The literature on inconsistency management hosts a wide array of measures of inconsistency. In this paper, we argue that these measures are not particularly suitable for measuring the inconsistency of the knowledge base of a logical agent. Logical agents are typically limited in their ability to discover inconsistency. Moreover, what is usually at stake is the effect inconsistency may have on the agent's behavior. To that end, we address the issue of practical measures of incon-sistency for logical agents by (i) developing a general notion of decidable, resource-bounded entailment and (ii) studying inconsistency measurement in the context of an assumption-based reason-maintenance system which keeps track of the premises underlying various conclusions made by the agent. This allows us to develop inconsistency measures for logical agents with limited reasoning capabilities and limited ability to discover inconsistency.

With the advent of social robots, precise accounts of an increasing number of social phenomena are called for. Although the phenomenon of secrets is an important part of everyday social situations, logical accounts of it can only be found, in a rather strict sense, within logical investigations of systems security. This paper is an attempt to formalize the logic of a commonsense notion of secrets as a contribution to ontologies of social and epistemological phenomena. We take a secret to be a five-way relation between a proposition, a group of secret-keepers, a group of nescients, a condition of secrecy, and a time point. A bare-bones notion of secrets is defined by providing necessary and sufficient conditions for said relation to hold. Special classes of secrets are then identified by considering an assortment of extra conditions. The logical language employed formalizes a classical account of belief and intention, a theory of groups, and a novel notion of revealing. In such a rich theory, interesting properties of secrets are proved.

A knowledge base is an integral part of a logic-based artificial intelligence system. In this paper, we present a variety of algorithms for knowledge base size reduction. Our approach differs from previous approaches in at least three aspects. First, it takes its objects to be support-structured sets of unconstrained, rather than flat sets of syntactically-constrained, logical formulas, which we refer to as belief states. Second, classical notions of minimality and redundancy are replaced by weaker, resource-bounded alternatives based on the support structure. Third, in “lossy” variants of compression, the compressed knowledge base logically implies only a practically-relevant subset of the original knowledge base.

We present a framework for incorporating perception-induced beliefs into the knowledge base of a rational agent. Normally, the agent accepts the propositional content of perception and other propositions that follow from it. Given the fallibility of perception, this may result in contradictory beliefs. Hence, we model high-level perception as belief revision. We overcome difficulties imposed by the highly idealistic classical belief revision in two ways. First, we adopt a belief revision operator based on relevance logic, thus limiting the derived beliefs to those that relevantly follow from the new percept. Second, we focus belief revision on only a subset of the agent's set of beliefs—those that we take to be within the agent's current focus of attention.

Effect axioms constitute the cornerstone of formal theories of action in AI. They drive standard reasoning tasks, especially prediction. These tasks need not be coupled with actual acting; the reasoning agent is, thus, typically given an ex post acto narrative of what actions took place. An acting agent, however, has no access to such knowledge; it needs to face what we call the event categorization problem, and figure out what actions it did. Until this is achieved, effect axioms will be useless. A careful review of the literature on effect axioms reveals that their syntax, semantics, and ontological commitments are so deeply entrenched in the armchair reasoning about action paradigm, that they cannot be used in resolving the event categorization problem. By enriching the ontology of action theories, we propose a different approach for representing effects of actions that unifies the two views. The enriched ontology is independently motivated by linguistic concerns.