

5.1 Revising epistemic states
5.2 Major classes of iterable operators

6 Alternative operators of change

6.1 Update
6.2 Non-prioritized change
6.3 Changes in the strength of beliefs
6.4 Resource-bounded change and inconsistency management
6.5 Multiple change
6.6 Indeterministic change
6.7 Some other operators of change

7 Applications and connections

7.1 Non-monotonic and defeasible logic
7.2 Description logic
7.3 Horn clause contraction functions
7.4 Game theory
7.5 Argumentation
7.6 Modal and dynamic logics
7.7 Belief Change by translation between logics
7.8 Truth
7.9 Use of choice functions and related preference orderings

8 Computability and implementation

Potential target audience

--The tutorial is intended for a wide audience. The main area of the course is

-- Knowledge Representation, Reasoning and Logic

However, it is also interesting for the people in the following areas:

-- Agent-based and Multi-agent Systems

-- Machine Learning

-- Robotics and Vision (in particular Cognitive Robotics)

-- Multidisciplinary Topics and Applications

-- Web and Knowledge-based Information Systems

-- Uncertainty in AI


Eduardo Fermé

-- I worked from 1991 in the area of Knowledge Representation and Reasoning – Non Monotonic Reasoning - Belief Revision.
My focus of research is the area of Belief Revision (Logic of Theory Change): Belief Revision theory studies the impact of acquiring new information. It is a fundamental activity of human intelligence, and it defines an exciting and significant research area in philosophy, logic and computer science. Belief revision theory provides sound modellings for changes of beliefs in response to new information.
In particular I work in: 1. Belief Bases: To complete the representation theorems in the area of belief bases models. 2. Multiple Change: Define constructive models of multiple change to be applied in agents like robots. 3. Iterated Change and non-prioritized change: Define new models of iterated and non-prioritized change. 4. Applications: Apply belief revision algorithms to solve consistent updates problems in systems. Ex. Belief revision applied to systems for neurorehabilitation therapy.

-- Homepage

-- Orcid CV

-- Ciência Vitae CV

-- Google Scholar Profile

-- ResearchGate Profile

-- DBLP Profile


Faculty of Exact Sciences and Engineering - University of Madeira

Campus Universitário da Penteada 9020-105 Funchal, Madeira, Portugal.

ferme @ uma .pt

(erase blank spaces)