Autonomous Agents & Multiagent Systems/Agent Theories, Architectures, and Languages, 2008
Service matchmaking is the process of finding suitable ser- vices given by the providers for the ... more Service matchmaking is the process of finding suitable ser- vices given by the providers for the service requests of con- sumers. Previous approaches to service matchmaking is mostly based on matching the input-output parameters of service requests and service provisions. However, such ap- proaches do not capture the semantics of the services and hence cannot match requests to services eectively.
Autonomous Agents & Multiagent Systems/Agent Theories, Architectures, and Languages, 2008
Communication among agents requires a common vocabulary to facilitate successful information exch... more Communication among agents requires a common vocabulary to facilitate successful information exchange. One way to achieve this is to assume the existence of a common ontology among communi- cating agents. However, this is a strong assumption, because agents may experience situations that result in independent evolution of their ontologies. When this is the case, agents need to form com- mon
Information systems must establish trust to cooperate effectively in open environments. We are de... more Information systems must establish trust to cooperate effectively in open environments. We are developing an agent-based approach for establishing trust, where information systems are modeled as agents that provide and consume services. Agents can help each other find trustworthy parties by providing referrals to those that they trust. We propose a graph-based representation of services for modeling the trustworthiness of agents. This representation captures natural relationships among service domains and provides a simple means to accommodate the accrual of trust placed in a given party. When interpreted as a lattice, it enables less important services (e.g., low-value transactions) to be used as gates to more important services (e.g., high-value transactions). We first show that, where applicable, this approach yields superior efficiency (needs fewer messages) and effectiveness (finds more providers) than a vector representation that does not capture the relationships between services. Next, we study trade-offs between various factors that affect the performance of this approach.
We view the Internet as supporting a peer-to-peer information system whose components provide ser... more We view the Internet as supporting a peer-to-peer information system whose components provide services to one another. The services could involve serving static pages, processing queries, or carrying out transactions. We model service providers and consumers as autonomous agents. Centralized indexes of the web are replaced by individual indexes kept by the agents. The agents can cooperate with one another. An agent may provide a service to another agent or give a referral that leads it in the right direction. Importantly, the agents can judge the quality of a service obtained and adaptively select their neighbors in order to improve their local performance. Our approach enables us to address two important challenges. One, in contrast with traditional systems, finding trustworthy parties is nontrivial in open systems. Through referrals, agents can help one another find trustworthy parties. Two, recent work has studied the structure of the web as it happens to have emerged mostly through links on human-generated, static pages. Whereas existing work takes an after-the-fact look at web structure, we can study the emerging structure of an adaptive P2P system as it relates to the policies of the members.
Proceedings of the second international joint conference on Autonomous agents and multiagent systems, 2003
Agents must decide with whom to interact, which is nontrivial when no central directories are ava... more Agents must decide with whom to interact, which is nontrivial when no central directories are available. A classical decentralized approach is referral systems, where agents adaptively give referrals to one another. We study the emergent properties of referral systems, especially those dealing with their quality, efficiency, and structure. Our key findings are (1) pathological graph structures can emerge due to some neighbor selection policies and (2) if these are avoided, quality and efficiency depend on referral policies. Further, authorities emerge automatically and the extent of their relative authoritativeness depends on the policies.
CP-nets have proven to be an effective representation for capturing preferences. However, their u... more CP-nets have proven to be an effective representation for capturing preferences. However, their use in automated negotiation is not straightforward because, typically, preferences in CP-nets are partially ordered and negotiating agents are required to compare any two outcomes based on a request and an offer in order to negotiate effectively. If agents know how to generate total orders from their CP-nets, they can make this comparison. This paper proposes heuristics that enable the use of CP-nets in utility-based negotiations by generating total orderings. To validate this approach, the paper compares the performance of CP-nets with our heuristics with the performance of UCP-nets that are equipped with complete preference orderings. Our results show that we can achieve comparable performance in terms of the outcome utility. More importantly, one of our proposed heuristics can achieve this performance with significantly smaller number of interactions compared to UCP-nets.
We are developing a decentralized approach to trust based on referral systems, where agents adapt... more We are developing a decentralized approach to trust based on referral systems, where agents adaptively give referrals to one another to find other trustworthy agents. Interestingly, referral systems provide us with a useful and intuitive model of how links may be generated: a referral corresponds to a customized link generated on demand by one agent for another. This gives us a basis for studying the processes underlying trust and authority, especially as they affect the structure of the evolving social network of agents. We explore key relationships between the policies and representations of the individual agents on the one hand and the aggregate structure of their social network on the other.
Proceedings of the International Conference on Autonomous Agents, 2006
Selecting the right parties to interact with is a fundamental problem in open and dynamic environ... more Selecting the right parties to interact with is a fundamental problem in open and dynamic environments. The problem is exemplified when the number of interacting parties is high and the parties' reasons for selecting others vary. We examine the problem of service selection in an e-commerce setting where consumer agents cooperate to identify service providers that would satisfy their service needs the most. Previous approaches to service selection are based on capturing and exchanging the ratings of consumers to providers. Contrary to previous, rating-based service selection, this paper advocates an objective experience-based approach for service provider selection, in which consumers record their experiences with service providers rather than the overall, subjective ratings for a provider. A consumer's experience with a service provider is represented using an ontology that can capture subtle details including the context in which the service was requested. When a service consumer decides to share her experiences with a second service consumer, the receiving consumer evaluates the experience using its own context and evaluation criteria. By sharing experiences rather than ratings, the service consumers can model service providers more accurately and thus can select the service providers for their needs more correctly.
Consumers use service selection mechanisms to decide on a service provider to interact with. Alth... more Consumers use service selection mechanisms to decide on a service provider to interact with. Although there are various service selection mechanisms, each mechanism has different strengths and weaknesses for different settings. In this paper, we propose a novel approach for consumers to learn how to choose the most useful service selection mechanism among different alternatives in dynamic environments. In this approach, consumers continuously observe outcomes of different service selection mechanisms. Using their observations and a reinforcement learning algorithm, consumers learn to choose the most useful service selection mechanism with respect to their trade-offs. Through the simulations, we show that not only the consumers choose the most useful service selection mechanism using the proposed approach, but also the performance of the proposed approach does not go below the lower-bound defined by the tradeoffs of the consumers.
Lecture Notes in Business Information Processing, 2010
Automated negotiation is important for carrying out flexible transactions. Agents that take part ... more Automated negotiation is important for carrying out flexible transactions. Agents that take part in automated negotiation need to have a concise representation of their user's preferences and should be able to reason on these preferences effectively. We develop an automated negotiation platform wherein consumer agents negotiate with producer agents about services. A consumer agent represents its user's preferences in a compact way using a CP-net, which is a structure that allows users to order their preferences based on the different value combinations of attributes. Acquiring user's preferences in a compact way is crucial since it significantly decreases the number of questions to be asked to the user by the consumer agent. We design strategies for consumer agents to reason on and negotiate effectively with the preference graph induced from a CP-net. These strategies are designed to generate deals that are acceptable by the provider and the consumer. We compare our proposed strategies in terms of how well and how quickly they can find desirable deals for the consumer.
In online and dynamic e-commerce environments, it is beneficial for parties to consider each othe... more In online and dynamic e-commerce environments, it is beneficial for parties to consider each other's preferences in carrying out transactions. This is especially important when parties are negotiating, since considering preferences will lead to faster closing of deals. However, in general may not be possible to know other participants' preferences. Thus, learning others' preferences from the bids exchanged during the negotiation becomes an important task. To achieve this, the producer agent may need to make assumptions about the consumer's preferences and even its negotiation strategy. Nevertheless, these assumptions may become inconsistent with a variety of preference representations. Therefore, it is more desired to develop a learning algorithm, which is independent from the participants' preference representations and negotiation strategies. This study presents a negotiation framework in which the producer agent learns an approximate model of the consumer's preferences regardless of the consumer's preference representation. For this purpose, we study our previously proposed inductive learning algorithm, namely Revisable Candidate Elimination Algorithm (RCEA). Our experimental results show that a producer agent can learn the consumer's preferences via RCEA when the consumer represents its preferences using constraints or CP-nets. Further, in both cases, learning speeds up the negotiation considerably.
We are developing an approach for P2P information systems, where the peers are modeled as autonom... more We are developing an approach for P2P information systems, where the peers are modeled as autonomous agents. Agents provide services or give referrals to one another to help find trustworthy services. We consider the important case of information services that can be cached. Agents request information services through highlevel queries, not by describing specific objects as in caching in traditional distributed systems. Moreover, the agents autonomously decide whom to contact for a service, whom to provide a service or referral, whether to follow a referral, and whether to cache a service. Thus the information system itself evolves as agents learn about each other and the contents of the caches of the agents change. We study the effect of caching on service location and on the information system itself. Our main results are that (1) even with a small cache, agents can locate services more easily; (2) since the agents that cache services can act like service providers, few initial service providers are enough to serve the information needs of the consumers; and (3) agents benefit from being neighbors with others who have similar interests.
Proceedings of the 6th international joint conference on Autonomous agents and multiagent systems, 2007
Most of the proposed approaches in automatic service selection assume the existence of a common o... more Most of the proposed approaches in automatic service selection assume the existence of a common ontology among communicating agents. However, this assumption becomes difficult to support in environments, where the agents' ontologies can evolve independently based on their individual experiences. In this paper, we propose an approach through which agents can cooperatively update their ontologies and teach one another concepts from their ontologies. This leads to a society of agents with different but overlapping ontologies. Our simulation results show that mutually accepted concepts emerge based on the interactions of the agents. Further, agents learn and use concepts that are created by other to express their own service needs.
We develop an approach in which we model communication protocols via commitment machines. Commitm... more We develop an approach in which we model communication protocols via commitment machines. Commitment machines supply a content to protocol states and actions in terms of the social commitments of the participants. The content can be reasoned about by the agents thereby enabling flexible execution of the given protocol. We provide reasoning rules to capture the evolution of commitments through the agents' actions. Because of its representation of content and its operational rules, a commitment machine effectively encodes a systematically enhanced version of the original protocol, which allows the original sequences of actions as well as other legal moves to accommodate exceptions and opportunities. We show how a commitment machine can be compiled into a finite state machine for efficient execution, and prove soundness and completeness of our compilation procedure. £ We would like to thank James Lester and Peter Wurman for helpful comments.
Proceedings of the 6th international joint conference on Autonomous agents and multiagent systems, 2007
In online, dynamic environments, the services requested by consumers may not be readily served by... more In online, dynamic environments, the services requested by consumers may not be readily served by the providers. This requires the service consumers and providers to negotiate their service needs and offers. Multiagent negotiation approaches typically assume that the parties agree on service content and focus on finding a consensus on service price. In contrast, this work develops an approach through which the parties can negotiate the content of a service. This calls for a negotiation approach in which the parties can understand the semantics of their requests and offers and learn each other's preferences incrementally over time. Accordingly, we propose an architecture in which both consumers and producers use a shared ontology to negotiate a service. Through repetitive interactions, the provider learns consumers' needs accurately and can make better targeted offers. To enable fast and accurate learning of preferences, we develop an extension to Version Space and compare it with existing learning techniques. We further develop a metric for measuring semantic similarity between services and compare the performance of our approach using different similarity metrics.
IEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems and Humans, 2005
Developing, maintaining, and disseminating trust in open, dynamic environments is crucial. We pro... more Developing, maintaining, and disseminating trust in open, dynamic environments is crucial. We propose self-organizing referral networks as a means for establishing trust in such environments. A referral network consists of autonomous agents that model others in terms of their trustworthiness and disseminate information on others' trustworthiness. An agent may request a service from another; a requested agent may provide the requested service or give a referral to someone else. Possibly with its user's help, each agent can judge the quality of service obtained. Importantly, the agents autonomously and adaptively decide with whom to interact and choose what referrals to issue, if any. The choices of the agents lead to the evolution of the referral network, whereby the agents move closer to those that they trust. This paper studies the guidelines for engineering self-organizing referral networks. To do so, it investigates properties of referral networks via simulation. By controlling the actions of the agents appropriately, different referral networks can be generated. This paper first shows how the exchange of referrals affects service selection. It identifies interesting network topologies and shows under which conditions these topologies emerge. Based on the link structure of the network, some agents can be identified as authorities. Finally, the paper shows how and when such authorities emerge. The observations of these simulations are then formulated into design recommendations that can be used to develop robust, self-organizing referral networks.
Uploads
Papers by PInar Yolum