Belief merging aims at extracting a coherent and informative view from a set of belief bases. A first requirement for belief merging operators is to obey basic rationality conditions. Another expected property is to preserve as much information as possible from the input bases. In this paper, we show how new merging operators, called compositional operators, can be defined from existing ones. Such operators aim at offering a higher discriminative power than the merging operators on which they are based, without leading to a complexity shift or losing rationality postulates. We identify some sufficient conditions for ensuring that rationality is fully preserved by composition.

An agent will generally have incomplete and possibly inaccurate knowledge about its environment. In addition, such an agent may receive erroneous information, perhaps in being misinformed about the truth of some formula. In this paper we present a general approach to reasoning about action and belief change in such a setting. An agent may carry out actions, but in some cases may inadvertently execute the wrong one (for example, pushing an unintended button). As well, an agent may sense whether a condition holds, and may revise its beliefs after being told that a formula is true. Our approach is based on an epistemic extension to basic action theories expressed in the situation calculus, augmented by a plausibility relation over situations. This plausibility relation can be thought of as characterising the agent's overall belief state; as such it keeps track of not just the formulas that the agent believes to hold, but also the plausibility of formulas that it does not believe to hold. The agent's belief state is updated by suitably modifying the plausibility relation following the execution of an action. We show that our account generalises previous approaches, and fully handles belief revision, sensing, and erroneous actions.

In Belief Revision the new information is generally accepted, following the principle of primacy of update. In some case this behavior can be criticized and one could require that some new pieces of information can be rejected by the agent because, for instance, of insufficient plausibility. This has given rise to several approaches of non-prioritized Belief Revision. In particular (Hansson et al. 2001) defined credibility-limited revision operators, where a revision is accepted only if the new information is a formula that belongs to a set of credible formulas. They provide several representation theorems in the AGM style. In this work we study credibility-limited revision operators when the information is represented in propositional logic, like in the Katsuno and Mendelzon framework. We propose a set of postulates and a representation theorem for credibility-limited revision operators. Then we explore how to generalize these definitions to the Iterated Belief Revision case, using epistemic states in the Darwiche and Pearl style.

Belief change studies how to update knowledge bases used for reasoning. Traditionally belief revision has been based on full propositional logic. However, reasoning with full propositional knowledge bases is computationally hard, whereas reasoning with Horn knowledge bases is fast. In the past several years, there has been considerable work in belief revision theory on developing a theory of belief contraction for knowledge represented in Horn form. Our main focus here is the computational complexity of belief contraction, and, in particular, of various methods and approaches suggested in the literature.

Belief revision is an operation that aims at modifying old beliefs so that they become consistent with new ones. The issue of belief revision has been studied in various formalisms, in particular, in qualitative algebras (QAs) in which the result is a disjunction of belief bases that is not necessarily representable in a QA. This motivates the study of belief revision in formalisms extending QAs, namely, their propositional closures: in such a closure, the result of belief revision belongs to the formalism. Moreover, this makes it possible to define a contraction operator thanks to the Harper identity. Belief revision in the propositional closure of QAs is studied, an algorithm for a family of revision operators is designed, and an open-source implementation is made freely available on the web.

In this paper we develop a general framework that allows for both knowledge acquisition and forgetting in the Situation Calculus. Based on the Scherl and Levesque (Scherl and Levesque 1993) possible worlds approach to knowledge in the Situation Calculus, we allow for both sensing as well as explicit forgetting actions. This model of forgetting is then compared to existing frameworks. In particular we show that forgetting is well-behaved with respect to the contraction operator of the well-known AGM theory of belief revision (Alchourron, Gardenfors, and Makinson 1985) but that knowledge forgetting is distinct from the more commonly known notion of logical forgetting (Lin and Reiter 1994).

A central result in the AGM framework for belief revision is the construction of revisionfunctions in terms of total preorders on possible worlds. These preorders encode comparative plausibility: r r' states that the world r is at least as plausible as r'. Indifference in the plausibility of two worlds, r, r', denoted r r', is defined as the absence of a preference between r and r'. Herein we take a closer look at plausibility indifference. We contend that the transitivity of indifference assumed in the AGM framework is not always a desirable property for comparative plausibility. Our argument originates from similar concerns in preference modelling, where a structure weaker than a total preorder, called a semiorder, is widely consider to be a more adequate model of preference.

In this paper, we investigate the revision of argumentation systems à la Dung. We focus on revision as minimal change of the arguments status. Contrarily to most of the previous works on the topic, the addition of new arguments is not allowed in the revision process, so that the revised system has to be obtained by modifying the attack relation only. We introduce a language of revision formulae which is expressive enough for enabling the representation of complex conditions on the acceptability of arguments in the revised system. We show how AGM belief revision postulates can be translated to the case of argumentation systems. We provide a corresponding representation theorem in terms of minimal change of the arguments statuses. Several distance-based revision operators satisfying the postulates are also pointed out, along with some methods to build revised argumentation systems. We also discuss some computational aspects of those methods.

This paper tackles the problem of evaluating the degree of inconsistency in spatial and temporal qualitative reasoning. We first introduce postulates to propose a formal framework for measuring inconsistency in this context. Then, we provide two inconsistency measures that can be useful in various AI applications. The first one is based on the number of constraints that we need to relax to get a consistent qualitative constraint network. The second inconsistency measure is based on variable restrictions to restore consistency. It is defined from the minimum number of variables that we need to ignore to recover consistency. We show that our proposed measures satisfy required postulates and other appropriate properties. Finally, we discuss the impact of our inconsistency measures on belief merging in qualitative reasoning.

Belief propagation over Markov random fields has been successfully used in many AI applications since it yields accurate inference results by iteratively updating messages between nodes. However, its high computation costs are a barrier to practical use. This paper presents an efficient approach to belief propagation. Our approach, Quiet, dynamically detects converged messages to skip unnecessary updates in each iteration while it theoretically guarantees to output the same results as the standard approach used to implement belief propagation. Experiments show that our approach is significantly faster than existing approaches without sacrificing inference quality.