Joana Urbano

Computing Confidence Values: Does Trust Dynamics Matter?

Computational Trust and Reputation (CTR) systems are platforms capable of collecting trust information about candidate partners and of computing confidence scores for each one of these partners. These systems start to be viewed as vital elements in environments of electronic institutions, as they support fundamental decision making processes, such as the selection of business partners and the automatic and adaptive creation of contractual terms and associated enforcement methodologies. In this article, we propose a model for the aggregation of trust evidences that computes confidence scores taking into account dynamic properties of trust. We compare our model with a traditional statistical model that uses weighted means to compute trust, and show experimental results that show that in certain scenarios the consideration of the trust dynamics allows for a better estimation of confidence scores.

ANTE: Agreement Negotiation in Normative and Trust-Enabled Environments

The ANTE framework encompasses results of research efforts on three main agreement technology concepts, namely negotiation, normative environments and computational trust. ANTE has been conceived as a general framework with a wide range of applications in mind. This chapter provides an overview of the main guidelines of this project, and explores two application domains for this framework: automated B2B electronic contracting, and disruption management in the context of an airline company operational control.

Trustworthiness Tendency Incremental Extraction Using Information Gain

Computational trust systems are getting popular in several domains such as social networks, grid computing and business-to-business systems. However, the estimation of the trustworthiness of agents is not trivial in scenarios where the existing trust evidences are scarce. We propose an online, situation-aware trust model that uses the information gain metric to dynamically extract tendencies of failure of target agents, improving the process of selection of partners in a relevant way. Experimental results presented in this paper show that our proposal outperforms other trust approaches in contextual scenarios.

A Socio-cognitive Perspective of Trust

Trust and reputation are two distinct social constructs of high complexity that have been studied for decades in different areas of knowledge. In order to allow for efficient models of computational trust and reputation, one must first understand the nature and dynamics of each one of these constructs. In this chapter, we focus on the social and cognitive aspects of the trust concept, and overview its fundamental characteristics, such as its determinants, nature, and dynamics. Then, we present two distinct hypothesis one can state for the interplay between trust and reputation: either reputation is an antecedent of trust, or both are considered as two distinct contributions to the ultimate decision making process. If they are seen as isolated components, trust is no longer directly influenced by reputation. Finally, we briefly refer to current existing computational trust models, including those that integrate the management of computational reputation.

Trust and Normative Control in Multi-agent Systems: An Empirical Study

Despite relevant insights from socio-economics, little research in multiagent systems has addressed the interconnections between trust and normative notions such as contracts and sanctions. Focusing our attention on scenarios of betrayal, in this paper we combine the use of trust and sanctions in a negotiation process. We describe a scenario of dyadic relationships between truster agents, which make use of trust and/or sanctions, and trustees characterized by their ability and integrity, which may influence their attitude toward betrayal. Both agent behavior models are inspired in socio-economics literature. Through simulation, we show the virtues and shortcomings of exploiting trust, sanctions and a combination of both.

Trust estimation using contextual fitness

Trust estimation is an essential process in several multi-agent systems domains. Although it is generally accepted that trust is situational, the majority of the Computational Trust and Reputation (CTR) systems existing today are not situation-aware. In this paper, we address the inclusion of the context in the trust management process. We first refer the benefits of considering context and make an overview of recently proposed situational-aware trust models. Then, we propose Contextual Fitness, a CTR component that brings context into the loop of trust management. We empirically show that this component optimizes the estimation of trustworthiness values in context-specific scenarios. Finally, we compare Contextual Fitness with another situation-aware trust approach proposed in the literature.

Trust Evaluation for Reliable Electronic Transactions between Business Partners

In this era of digital economy, commercial relationships between business partners are increasing in flexibility, with new business binds being created whenever a business opportunity arises. Moreover, the instability in demand increases the need for enterprises to procure new partners as well as the associated risk of dealing with partners that may be unknown beforehand. Therefore, enterprises need mechanisms that allow to evaluate the confidence they have on their current and potential new, unknown, partners, and to monitor this confidence in a continuous and automatic way. This paper presents our computational trust model, which was inspired in the concept of the hysteresis of trust and betrayal and in the asymmetry principle of human psychology. Our model allows to estimate the trustworthiness of agents using different features of the dynamics of trust. Additionally, we present a study on the effect of preselecting partners based on their trustworthiness in automated negotiation processes. The study was conducted experimentally using our agent-based Electronic Institution framework for e-Contracting, which includes a normative environment and an automatic negotiation service, as well as the mentioned computational trust service. The results obtained show that, in identified conditions, business clients benefit from preselecting partners based on trust prior to the negotiation phase.

A situation-aware computational trust model for selecting partners

Trust estimation is a fundamental process in several multiagent systems domains, from social networks to electronic business scenarios. However, the majority of current computational trust systems is still too simplistic and is not situation-aware, jeopardizing the accuracy of the predicted trustworthiness values of agents. In this paper, we address the inclusion of context in the trust management process. We first overview recently proposed situation-aware trust models, all based on the predefinition of similarity measures between situations. Then, we present our computational trust model, and we focus on Contextual Fitness, a component of the model that adds a contextual dimensional to existing trust aggregation engines. This is a dynamic and incremental technique that extracts tendencies of behavior from the agents in evaluation and that does not imply the predefinition of similarity measures between contexts. Finally, we evaluate our trust model and compare it with other trust approaches in an agent-based, open market trading simulation scenario. The results obtained show that our dynamic and incremental technique outperforms the other approaches in open and dynamic environments. By analyzing examples derived from the experiments, we show why our technique get better results than situation-aware trust models that are based on predefined similarity measures.

Trust-Based Selection of Partners

The community of multi-agent systems has been studying ways to improve the selection of partner agents for joint action. One of such approaches consists in estimating the trustworthiness of potential partners in order to decrease the risk inherent to interacting with unknown agents. In this paper, we study the effect of using trust in the process of selecting partners in electronic business. We empirically evaluate and compare different trust-based selection methods, which either use trust in a preselection phase previous to the negotiation, in the negotiation process, or in both of these stages. We here briefly introduce a computational model of trust that uses a simple machine learning mechanism to dynamically derive the expected tendencies of behavior of potential candidate partner agents. The results obtained in our comparison study allow us to point to the best trust-based selecting methods to use in specific situations.

The impact of benevolence in computational trust

Trust is a construct of paramount importance in society. Accordingly, computational trust is evolving fast in order to allow trust in artificial societies. Despite the advances in this research field, most computational trust approaches evaluate trust by estimating the trustworthiness of the agents under evaluation (the trustees), without however distinguishing between the different dimensions of trustworthiness, such as ability and benevolence. In this paper, we propose different techniques to extract the ability of the trustee in the task at hand and to infer the benevolence of the trustee toward the truster when the trust judgment is made. Moreover, we propose to dynamically change the relative importance and impact of both ability and benevolence on the perceived trustworthiness of the trustee, taking into consideration the development of the relationship between the truster and the trustee and the disposition of the truster in the specific situation. Finally, we set an experimental scenario to evaluate our approach. The results obtained from these experiments show that the proposed techniques significantly improve the reliability of the estimation of the trustworthiness of agents.