Showing papers by "Munindar P. Singh published in 1999"

An ontology for commitments in multiagent systems

Abstract: Social commitments have long been recognized as an important concept for multiagent systems. We propose a rich formulation of social commitments that motivates an architecture for multiagent systems, which we dub spheres of commitment. We identify the key operations on commitments and multiagent systems. We distinguish between explicit and implicit commitments. Multiagent systems, viewed as spheres of commitment (SoComs), provide the context for the different operations on commitments. Armed with the above ideas, we can capture normative concepts such as obligations, taboos, conventions, and pledges as different kinds of commitments. In this manner, we synthesize ideas from multiagent systems, particularly the idea of social context, with ideas from ethics and legal reasoning, specifically that of directed obligations in the Hohfeldian tradition.

Agents for process coherence in virtual enterprises

Abstract: Open environments such as the Internet— and even corporate intranets— enable a large number of interested parties to use and enhance vast quantities of information. These environments support modern applications, such as manufacturing, virtual enterprises, and ubiquitous information access, which involve a number of information sources and component activities. However, without principled techniques to coordinate the various activities, any solutions would yield disjointed and error-prone behavior, while consuming excessive effort to build and maintain. The agent metaphor, long in study in artificial intelligence, has recently become popular in mainstream computing, largely due to its suitability for open environments. Agents can be thought of as active objects with some special properties tailored to open environments. For our purposes, the key aspects of agents are their autonomy and abilities to perceive, reason, and act in their environment, as well as to socially interact and communicate with other agents [7]. When agents interact with one another they form a multiagent system. As part of a multiagent system, agents can capture and apply the semantic constraints among heterogeneous components in order to enact distributed workflows. Autonomy is critical in open environments. Consider a manufacturing scenario requiring supply-chain coordination. It is natural to model independent companies in a supply chain as represented by autonomous agents. But, at first sight, autonomy is a mixed blessing if the companies behaved arbitrarily, the supply chain would break. Consequently, our main technical challenge is to manage autonomy that is, how to maximize freedom without letting it devolve into chaos. We propose that the main basis for managing autonomy lies in the notion of commitments. A flexible formulation of commitments can provide a natural means through which autonomous agents may voluntarily constrain their behavior. By flexible, we mean that it should be possible to cancel or otherwise modify the commitments. Consider a situation where a purchaser is trying to obtain some parts from a vendor. We would like the vendor to commit to delivering the parts of the right quality to the purchaser. However, it is important that the supply chain be able to survive exceptions such as when the manufacturing plant goes down in an earthquake, or when the purchaser decides that it needs the parts to be of a lower error tolerance than initially ordered. Information cannot be understood independently of the processes that create or consume it. The desired flexibility of behavior and the ability to recover from failures require an approach that is sensitive to how those processes interact. We show that when agents are associated with each independent process, our flexible notion of commitments can capture the desired interactions among those processes

Multiagent systems for workflow

Abstract: Workflows are ubiquitous in business computing. They arise not only within an enterprise, but increasingly across enterprises as well—in situations such as virtual enterprises and applications such as supply-chain management. Although the importance of workflows as a basis for understanding and automating business activities is widely recognized, current workflow practice leaves much to be desired. To a large extent, this problem arises because of the rigidity of current technology, which does not accord well with the complex, heterogeneous, dynamic environments in which workflows are applied. Agent technology promises to alleviate many of these problems and hence enable adaptive workflows in realistic settings. We consider interaction-oriented programming (IOP), an approach to software engineering based on multiagent systems that we have been developing. We focus on one aspect of IOP, which deals with social commitments and enables agents to flexibly enact a multienterprise workflow by entering into and behaving according to their commitments to each other. The agents can cancel or modify their base-level commitments only if they satisfy the metacommitments that then go into effect.

Be Patient and Tolerate Imprecision: How Autonomous Agents can Coordinate Effectively

Abstract: A decentralized multiagent system comprises agents who act autonomously based on local knowledge. Achieving coordination in such a system is nontrivial, hut is essential in most applications, where disjointed or incoherent behavior would be undesirable. Coordination in decentralized systems is a richer phenomenon than previously believed. In particular, five major attributes are crucial: the extent of the local knowledge and choices of the member agents, the extent of their shared knowledge, the level of their inertia, and the level of precision of the required coordination. Interestingly, precision and inertia turn out to control the coordination process. They define different regions within each of which the other attributes relate nicely with coordination, but among which their relationships are altered or even reversed. Based on our study, we propose simple design rules to obtain coordinated behavior in decentralized multiagent systems.

The E-Commerce Inversion

Abstract: about everyone seems to believe that business-to-consumer e-commerce (B2C) is a solved problem: you set up a Web site with the right forms, add a bit of security, establish accounts with a payment service and a shipper, maybe find a trusted server to certify your transactions, and you're ready to go. And in a way, everyone is right, because that's about all it takes to get started in e-business. Depending on which variant you look at, even business-to-business e-commerce (B2B) might seem like a solved problem. Like B2C, B2B is simpler when your site sells goods from just one manufacturer or vendor and harder when your site sells goods from a number of manufacturers or vendors. In the latter case, complexity arises from the need to maintain a uniform representation for goods from different sources and match the information flows with the materiel flows. While B2B and B2C e-commerce share similar problems, many people believe that B2B is much harder than B2C. As evidence, they point out that there are several general B2C sites but none such B2B sites. B2B sites sprout mostly in vertical nich-es, dealing, for example, with lab equipment or agricultural chemicals. Setting up a B2B site is deemed harder today because all e-commerce requires the ability to structure and present information properly to users. In typical B2B domains, there is far more complexity to represent and display than in typical consumer domains. A standard office supplies catalog will have many more entries than an ordinary clothing store catalog—there are probably more kinds of notebooks, for example, than men's shirts. Thus B2B domains require more features and impose more subtle constraints. Just the Numbers Part of the increasing interest in B2B e-commerce is due to widely cited studies indicating that it will generate revenues exceeding those of B2C by an order of magnitude or more. I don't question this view for the near term, although fundamentally it is puzzling, and one has to wonder how long it will hold. Whatever companies buy from each other is eventually sold to consumers or is used in the process of developing something else for the consumer market. The only exceptions would be public works and military expenditures, which don't directly result in consumer spending. Except for these two categories , consumer spending should exceed business spending. If B2C revenues don't match those of B2B, then conventional commerce must be picking …

Stored data object management and archive control

Abstract: A method for maintaining data objects in memory introduces metrics for freshness and interestingness of the data object. A freshness estimate is defined as the length of time that the data contained within the data object is expected to remain correct from the date that the object was last known to be valid. The interestingness estimate is defined as the length of time that the data object is expected to retain its usefulness to the user from the time that the data object was last used. The freshness estimate and interestingness estimate are stored in the data object along with a reference date. The freshness estimate, interestingness estimate, and reference date are updated or confirmed each time the data object is accessed. A maintenance program is periodically invoked to determine whether the data objects are outdated by comparing the time estimates associated with each data object to that object's reference date. Objects found to be outdated are candidates for removal from the system.

Intelligent agents V : agent theories, architectures, and languages : 5th International Workshop, ATAL'98 : Paris, France, July 4-7, 1998 : proceedings

Abstract: Belief-Desire-Intention.- The Belief-Desire-Intention Model of Agency.- BDI Models and Systems: Reducing the Gap.- Information-Passing and Belief Revisionin Multi-agent Systems.- On the Relationship between BDI Logics and Standard Logics of Concurrency.- Intention Reconsideration Reconsidered.- Making SharedPlans More Concise and Easier to Reason About.- Theories.- Autonomous Norm Acceptance.- Moral Sentiments in Multi-agent Systems.- Social Structure in Artificial Agent Societies: Implications for Autonomous Problem-Solving Agents.- The Bases of Effective Coordination in Decentralized Multi-agent Systems.- A Model Checking Algorithm for Multi-agent Systems.- Compositional Verification of Multi-agent Systems in Temporal Multi-epistemic Logic.- Emergent Mental Attitudes in Layered Agents.- Architectures.- The Right Agent (Architecture) to Do the Right Thing.- Representing Abstract Agent Architectures.- HEIR - A Non-hierarchical Hybrid Architecture for Intelligent Robots.- A-Teams: An Agent Architecture for Optimization and Decision-Support.- Goal-Satisfaction in Large-Scale Agent Systems: A Transportation Example.- Task Decomposition and Dynamic Role Assignment for Real - Time Strategic Teamwork.- Languages.- Agent Languages and Their Relationship to Other Programming Paradigms.- A Survey of Agent-Oriented Methodologies.- The Agentis Agent InteractionModel.- Content-Based Routing as the Basis for Intra-Agent Communication.- Agent Communication Language: Towards a Semantics based on Success, Satisfaction, and Recursion.- Control Structures of Rule-Based Agent Languages.- A Reactive Approach for Solving Constraint Satisfaction Problems.- Increasing Resource Utilization and Task Performance by Agent Cloning.- An Index to Volumes 1-5 of the Intelligent Agents Series.- An Index to Volumes 1-5 of the Intelligent Agents Series.

Social Abstractions for Information Agents

Abstract: Most of the modern applications of computing technology arise in large-scale, open, information-rich environments Open environments are distinguished by fact of having a number of networked and interrelated, but heterogeneous, information resources The applications include ubiquitous information access, electronic commerce, virtual enterprises, logistics, and sensor integration, to name but a few These applications differ from conventional database applications not only in the nature and variety of information they involve, but also in including a significant component that is beyond the information system An information system must inevitably perform its immediate duties: create, store, transform, use, and dispose of information However, as the above applications suggest, there is a large world outside the realm of pure information where reside physical and economic objects and organizational processes that manipulate them

Multiagent treatment of agenthood

Abstract: There have been numerous attempts to provide a standardized definition ofa computational agent, but little consensus has emerged. We propose a simple test for agenthood that can be applied to a putative computational agent. Roughly, this test seeks to capture the intuition that an agent is an entity that can function as part ofa multiagent system. The test depends on the observed behavior of the supposed agent and not on the internals of it. We apply the test to some well-known kinds of systems of (supposed) agents and discuss the results. We present a formulation of the test and some variants with a semantics based on sociability. Our treatment of agenthood can thus serve as a methodological basis for evaluating putative agents and agent toolkits.

E-commerce over communicators: challenges and solutions for user interfaces

Abstract: which are substantially larger than those reached by traditional computers. Communicators are small handheld devices that combine the features of PDAs and palmtop computers with those of telephones. Communicators are widely believed to be the next big wave of devices with projected worldwide market penetration exceeding that of conventional desktop computers. Because of their being ready at hand as people go about their daily business, Communicators provide a natural platform on which to conduct e-commerce. However, Communicators pose major technical challenges because of their physical limitations and likely modes of usage. They offer small screens for presentation, difficult input, low bandwidth, and intermittent and unreliable connectivity. Further, people use their Communicators while on the run, in awkward settings, and may break, lose, or have their Communicators stolen. We describe an approach that addresses the above challenges while retaining the convenience of using Communicators. This approach integrates innovations in user interfaces and backend systems for information access and transactions. Norman argues forcefully that information appliances (or “invisible computers”) of the sort of communicators can fundamentally be more usable and thus more popular than traditional computers [5]. We agree with this assessment and think of communicators and their relatives as becoming ever more popular. (We revisit other aspects of Norman’s position in section 4.4.) As communicators (and related devices such as palmtops and PDAs) become more prevalent, it is natural that they will be used to conduct electronic commerce. Indeed, because communicators are carried by people all the time, in many cases, communicators are likelier to be used for e-commerce than traditional desktop computers. E-Commerce involves a number of activities. Of these, the following are especially relevant from the perspective of the consumer:

Conceptual Modeling for Multiagent Systems: Applying Interaction-Oriented Programming

Abstract: Multiagent systems (MAS) are an important paradigm for building complex systems, especially cooperative information systems. Despite much interest in MAS construction, there has not been sufficient progress on the corresponding conceptual modeling representations and techniques. We believe that further extensions to conceptual modeling to include aspects of actions and organizations will be essential for MAS development. These goals are broader than conceptual modeling is traditionally understood, but are essential to deal with the interactive and dynamic aspects of modern applications. We describe an approach termed interaction-oriented programming, which incorporates functionality geared toward coordination, commitment management, and collaboration. This functionality is naturally thought of as providing a conceptual metamodel for describing MAS. We suggest some preliminary methodologies pertaining to the design of coordination and commitment requirements.

Write asynchronous, run synchronous

Abstract: Last issue, I wrote about interaction via " live, " or dynamic, documents as analogs of communication. This time I will probe the concept of communication a little further. Why should an Internet specialist care about communication? Because, fundamentally, communication is all you ever do with the Internet. And communication is more than just bit transport. Communication happens not only when you talk to someone, but also when you work together or carry out a trade. This isn't simply because information is passed around in these activities, but because the activities are based upon the parties interacting in certain ways. Likewise, in the realm of computing , we must think of communication as more than just bit transport, and consider explicitly the interactions among the computations. Briefly, my claim is that the forms of communication we see in current programming models are limited and, consequently, so are the applications we can build over the Internet. If we are to engineer the next generation of applications—from negotiation-based electronic commerce to tools for people to come together in online communities— we must come up with higher level abstractions than, say, push and pull. These abstractions will prove as important as having interpretive programming environments, such as Java, that support the " write once, run anywhere " paradigm. Synchrony vs. Asynchrony There are two main kinds of communication: synchronous and asynchronous. Sometimes we distinguish synchronous communication by calling it " real time, " but in my view, synchrony is not about timing guarantees but about the communicating parties being simultaneously engaged in communication. Over the Internet, asynchrony is the default and synchrony is something you have to work for. However, syn-chrony is simpler both conceptually and computationally. Under synchronous communication, the parties involved share more of their context and can thus make stronger assumptions about each other. By contrast, under asynchronous communication, none of the participants knows for sure when the other participants will be engaged, and thus the communications must have a meaning that to some extent is independent of the context in which they were created. That's why programming under the assumptions of synchrony is easier and that's why many of the earlier programming models assumed synchrony. Expressiveness vs. Reusability Independence from context is essential to reuse, whereas sensitivity to context is essential for capturing the nuances of dynamic situations. For example, a traditional electronic commerce system might provide forms for …

Value-Oriented Electronic Commerce

Abstract: Internet is abuzz with electronic commerce. All too often these days, however, electronic commerce is little more than a euphemism for a Web site with an online catalog and the minimal technology needed to have buyers fill out forms to place orders. As with any buzzword, \" electronic commerce \" will keep getting used for the most-restrictive, least-common denominator scenarios. That is only to be expected. However, commerce involves a number of important activities besides the customer reviewing catalogs and placing orders, and the seller creating catalogs and shipping goods. Some upcom-ing approaches concentrate on how the customer and seller may negotiate on a price, how they may participate in an auction and come to an agreement about the prices at which the merchandise ought to move. These are all useful, if not essential, components of electronic commerce. But let's look beyond them to see what we can achieve over the Internet. Price-Oriented Electronic Commerce It is said that an economist is a person who knows the price of everything and the value of nothing. It looks like there are a lot of economists on the Internet. What I mean is that many people seem to find simple market mechanisms such as auctions enticing. Accordingly, many auction sites have been set up, and a lot of current research on electronic commerce emphasizes auctions. While the early research considered the familiar kinds of auctions, the newer work deals with more complex forms that can handle more realistic problems. Approaches are emerging that can handle bundled goods respecting the constraints among goods that someone may wish to purchase. An example of a bundled good is a trip that includes an air ticket and a hotel stay. Simply put, markets are an approach to achieve resource allocation among a group of interested parties. These parties may be distributed, although the markets themselves are centralized. As Herb Simon (who won both the Turing Award for his work on computer science and the Nobel Prize for his work on economics and organizations) observed, if markets have one good property, it is that they clear. In other words, they find a buyer for each piece of merchandise by matching the buyers' and sellers' prices. For this reason , I think auctions will settle down as a useful computational metaphor for various kinds of resource allocation problems. I rather suspect, however, that auctions will find most of …

Handling Semantic Exceptions in the Large: A Multiagent Approach

Abstract: We consider semantic exceptions, which arise when a task yields results that are incorrect, inconsistent with related tasks, or incomplete. Semantic exceptions are especially prominent in the large, i.e., when we construct and execute a workflow. A workflow is a composite computation with several interoperating components and interacting processes. Detecting and resolving semantic exceptions is critical to the functioning of a workflow, especially when its member activities are autonomous, heterogeneous, long-lived, and interact in subtle ways. Unfortunately, present workflow techniques offer little support for exceptions. For modeling, they provide few abstractions beyond activity charts. For enactment, they are not flexible enough to allow a component to detect and resolve semantic exceptions properly. We describe a multiagent approach for handling semantic exceptions. Our approach is based on high-level abstractions such as commitments, a process metamodel that accommodates commitments and allied concepts, a behavior model to specify agents, and an execution architecture that handles persistent and dynamic (re)execution. We can handle a variety of semantic exceptions by formulating a number of commitment patterns that cover the important situations. The behavior model and the commitment patterns are expressed as statecharts and executed in a rule-based system. In this way, advanced abstractions based on multiagent systems are mapped into conventional software techniques. Topic: Models and paradigms This research has been supported by the U.S. National Science Foundation under grants IIS-9529179 and IIS9624425, and by IBM Corporation.

The true Meaning Of Analog

Abstract: only its newest applications but also its infrastructure are inherently interactive. When you buy a book over the Internet, you interact at the application level; when your router forwards along a packet, it interacts with its peers. However, these interactions are not especially sophisticated—query-response in the first case and forwarding in the second. Therein lies the problem. The paradox of Internet computing is that although the Internet requires interactivity and the underlying technology can support it, our architectures and paradigms are still based mostly in design for isolated activities. Any interaction is of the least-common-denominator variety. Documents as Communication A simple example of interaction, related to the present theme issue, occurs in the notion of documents , which underlie much of what happens in the Web. Documents used to be just pieces of text, typically tied to a specific presentation and to the physical medium of paper. The arrival of hypertext and its realization on the Web opened documents up to varieties of structure and to the separation of content from presentation, and presentation from the physical medium. Well, almost. As a community of techies, we haven't quite separated content from style and we haven't quite nailed down the metadata. But, as the present issue shows, we are making progress. Let's step back and consider what documents really are. They are not just bytes or files, but ways of communicating. As currently realized, documents are a peculiar way of communicating. Someone viewing a document is essentially subjected to a long, possibly annoying monologue. To become analogs of real human communication, documents should be modeled as dialogues: They should be interactive, be presented incremen-tally, and carry a deep argument— rhetorical—structure. At the same time, they ought to go beyond usual verbal communication, because unlike normal human conversation, documents are meant to persist and to be communicated or \" played back \" in different contexts. The viewers of a document may have different needs for information and may be working on platforms of different connectivity. For instance, imagine a document consisting of a variety of media types: text, video, or whatever. The document might be making an argument, say, about the weather in the Raleigh area. Based on the structure of the argument, the available bandwidth, and what the viewer cares about, the document might be streamed and presented as a set of temperature ranges, textual excerpts from magazines, still images, or …

Information management for co­ operative engineering

Abstract: In this paper we describe how a set of autonomous computational agents can co­ operate in providing coherent management of information in environments where there are many diverse information resources. The agents use models of themselves and of the resources that are local to them. Resource models may be the schemas of databases, frame systems of knowledge bases, or process models of business operations. Models enable the agents and resources to use the appropriate semantics when they inter-operate. This is accomplished by specifying the semantics in terms of a common ontology. We discuss the contents of the models, where they come from, and how the agents acquire them. We then describe a set of agents for telecommunication service provisioning and show how the agents use such models to co-operate. Their interactions produce an implementation of relaxed transaction processing.

Information Management for Cooperative Engineering

Abstract: In this paper we describe how a set of autonomous computational agents can cooperate in providing coherent management of information in environments where there are many diverse information resources. The agents use models of themselves and of the resources that are local to them. Resource models may be the schemas of databases, frame systems of knowledge bases, or process models of business operations. Models enable the agents and resources to use the appropriate semantics when they inter-operate. This is accomplished by specifying the semantics in terms of a common ontology. We discuss the contents of the models, where they come from, and how the agents acquire them. We then describe a set of agents for telecommunication service provisioning and show how the agents use such models to co-operate. Their interactions produce an implementation of relaxed transaction processing.