The needs of a real-time reasoner situated in an environment may make it appropriate to view error-correction and non-monotonicity as much the same thing. This has led us to formulate situated (or step) logic, an approach to reasoning in which the formalism has a kind of real-time self-reference that affects the course of deduction itself. Here we seek to motivate this as a useful vehicle for exploring certain issues in commonsensereasoning. In particular, a chief drawback of more traditional logics is avoided: from a contradiction we do not have all wffs swamping the (growing) conclusion set. Rather, we seek potentially inconsistent, but nevertheless useful, logics where the real-time self-referential feature allows a direct contradiction to be spotted and corrective action taken, as part of the same system of reasoning. Some specific inference mechanisms for real-time default reasoning are suggested, notably a form of introspection relevant to default reasoning. Special treatment of ...

Artificial intelligence research falls roughly into two categories: formal and implementational. This division is not completely firm: there are implementational studies based on (formal or informal) theories (e.g., CYC, SOAR, OSCAR), and there are theories framed with an eye toward implementability (e.g., predicate circumscription). Nevertheless, formal /theoretical work tends to focus on very narrow problems (and even on very special cases of very narrow problems) while trying to get them "right" in a very strict sense, while implementational work tends to aim at fairly broad ranges of behavior but often at the expense of any kind of overall conceptually unifying framework that informs understanding. It is sometimes urged that this gap is intrinsic to the topic: intelligence is not a unitary thing for which there will be a unifying theory, but rather a "society" of subintelligences whose overall behavior cannot be reduced to useful characterizing and predictive principles.

We propose a context-logic style formalism, Timed Reasoning Logics (TRL), to describe resource-bounded reasoners who take time to derive consequences of their knowledge. The semantics of TRL is grounded in the agent's computation, allowing an unambiguous ascription of the set of formulas which the agent actually knows at time t. We show that TRL can capture various rule application and conflict resolution strategies that a rule-based agent may employ, and analyse two examples in detail: TRL(STEP) which models an all rules at each cycle strategy similar to that assumed in step logic [4], and TRL(CLIPS) which models a single rule at each cycle strategy similar to that employed by the CLIPS [21] rule based system architecture. We prove a general completeness and decidability results for TRL(STEP).

Logics are powerful tools with which to reason about the behaviour of AI agents. A particular logic may provide an agent with an internal language with which it can represent and reason about the world, and act based upon its internal manipulation of this language. We can also use logics to reason about the behaviour of agents from an external perspective. For example, we could use a dynamic logic to represent the interaction between a simple deterministic automaton and its environment by introducing environment state update variables into whatever logic we use. The resulting logic may state true facts about the agent-environment pair, even if the agent does not make use of our dynamic variables. The agent may, for instance, have no internal reasoning capabilities. It may simply have its inputs hard-wired to its outputs, and so operate on no internal logic at all. However, we can still reason about the behaviour of the agent using complicated logics. We can reason about...

Abstract. Practical reasoners are resource-bounded—in particular they require time to derive consequences of their knowledge. Building on the Timed Reasoning Logics (TRL) framework introduced in [1], we show how to represent the time required by an agent to reach a given conclusion. TRL allows us to model the kinds of rule application and conflict resolution strategies commonly found in rule-based agents, and we show how the choice of strategy can influence the information an agent can take into account when making decisions at a particular point in time. We prove general completeness and decidability results for TRL, and analyse the impact of communication in an example system consisting of two agents which use different conflict resolution strategies. 1

Concern with the real-time nature of effective reasoning led us to develop a memory-based model of reasoning. Later efforts convinced us that a formal treatment of this approach would be fruitful. The current paper surveys the evolution of this work and discusses potential future research endeavors. 1 Motivation Traditional theoretical treatments of reasoning do not, in our view, address reasoning per se at all. Rather they seek to characterize the end results of reasoning. We have four complaints with this. First, it is not clear that reasoning has clearly identifiable end results, but rather goes on and on as part of the active history of an individual reasoner. Second, reasoning is a process or activity, and this is simply ignored in traditional studies. Third, as a consequence, crucial issues of temporal and spatial resources are not taken into account. Finally, if we are going to understand realizable intelligent systems, we must look at what they actually do, i.e., we must accep...

michal,bezem¡ We discuss briefly the duality (or rather, complementarity) of system descriptions based on actions and transitions, on the one hand, and states and their changes, on the other. We settle for the latter and present a simple language, for describing state changes, which is parameterized by an arbitrary language for describing properties of the states. The language can be viewed as a simple fragment of step logic, admitting however various extensions by appropriate choices of the underlying logic. Alternatively, it can be seen as a very specific fragment of temporal logic (with a variant of ‘until ’ or ‘chop ’ operator), and is interpreted over dense (possibly continuous) linear time. The reasoning system presented here is sound, as well as strongly complete and decidable (provided that so is the parameter logic for reasoning about a single state). We give the main idea of the completeness proof and suggest a wide range of possible applications (action based descriptions, active logic, bounded agents), which is a simple consequence of the parametric character of both the language and the reasoning system. 1

Reason-based actions plunge the reasoner into temporal considerations from all angles. We see this not only when time enters explicitly into the problem statement, but also in formal robot blocks-world scenarios, in the Yale Shooting Problem and other associated versions of the frame problem (e.g., [Hanks and McDermott, 1986] ), in various specialized actions (e.g., hiding, as in [Allen, 1984] ), and so on. In short, where there is action, there is time, and where there is time, there is a potential need for reasoning about time. Where, then, is the action? It certainly includes the usual overt physical acts of motion, and also certain covert behaviors such as hiding or watching. In these of course time is important. But there is another angle that is not usually noted, one that we have been exploring for the past several years ([Drapkin and Perlis, 1986b] , [Drapkin and Perlis, 1986a] , [ElgotDrapkin, 1988] , [Elgot-Drapkin and Perlis, 1990] ). Namely, action also occurs in the form o...

Abstract. We propose a two-fold use of context in reasoning about agents. The first concerns the modelling of beliefs which are entertained only within the scope of some assumption. We illustrate this use by describing as logical model of an agent which reasons in a natural deduction style by making assumptions to derive new formulas. The second use of context we propose here concerns the modelling of reasoning as a step-by-step temporal process, allowing us to model agents whose resources are temporally bounded. In such a setting, the beliefs ascribed to an agent need not be closed under consequence, as they are in many traditional epistemic logics. 1

We argue that there are not one, and not many, but rather two frame problems. One frame problem is simply a computational one, the other conceptual. We argue that reasoning in the commonsense world necessarily involves the conceptual problem. This is closely linked to (and even subsumed by) the familiar qualification problem in philosophy which, in turn, is related to the problem of natural and artifactual kinds.-2-I. Introduction Descriptions and explanations of ‘‘the frame problem’ ’ seem to be as numerous and as var-ied as those of Artificial Intelligence (AI) itself. We argue, however, that there are essentially two different frame problems. One pertains to a mathematically precise world, and the other to the less precise common-sense world. This distinction seems not to have been examined before, and appears to be the source of some confusion in discussions of the frame problem. Dealing with the first type of frame problem is simi-