A simulation testbed for analyzing trust and reputation mechanisms in unreliable online markets

https://doi.org/10.1016/j.elerap.2014.07.001Get rights and content

Highlights

  • Open-source approach for building market scenarios with varying population mixes.

  • The effect of sources of information on consumer decisions.

  • Benchmarking of various trust and reputation models.

  • Based on the FIRE modelling approach and built in Eclipse AMP.

  • Example experimental scenario to demonstrate testbed setup and functionality.

Abstract

Modern online markets are becoming extremely dynamic, indirectly dictating the need for (semi-) autonomous approaches for constant monitoring and immediate action in order to satisfy one’s needs/preferences. In such open and versatile environments, software agents may be considered as a suitable metaphor for dealing with the increasing complexity of the problem. Additionally, trust and reputation have been recognized as key issues in online markets and many researchers have, in different perspectives, surveyed the related notions, mechanisms and models. Within the context of this work we present an adaptable, multivariate agent testbed for the simulation of open online markets and the study of various factors affecting the quality of the service consumed. This testbed, which we call Euphemus, is highly parameterized and can be easily customized to suit a particular application domain. It allows for building various market scenarios and analyzing interesting properties of e-commerce environments from a trust perspective. The architecture of Euphemus is presented and a number of well-known trust and reputation models are built with Euphemus, in order to show how the testbed can be used to apply and adapt models. Extensive experimentation has been performed in order to show how models behave in unreliable online markets, results are discussed and interesting conclusions are drawn.

Graphical abstract

The consumer decision/action model in Euphemus.

  1. Download : Download full-size image

Introduction

Open distributed environments have successfully been modeled as multi-agent systems comprising agents that interact with each other under specific protocols and service-level agreements. Most of these systems take the availability of information for granted and just focus on the decision-making strategies of agents. In real-life environments, though, it is practically impossible for agents to have perfect information about the environment, properties and possible strategies or interests of others. Thus, agents have to make decisions under uncertainty.

One way to tackle uncertainty in open distributed systems is through the definition of a trust and reputation scheme. Trust and reputation (TR) models may guide an agent in deciding with whom to (prefer to) interact. In fact, trust and reputation have been recognized as key issues in autonomic, peer-to-peer and grid computing, as well as in service-oriented architectures and e-commerce applications.

In online markets, with millions of nearly-anonymous agents buying and selling a plethora of goods, self-interested selling agents may act maliciously by not delivering products with the same quality as promised. Thus, trust is a critical issue and it is important for buying agents to reason about the trustworthiness of sellers and determine with which sellers to interact. The higher the value of the products being transacted, the higher the importance of trust for the successful engagement of buyers and sellers.

Trust in online markets seems to be more important than in physical ones (Bakos and Bailey, 1997), since neither seller identity nor product characteristics can be evaluated during the transaction. For this reason, users usually request reliable reports on past performance and truthful statements of future guarantees, and are more likely to participate in web transactions and relationships if they receive strong assurances that they are engaging in a trusting relationship.

As commerce in high-value items becomes increasingly profitable on the Internet, online merchants and auctioneers face enormous challenges in overcoming the trust problem and creating attractive trading environments. In this context, trust and reputation systems provide a foundation for security, stability, and efficiency in the online environment because of their ability to stimulate quality and to sanction poor quality. Trust and reputation scores are assumed to represent and predict future quality and behaviour and thereby to provide valuable decision support for relying parties.

Current work aspires to investigate issues related to trust and reputation in open online markets. Towards this aim we have developed Euphemus, a multivariate agent-based platform for simulating online markets, where agents offer a specific service that others may consume. Setting up a set of example scenarios, we have studied the impact of various sources of information and evaluation criteria in decision making, as well as the effect of altering agents’ behavior for various trust and reputation models.

The paper is organized as follows: Section 2 discusses the state-of-the-art on the available trust and reputation models and testbeds. Section 3 discusses the basic trust and reputation concepts that Euphemus builts upon, the models that have been developed through Euphemus, the architecture and the data flow in the developed framework. Finally, Section 4 discusses the categories of the experiments conducted, while Section 5 summarizes work performed, probes on future extensions and concludes the paper.

Section snippets

Trust and reputation models

The terms of trust and reputation have been used in the literature in various ways, but there is no commonly accepted definition. Josang et al. (2007) divide trust in reliability trust and decision trust, where reliability trust focuses on dependence on the trusted party, as seen by the trusting party (Gambetta, 1990), while decision trust is the extent to which one party is willing to depend on something or somebody in a given situation with a feeling of relative security (McKnight and

Euphemus modeling, architecture and data flow

In order to study issues related to trust and reputation in online markets, we have developed a framework where agents interact taking under consideration the trustworthiness of others. In this section we discuss the main building blocks of the framework, its architecture and the flow of data.

Experiments

In this section we demonstrate the capabilities of the Euphemus framework and perform indicative benchmarking between different TR models against an example market scenario. Focus is given on the methodology to follow in order to define a hypothesis and assess the performance of the model, as well as on the way that different TR models can be implemented and run through Euphemus. In this context, we have performed three sets of experiments: a) we have developed the FIRE model (Huynh et al., 2006

Conclusions and future work

Within the context of our work we have developed a multivariate adaptive testbed, Euphemus, for the analysis of multi-objective trust and reputation mechanisms in agent-based online markets. In Euphemus agents are split in two groups, either providing or consuming a service. The Euphemus TR model draws primitive from the FIRE model, but also provides the ability to easily implement other TR models. It has been developed on AMP, in order to allow for easily reconfigurability, fast adaptation and

References (48)

  • Rosaria. Conte et al.
    (2002)
  • D. Dubois et al.

    Fuzzy Sets and Systems: Theory and Applications

    (1980)
  • R. Falcone et al.

    A fuzzy approach to belief-based trust computation

    (2005)
  • B. Friedman et al.

    Trust online

    Commun. ACM

    (2000)
  • Fullam, K.K., Klos, T.B., Muller, G., Sabater, J., Schlosser, A., Topol, Z., Barber, K.S, Rosenschein, J.S., Vercouter,...
  • D. Gambetta

    Trust: Making and Breaking Cooperative Relations, Chapt. Can We Trust Trust?

    (1990)
  • T. Grandison et al.

    A survey of trust in internet applications

    IEEE Commun. Surv. Tutorials

    (2000)
  • N. Griffiths

    A fuzzy approach to reasoning with trust, distrust and insufficient trust

  • R. Guha et al.

    Propagation of trust and distrust

  • Gyongyi, Z., Garcia-Molina, H., Pedersen, J., 2004. Combating web spam with trustrank. In: Proceedings of the 30th...
  • Hang, Chung-Wei, Yonghong Wang, Munindar P. Singh, 2008. An adaptive probabilistic trust model and its evaluation. In:...
  • Harbers M., Verbrugge R., Sierra C., Debenham J., 2007. The examination of an information-based approach to trust. In:...
  • H. Huang et al.

    Revisiting Trust and Reputation in Multi-agent Systems

    ISECS Int. Colloquium Comput. Commun. Control Manage.

    (2008)
  • T.D. Huynh et al.

    An integrated trust and reputation model for open multi-agent systems

    Auton. Agent. Multi-Agent Syst.

    (2006)
  • Cited by (15)

    • Recommendation quality, transparency, and website quality for trust-building in recommendation agents

      2016, Electronic Commerce Research and Applications
      Citation Excerpt :

      Transparency means that the recommendation agent is capable of conveying to the user – for example with the help of system-generated explanations – why certain items were recommended. Looking beyond trust-related aspects of recommendation agents, the topic of trust in e-commerce has been extensively discussed in the research literature, and trust is often considered as a key enabler for the success of e-commerce in general (Adamopoulou and Symeonidis, 2014; Gefen, 2000; Grabner-Kräuter and Kaluscha, 2003; Jarvenpaa et al., 1999; Moody et al., 2014). In the e-commerce settings considered in this paper, the online shop is often the only way in which a company communicates with its customers (Chang and Chen, 2008), and one common assumption in this stream of research is that Internet users at least partially rely on website attributes like information quality, interaction quality, or performance to judge the trustworthiness of the site.

    • Effect of security investment strategy on the business value of managed security service providers

      2019, Electronic Commerce Research and Applications
      Citation Excerpt :

      This is a universal rule in the market and forms a typical reinforcing loop because a stronger reputation results in more clients, thus strengthening the reputation. Several studies have investigated the issues of reputation under traditional commerce and e-commerce (Adamopoulou and Symeonidis, 2014; Fang et al., 2014). Research indicates that the reputation effect can mitigate the moral hazard problem (Ding et al., 2005).

    • Identifying Colluding Actors in Social Communities by Reputation Measures

      2022, Communications in Computer and Information Science
    View all citing articles on Scopus
    View full text