2003 Business Network Fellow
Robert Ghanea-Hercock, BTexact Technologies
The Santa Fe Institute is pleased to announce that Dr. Robert Ghanea-Hercock, Principal Research Scientist with BTexact Technologies, Ltd., is the recipient of the first annual Business Network Fellowship.
Dr. Ghanea-Hercock's Fellowship will cover a two-year period of time in which he will pursue collaborations with members of the SFI research community in support of his choosen research topic (see below), and spend up to four weeks in residence at the Institute. The fellowship will culminate in the submission of a detailed technical report intended for publication.
The Dynamics of Collective Reputation Development In Multi-Agent Systems
Abstract
This project will explore the dynamic formation and interactions of trust and reputation processes, within multi-agent systems (MAS). The application of software agents in e-commerce and computing security applications is increasingly dependent on the ability of such agents to form coalitions of trusted partners and reputation networks. However, this remains a poorly understood theoretical domain, and commercial systems urgently require applicable algorithms for implementing and managing reputation within MAS. This project would, therefore, involve an empirical and theoretical study of reputation formation, using a MAS simulation platform. The specific aim being to establish agent protocols and behaviors that lead to stable reputation networks.
Introduction
The use of multi-agent systems within the commercial domain is increasing and is expanding from traditional applications in e-commerce [7] to include agents in embedded systems, and network security applications [3] [4]. A common feature of these applications is the need for a distributed community of agents to establish networks of trust and reputation. Examples include agent-based network authentication systems [6], and reputation development between agent brokers in e-commerce systems [2].
The Problem
While a number of mechanisms have been utilized to develop agent reputation systems [1] [6] [10], there is still a lack of understanding in terms of how such processes will evolve over time, or the dynamic behavior of such systems. For example, the following quote from Braynov [2], illustrates the importance of inter-agent trust:
"As electronic commerce develops, its success crucially depends on establishing, maintaining, and managing trust in on-line transactions. Economic agents can fail to perform their tasks or to meet their commitments due to lack of incentives, lack of ability, or due to circumstances beyond their control. In risky environments trust enables cooperation and permits voluntary participation in mutually beneficial transactions which are otherwise costly to enforce or cannot be enforced by third parties." [2]
In this respect the problem mirrors that of trust and social capital generation within human societies and economic systems [11].
Scientific and Technical Objectives
Based on the short duration of the project, two specific topics would be studied, these are:
Research of behavioral rules that may engender stable trust relationships between software agents.
Mapping of the dynamics of coalition formation driven by reputation mechanisms in MAS.
Research Environment
The proposed research platform will be an agent simulation testbed based on the REPAST toolkit from University of Chicago. (Preliminary work has been conducted using this platform within BTexact by the author.)
Relation to SFI Program
The specified problem displays a number of properties common to complex adaptive systems. For example, the dynamics of agent interaction with variable or discrete levels of trust can generate highly complex or chaotic sequences of agent behavior [10].
The proposed project also has an overlap with the SFI Robustness program, directed by Dr. Erica Jen. In particular, it considers processes that impact the resilience and robustness of computing and communication networks. Finally, the proposed subject of study is interdisciplinary in nature and calls on research from economics, game theory, social studies, and distributed computing systems.
Summary
This project would be of scientific and commercial benefit for the following reasons:
- There is an urgent industrial requirement for practical guidance on how to develop MAS that can autonomously sustain stable reputation and trust relationships.
- The domain provides an excellent testbed for theories of stability and adaptability in complex adaptive systems.
- There is scope for significant expansion and further development in this field, across both the academic and commercial domains.
- A key objective of the proposal is to develop further collaborative interaction with SFI sponsored researchers, in order to generate a sustained research programme of trust in MAS.
References