Trust and reputation are central to effective interactions in open multi-agent systems in which agents, that are owned by a variety of stakeholders, can enter and leave the system at any time. This openness means existing trust and reputation models cannot readily be used. To this end, we present FIRE, a trust and reputation model that integrates a number of information sources to produce a comprehensive assessment of an agent's likely performance. Specifically, FIRE incorporates interaction trust, role-based trust, witness reputation, and certified reputation to provide a trust metric in most circumstances. FIRE is empirically benchmarked and is shown to help agents effectively select appropriate interaction partners.

Abstract. In many dynamic open systems, agents have to interact with one another to achieve their goals. Here, agents may be self-interested and when trusted to perform an action for another, may betray that trust by not performing the action as required. In addition, due to the size of such systems, agents will often interact with other agents with which they have little or no past experience. There is therefore a need to develop a model of trust and reputation that will ensure good interactions among software agents in large scale open systems. Against this background, we have developed TRAVOS (Trust and Reputation model for Agent-based Virtual OrganisationS) which models an agent’s trust in an interaction partner. Specifically, trust is calculated using probability theory taking account of past interactions between agents, and when there is a lack of personal experience between agents, the model draws upon reputation information gathered from third parties. In this latter case, we pay particular attention to handling the possibility that reputation information may be inaccurate. 1

Trust is an integral component in many kinds of human interaction, allowing people to act under uncertainty and with the risk of negative consequences. For example, exchanging money for a service, giving access to your property, and choosing between conflicting sources of information all may utilize some form of trust. In computer science, trust is a widelyused term whose definition differs among researchers and application areas. Trust is an essential component of the vision for the Semantic Web, where both new problems and new applications of trust are being studied. This paper gives an overview of existing trust research in computer science and the Semantic Web.

This research aims to develop a model of trust and reputation that will ensure good interactions amongst software agents in large scale open systems. The following are key drivers for our model: (1) agents may be self-interested and may provide false accounts of experiences with other agents if it is beneficial for them to do so; (2) agents will need to interact with other agents with which they have little or no past experience. Against this background, we have developed TRAVOS (Trust and Reputation model for Agentbased Virtual OrganisationS) which models an agent’s trust in an interaction partner. Specifically, trust is calculated using probability theory taking account of past interactions between agents. When there is a lack of personal experience between agents, the model draws upon reputation information gathered from third parties. In this latter case, we pay particular attention to handling the possibility that reputation information may be inaccurate.

Abstract. This research aims to develop a model of trust and reputation that will ensure good interactions amongst software agents in large scale open systems in particular. The following are key drivers for our model: (1) agents may be self-interested and may provide false accounts of experiences with other agents if it is beneficial for them to do so; (2) agents will need to interact with other agents with which they have no past experience. Against this background, we have developed TRAVOS (Trust and Reputation model for Agent-based Virtual OrganisationS) which models an agent’s trust in an interaction partner. Specifically, trust is calculated using probability theory taking account of past interactions between agents. When there is a lack of personal experience between agents, the model draws upon reputation information gathered from third parties. In this latter case, we pay particular attention to handling the possibility that reputation information may be inaccurate. 1

The majority of existing trust models is based on three underlying assumptions: (i) proven identity of agents, (ii) repetitive interactions and (iii) similar trusting situations. In our work, we address these assumptions by introduction of simple classification techniques in our mechanism that extends existing trust models, rather than by introduction of a new model. The proposed approach formalizes the situation (context) and/or trusted agent identity in a multi-dimensional Identity-Context space, and attaches the trustworthiness evaluations to individual elements from this metric space, rather than to fixed identity tags (e.g. AIDs, addresses). Trustworthiness of the individual elements of the Identity-Context space can be evaluated using any trust model that supports weighted aggregations and updates, allowing the integration of the mechanism with most existing work. Trust models with the proposed extension are appropriate for deployment in dynamic, ad-hoc and mobile environments, where the agent platform can not guarantee the identity of the agents and where the cryptography-based identity management techniques may be impractical due to the unreliable and costly communication.

In this paper we develop a novel probabilistic model of computational trust that explicitly deals with correlated multi-dimensional contracts. Our starting point is to consider an agent attempting to estimate the utility of a contract, and we show that this leads to a model of computational trust whereby an agent must determine a vector of estimates that represent the probability that any dimension of the contract will be successfully fulfilled, and a covariance matrix that describes the uncertainty and correlations in these probabilities. We present a formalism based on the Dirichlet distribution that allows an agent to calculate these probabilities and correlations from their direct experience of contract outcomes, and we show that this leads to superior estimates compared to an alternative approach using multiple independent beta distributions. We then show how agents may use the sufficient statistics of this Dirichlet distribution to communicate and fuse reputation within a decentralised reputation system. Finally, we present a novel solution to the problem of rumour propagation within such systems. This solution uses the notion of private and shared information, and provides estimates consistent with a centralised reputation system, whilst maintaining the anonymity of the agents, and avoiding bias and overconfidence.

Negotiation is a fundamental interaction mechanism in multi-agent systems because it allows self-interested agents to come to mutually beneficial agreements and partition resources efficiently and effectively. Now, in many situations, the agents need to negotiate with one another many times and so developing strategies that are effective over repeated interactions is an important challenge. Against this background, a growing body of work has examined the use of Persuasive Negotiation (PN), which involves negotiating using rhetorical arguments (such as threats, rewards, or appeals), in trying to convince an opponent to accept a given offer. Such mechanisms are especially suited to repeated encounters because they allow agents to influence the outcomes of future negotiations, while negotiating a deal in the present one, with the aim of producing results that are beneficial to both parties. To this end, in this paper, we develop a comprehensive PN mechanism for repeated interactions that makes use of rewards that can be asked for or given to. Our mechanism consists of two parts. First, a novel protocol that structures the interaction by capturing the commitments that agents incur when using rewards. Second, a new reward generation algorithm that

Trust is a crucial basis for interactions among parties in large, open systems. Yet, the scale and dynamism of such systems make it infeasible for each party to have a direct basis for trusting another party. For this reason, the participants in an open system must share information about trust. However, they should not automatically trust such shared information. This paper studies the problem of propagating trust in multiagent systems. It describes a new algebraic approach, shows some theoretical properties of it, and empirically evaluates it on two social network datasets. This evaluation incorporates a new methodology that involves dealing with opinions in an evidential setting. 1

Abstract. We present a trust-based mechanism for the acquisition of information from possibly unreliable sources. Our mechanism addresses the case where the acquired information cannot be verified. The idea is to intersperse questions (“challenges”) for which the correct answers are known. By evaluating the answers to these challenges, probabilistic conclusions about the correctness of the unverifiable information can be drawn. Less challenges need to be used if an information provider has shown to be trustworthy. This work focuses on three major issues of such a mechanism. First, how to estimate the correctness of the unverifiable information. Second, how to determine an optimal number of challenges. And finally, how to establish trust and use it to reduce the number of challenges. Our approach can resist collusion and shows great promise for various application areas such as distributed computing or peer-to-peer networks.