Jim Cunningham

Developing Agent Interaction Protocols Using Graphical and Logical Methodologies

Although interaction protocols are often part of multi-agent infrastructures, many of the published protocols are semi-formal, vague or contain errors. Formal presentations can counter such disadvantages since they are amenable to verification of correctness. On the other hand, a diagrammatic representation of system structure is easier to comprehend. To this end, this paper bridges the gap between formal specification and intuitive development by: (1) proposing an extended form of propositional dynamic logic for expressing protocols completely, with clear semantics, that can be converted to a programming language for interaction protocols and (2) developing extended statecharts as a diagrammatic counterpart.

Ensuring consistency in the joint beliefs of interacting agents

Agent interaction in realistic applications is subject to many forms of uncertainty -- including information and network uncertainty, trust of and conflicts with other participants, lack of stability in a deal and risks about agreements and commitments. However, one of the most common forms of uncertainty occurs when a group has divergent beliefs about the interaction they are engaged in -- some agents believe an agreement has been reached, while others believe it has been rejected or that they are still bargaining. Such misunderstandings can arise because of loss of network performance, spurious connections, message loss or delays. Against this background, this paper develops synchronisation protocols for a group of agents to attain the same beliefs about an interaction, independent of the reliability of the underlying communication layer. This paper includes and proves theorems about a groups mutual beliefs, on which the safety of an interaction relies. Specifically, protocols for message exchange and belief revision and the reasoning for reachability of states during interactions are presented. Each protocol is proved to show that an increasing level of mutual and consistent belief is reached, thereby guaranteeing an interactions integrity.

Normalized interactions between autonomous agents: a case study in inter-organizational project management

The CEC Project GOAL (Esprit 6283) aims to develop generic software tools to support a new project management paradigm, in which projects are collaborative, decentralised and inter-organizational. To support inter-organizational interaction, communication and cooperation, we are developing a design framework for formalizing the flow of information between organizations, specifying access to and provision of project services, and defining project-wide standards and procedures. This framework is based on normalizing interactions between autonomous software agents by specifying messages and protocols for inter-agent communication and cooperation. This paper reviews the framework, and then focusses on the specification and implementation of a case study, the automation of a distributed document review procedure. This is both a successful proof of concept and a demonstration of how Artificial Intelligence technologies can support inter-organizational project management. It also points the way to ‘agent brokering’, an enhancement of object brokering in distributed open systems, where the satisfaction of service requests can be subject to negotiation.

Distributed Modal Theorem Proving with KE

This paper describes an approach to distributed modal theorem proving by bringing together and exploiting two software packages. The first is the implementation of a theorem prover for normal modal logics based on KE and a generalization of Fittings prefixed tableaux. The second is a library for implementing brokered inter-process communication over the internet. We describe three demonstrators which combine these implementations and illustrate potential applications of the new technology, enabling theorem provers connected to the internet to cooperate, compete, or be used by third parties.

A formal framework for agent interaction semantics

Although informative, the semantic definition proposed for the most standard agent communication language (FIPA ACL 1997) is complicated and contentious, while published interaction protocols (IPs) tend to be ambiguous, incomplete, and unverified with respect to message semantics. To clarify and help rectify these problems, this paper proposes an integrated framework based on Propositional Dynamic Logic and Belief and Intention modalities (called the PDL-BI language). Specifically, we provide an axiomatisation of PDL-BI and for an agents propositional attitudes (beliefs and intentions) and social attitudes (such as sincerity and trustworthiness). Then we suggest a revised and simpler core semantics for many of the FIPA ACL speech acts, which, in turn, lead to the specification of the semantics of IPs. As a case study, we specify the semantics of the contract net protocol (CNP) in PDL-BI, which allows to prove that the CNP terminates.

Satisfying Requirements for Electronic Commerce

We first present the findings of 1996 research on on-line commercial activities showing a lack of technological support for such electronic commerce. In fact, there were few signs of trading and negotiation was virtually nonexistent on-line. We also give a case study of an information retrieval system that has the potential to become an on-line brokerage service, then identify and expand the key requirements for electronic commerce. We give an overview of the OSM1 architecture for an electronic market-place and we show how its framework may be used. The paper closes with a bibliography.

Attributed Translation and the Semantics of Natural Language

When Knuth introduced attribute grammars, he observed that although “oriented primarily towards programming languages, the same methods appear to be relevant also in the study of natural languages”. We demonstrate that his intuition is computationally justifiable, based on the algebraic equivalence of attribute grammars and Montagues theory of Universal Grammar. We discuss the relationship between attribute grammars, axiomatic theories and logic programming. We find that attribute grammars can be used to encode an algebraic specification of a natural language, attributed translation used to compute representations of the ‘meaning’ of a sentence at different levels of abstraction, and that the specifications can be implemented as logic programs. We illustrate the application of non-deterministic attributed translation to natural language by the specification of a subset of Montagues PTQ, including a treatment of some semantic ambiguities.

Modelling Linguistic Events

In this paper we investigate the logical representation of linguistic categories of tense and aspect with an emphasis on ease of computation. We show how temporal readings of simple sentences involving tense, the progressive and the perfect may be expressed concisely using a fragment of Halpern and Shoham’s interval logic HS, and describe how to obtain the readings using a compositional semantic approach. By modelling the readings as timelines we show how linguistic entailments may be checked simply and systematically.

Towards an axiomatic theory of consciousness

In this paper we seek to provide elements of an axiomatic theory for a sentient consciousness as a quantified form of introspective awareness. A crucial step for its formulation is use of an interval temporal logic to give expression to on-going conditions such as those represented by the progressive aspect in natural language. In this way we are able to enrich more stative mental models so that an agent’s internal activities and its perception of external processes can be represented more faithfully. The need for agent consciousness research is briefly discussed. 1 Some Pragmatics of Consciousness Although part of the ancient mind-body problem of philosophy, the concept of consciousness itself is well enough recognised for it to be an ordinary word of our language. A conscious individual is aware, and knowing; the unconscious condition is normally recognisable. Numerous popular and contemporary books by Aleksander [1], Dennett [4], Searle [13], and others, show its explication to be contentious and a challenge to our suppositions on reality, a hazardous topic indeed for a would-be engineer of artificial intelligence. Yet we must admit the possibility that consciousness has utilitarian function, evolved to ensure survival. Our justification for addressing the subject is that artificial agents which display elements of intelligent behaviour already exist, in the popular sense of these words, but that we would doubt the real intelligence of an agent which seemed to us to have no sense of “self”, or awareness of its capabilities and its senses and their current state. So although consciousness, in a sentient albeit non-emotive sense, seems more allied to awareness than reasoning, an approximation to human consciousness could enable us to converse more naturally with an individual agent. This makes it an unusual topic of enquiry because we need an account for the first person and second person perspective as well as the more usual third person of objective science. Contention arises over whether consciousness can be considered a mental state of the human mind, for this brings presuppositions of the intentional stance and issues of its faithfulness to the human brain. But lack of faithfulness to a biological model is not a barrier to engineering, as the wheel, the fixed wing, and the computer itself demonstrate. Software agents are already designed using notions of mental state and practical reasoning which have emerged as abstractions from rational enquiry rather than any physical brain model. While agent designers may also eschew such models, and instead rely on a variety of physical and computational devices, in well known cases the management of complexity leads to design architectures with layers of abstraction, some of which are comparable with intentional models of the mind. To bypass the metaphysics of consciousness in favour of pragmatic considerations, 341 L. J. of the IGPL, Vol. 9 No. 2, pp. 341–347 2001 c ©Oxford University Press 342 Towards an Axiomatic Theory of Consciousness there is evidence to consider, the view of peers in rational enquiry, and the need for guidance in an artificial construction. Clinical reports, psychological experiment, and philosophical enquiry lead to a variety of theories which partially explain the phenomena and suggest layers of consciousness [10]. Problematic issues range from neurological syndromes such as phantom limbs, denial of paralysis, and involuntary manipulation, through issues of identity and the effect of emotion and habit, to an explication of context and presence in perception and its links with language. Our formalism arises from attempts to bridge the gap between agents designed with mental states, and credible multi-processing implementations. It may be compatible with the implementations of a psychologically motivated, but non-sentient theory like that of [2], which can be realised as a computational agent with a myriad of heterogeneous processes. But to explicate conscious behaviour we certainly require layers of conception which we hardly discuss here. 2 Refining the Intentional Stance Mental models of the intentional stance encroach on two areas of agent design. One is as an abstract basis for incorporating plans and selecting actions through meansend reasoning in software agents, notably in variants of the Belief, Desire, Intention paradigm. (See, for instance, Bratman[3], Rao and Georgeff [12]). The other area is the related basis for giving definition to standard acts of communication as realisations of speech act theory, so that there are ingredients of a coherent basis for dialogue between agents in terms of what we can loosely call knowledge interchange. (See, for instance, Labrou and Finin [8]). Our proposal for steps towards an axiomatisation of consciousness depends critically on a refinement of traditional ideas of intentionality. From the perspective of an agent designer, extant intentional theories of rational agents focus on stative concepts of belief, desire, intention, knowledge and commitment, each of which can be regarded intuitively as expressing computational data states. Agent activities, or processes, which Vendler (1967) and later workers have considered equally important for the modelling of our linguistic descriptions of behaviour, have been ignored, or rather, buried in naive computational models. But activity states like planning, learning and sleeping, and the sensing and perceiving of external conditions allow a more refined computational model of rationality. Their absence is a serious deficiency in the usual perception of mental state. However, there is also another defect. The usual axiomisation for belief, and of knowledge, presumes introspection; e.g. for knowledge, that which is known is known, that which is not known is known to be not known. These are strong conditions which make such states already too “conscious” for some forms of memory recall and learnt behaviour. The limitations of stative mental states can be overcome simply by allowing activity states as well. Both stative and activity states can be considered durative on a temporal frame. They can be distinguished informally by the observation that a stative condition is basically atemporal, but becomes homogeneous on an interval as an artifact of a temporal frame, whereas an activity is essentially durative, a process which may be composed from sub-processes. However, we gloss over finer semantic issues by emphasising one facit, an explicit progressive expression for an activity, captured by a modal operator prog to modify a singular predicate for a dynamic 2. REFINING THE INTENTIONAL STANCE 343

Deriving agent-centred representations of protocols described using propositional statecharts

Diagrammatic methodologies for the representation of agent interaction protocols can be classified as joint representations that describe an interaction in terms of a single sequential process, or agent-centred representations that provide a distinct description of the interaction protocol for each agent (or role) in the interaction. Here we discuss the process of deriving agent centred representations from joint representations and vice versa, using a variant of UML statecharts.