Abstract: In this overview paper, I will attempt to identify and describe some of the common threads that tie together work in reasoning about knowledge in such diverse fields as philosophy, economics, linguistics, artificial intelligence, and theoretical computer sciencce. I will briefly discuss some of the more recent work, particularly in computer science, and suggest some lines for future research.

Over the years, many different agent programming languages have been proposed. In this paper, we propose a concept called Agent Programs using which, the way an agent should act in various situations can be declaratively specified by the creator of that agent. Agent Programs may be built on top of arbitrary pieces of software code and may be used to specify what an agent is obliged to do, what an agent may do, and what an agent may not do. In this paper, we define several successively more sophisticated and epistemically satisfying declarative semantics for agent programs. We further show that agent programs cleanly extend well understood semantics for logic programs, and thus are clearly linked to existing results on logic programming and nonmonotonic reasoning.

Classical planning started with goals and produced plans. To do something similar in a reactive framework, it is necessary to treat the achievement or maintenance of a goal as specifying a default behavior, while at the same time being able to deploy tools to generate more complex plans. These tools rely on representing the relations between goals and plan patterns that achieve them. An important component of the representation is a set of fast, local methods for making probabilistic estimations of the quality of proposed plans, including their robustness, completeness, and efficiency. 1 Introduction A reactive plan is a program that specifies how an agent is to react to its environment. There is currently no widely accepted theory of how to generate reactive plans, or even if it makes sense for an agent to have an explicit plan when there are a lot of events for it to react to quickly. If things are happening too fast for the agent to compare the expected consequences of alternative...

Agents are small programs that autonomously take actions based on changes... In this paper, we propose the concept of a probabilistic agent program and show how, given an arbitrary program written in any imperative language, we may build a declarative "probabilistic" agent program on top of it which supports decision making in the presence of uncertainty. We provide two alternative semantics for probabilistic agent programs. We show that the second semantics, though more epistemically appealing, is more complex to compute. We provide sound and complete algorithms to compute the semantics of positive agent programs.

Abstract not found

In [17], two of the authors have introduced techniques to build agents on top of arbitrary data structures, and to "agentize" new/existing programs. They provided a series of successively more sophisticated semantics for such agent systems, and showed that as these semantics become epistemically more desirable, a computational price may need to be paid. In this paper, we identify a class of agents that are called weakly regular---this is done by first identifying a fragment of agent programs [17] called weakly regular agent programs (WRAPs for short).

Intentions are an important concept in Artificial Intelligence and Cognitive Science. We present a formal theory of intentions...

Qualitative decision tools have been used in AI and CS in various contexts. However, their adequacy is unclear. Following Brafman and Tennenholtz, we use the axiomatic approach to investigate the adequacy and usefulness of various decision rules. We present constructive representation theorems for a number of qualitative decision criteria, including minmax regret , competitive ratio, and maximax , and characterize conditions under which a maximin agent can be ascribed qualitative beliefs. Introduction Decision theory plays a central role in various disciplines, including mathematical economics, game theory, operations research, industrial engineering, and statistics. It is widely recognized by now that decision making is crucial to AI as well, since artificial agents are, in fact, automated decision makers (RN95). However, many decision making techniques found in the AI literature are quite different from those found in other fields. Work in other disciplines has mostly adopted the v...

Recently there has been a good deal of interest in using techniques developed for learning from reinforcement to guide learning in robots. Motivated by the desire to find better robot learning methods, this thesis presents a number of novel extensions to existing techniques for controlling exploration and inference in reinforcement learning. First I distinguish between the well known exploration-exploitation trade-off and what I term exploration for future exploitation. It is argued that there are many tasks where it is more appropriate to maximise this latter measure. In particular it is appropriate when we want to employ learning algorithms as part of the process of designing a controller. Informed by this insight I develop a number of novel measures of the agent's task knowledge. The first of these is a measure of the probability of a particular course of action being the optimal course of action. Estimators are developed for this measure for boolean and non-boolean processes. These...

As part of an effort to define a unified formal semantics for beliefs, desires and action, this paper sketches a model theory for the axiological aspects of agent theory: hedonic states, likes, goals and values. Particular attention is paid to modelling the intensity of likes. The main intuition underlying the model theory is that the axiological aspects of agent theory can be modelled through computational generalisations of physical dynamics. Computational analogues of force, mass and potential are offered. Introduction An important part of agent theory appears to be the notion of desires. Several formulations of agent theory have adopted beliefs, desires and intentions as a set of basic notions (the so-called BDI models). However, to our knowledge, so far relatively little has been said explicitly in the AI literature about a theory of desires (Cohen and Levesque, 1985 and in press, Moore, 1985a; Kiss, 1988, Shoham, 1989). This paper takes some initial steps towards the explicit f...