This work is concerned with the conduct of Multiple Criteria Decision Making (MCDM) by intelligent software agents trading digital commodities in Application Layer Networks (ALN) such as grids or clouds.
These agents consider trustworthiness in their course of negotiation and select offers with respect to product price and seller reputation. To automate the selection process, we seek an appropriate MCDM method that provides clear advice for an agent prior to negotiating. We compare eleven well-known MCDM methods and choose the TOPSIS approach of Hwang and Yoon since it produces
comprehensible and plausible results with a justifiable amount of effort. We modify the method and present a draft named xTOPSIS that promises intertemporal performance analysis for further automatation. The resulting tool is finally tested and
evaluated in the context of a scenario similar to the eRep - Social Knowledge for e-Governance project.
Table of Contents
1 Introduction
1.1 Starting Position: Trust in eCommerce
1.2 Objectives of this Study
1.3 Conduct of this Study
2 Interactions in Application Layer Networks
2.1 Depicting the Environment
2.2 Coordination in Application Layer Networks
2.2.1 Application Layer Networks
2.2.2 Catallactic Information Systems
2.3 Software Agents in Multi Agent Systems
2.3.1 Software Agents
2.3.2 Multi Agent Systems
2.3.3 Disseminating and Gathering Information
2.4 The Object of Interaction: Trading Goods
2.4.1 Homogeneous and Heterogeneous Goods and Services
2.4.2 Price Formation Mechanisms
2.5 Differentiation through Reputation
2.5.1 Reputation and Image
2.5.2 Reputation Systems
2.6 Decision Making
2.6.1 Theory of Decision
2.6.2 The Classical Model for Decision Making
2.6.3 Multiple Criteria Decision Making
2.6.4 Preference Modeling through Utility and Values
3 Multiple Criteria Decision Making
3.1 Classification of MCDM Methods
3.2 On Data and Weights
3.2.1 Scales of Data
3.2.2 Normalization Techniques for Equalizing Diverse Scales
3.2.3 Weights as Means for Relative Importance of Criteria
3.3 Multiple Attribute Decision Making
3.3.1 A Taxonomy of MADM Methods
3.3.2 Deciding without Preference Information
3.3.3 Satisficing (Conjunctive and Disjunctive Approaches)
3.3.4 Sequential Elimination
3.3.5 Value Function Methods
3.4 Multiple Objective Decision Making
3.4.1 Overview of MODM Methods
3.4.2 Goal Programming
3.5 Decision Aids
3.5.1 Outranking Relations
3.5.2 The ELECTRE Approach
4 Application of the Extended TOPSIS to the Scenario
4.1 Structure
4.2 A Synthesis of ALN and MCDM
4.2.1 Summary of Environment Characteristics
4.2.2 Comparison of MCDM Methods
4.2.3 Conclusion for Method Application
4.3 Scenario Specifications
4.3.1 Environment and Actors
4.3.2 Interaction between Actors
4.3.3 Offer Attributes
4.3.4 Principal’s Preference Information
4.4 The Extended TOPSIS
4.4.1 Description of the Technique
4.4.2 Application of the xTOPSIS: A Numerical Example
4.5 Findings from the Scenario Application
4.5.1 Intertemporal Comparison of Reached Agreements
4.5.2 Seller Evaluation
4.5.3 Summary
5 Conclusion
5.1 Results
5.2 Suggestions for Research and some Critical Annotations
Objectives and Research Themes
This work aims to determine an efficient procedure for intelligent software agents in Application Layer Networks (ALNs) to make purchasing decisions. Faced with anonymous global traders, these agents must process multiple criteria—specifically product price, image, and social reputation—to provide clear advice to their human principals before engaging in negotiations.
- Analysis of decision-making frameworks for autonomous agents in competitive environments.
- Evaluation of Multiple Criteria Decision Making (MCDM) methods in the context of service trading.
- Implementation and modification of the TOPSIS approach into "xTOPSIS" for intertemporal performance analysis.
- Application of the proposed model to a simulation scenario based on the eRep project.
- Examination of constraints regarding human-agent trust, privacy, and automated outcome evaluation.
Excerpt from the Book
2.5.1.1 The Function of Reputation
In modern economies, companies build reputation to differentiate themselves from competitors and gain competitive advantage [Fomb96, 80]. A good reputation can stimulate product sales, increase the chance of hiring the best employees and attract potential investors [FoSh90, 233–234]. As an intangible asset, reputation is of paramount importance for service providers and can be shaped through various practices such as conducting pro bono activities or by company advertising campaigns [Fomb96, 112–136].
The peculiar importance for service providers can be explained as follows: In comparison to commodities, services are, among other things, characterized by a high degree of uncertainty. As a result of asymmetrical information between service providers and customers, potential buyers face high risks of being exploited after signing a contract (moral hazard) [MeBr06, 97], [Wora96, 62]. Taking those risks into account, consumers’ willingness-to-pay decreases and eventually leads to the destruction of markets for high-quality products [Aker70, 490–491].
Reputation is an effective panacea which indicates reliability and mitigates moral hazard [Wora98, 47]. Building a good record is expensive and time-consuming; and since reputation is sensitive to dishonesty, deceitfulness and fraudulence, a good name promotes self-commitment by encouraging the owner to continue fair business practices [MeBr06, 98]. In consequence, customers believe past behavior is a good predictor for future conduct [Roth01, 59]. This corresponds with Axelrod’s “shadow of the future”, a phenomenon describing why future interactions are constrained by behavior in the past [Axel88, 11].
Summary of Chapters
1 Introduction: Discusses the rise of the Internet economy, the necessity of trust in electronic commerce, and establishes the objectives and research methodology of the study.
2 Interactions in Application Layer Networks: Describes the technical environment of ALNs, the role of software agents, trading goods, reputation systems, and the theory behind human and agent decision-making.
3 Multiple Criteria Decision Making: Provides a comprehensive overview of MCDM classification, data normalization, weighting methods, and various decision-making models like MADM, MODM, and outranking relations.
4 Application of the Extended TOPSIS to the Scenario: Applies the derived prerequisites to a specific simulation scenario, introduces the modified xTOPSIS method, and demonstrates its effectiveness through numerical examples and comparisons.
5 Conclusion: Summarizes the study’s results, discusses the applicability of xTOPSIS to service procurement, and offers critical annotations regarding future research, human trust, and agent autonomy.
Keywords
Multiple Criteria Decision Making, MCDM, Application Layer Networks, ALN, Software Agents, Reputation Systems, TOPSIS, xTOPSIS, Decision Support, Electronic Commerce, eRep, Trust, Negotiation, Auction Theory, Multi Agent Systems
Frequently Asked Questions
What is the fundamental focus of this research?
The thesis investigates how intelligent software agents, acting as Digital Business Agents (DBAs) in Application Layer Networks, can make rational purchasing decisions while dealing with uncertainty in online marketplaces.
What are the primary fields of study explored?
The work integrates research from Multi-Agent Systems (MAS), electronic market design, reputation theory, and Multiple Criteria Decision Making (MCDM) to provide a holistic framework for automated negotiation.
What is the central research objective?
The primary goal is to identify and customize a robust MCDM method that processes variables like price, image, and social reputation to provide clear decision-making guidance to autonomous trading agents.
Which methodology is adopted for the agent's decision-making process?
The author evaluates eleven standard MCDM approaches and concludes that the TOPSIS method is most suitable, subsequently extending it into a new, performance-tracking method called xTOPSIS.
What aspects of the trade are covered in the main section?
The main part details the environmental determinants of Application Layer Networks, establishes how information is disseminated, analyzes price formation through English auctions, and implements specific reputation models like ReGreT.
Which terms best characterize this work?
The work is defined by the intersection of computational economics and information systems, emphasizing scalability, automated reasoning, and the practical application of game-theoretic concepts.
How does the xTOPSIS approach improve upon traditional TOPSIS?
Unlike standard TOPSIS, which resets calculations for every new instance, xTOPSIS maintains dynamic ideal solution vectors over time, enabling the buyer agent to track performance and adapt to market changes through a "packed memory" database.
What is the significance of the "intertemporal comparison" mentioned in the findings?
It allows the agent to treat past transactions as reference points, making it possible to identify which hubs provide reliable service and which sellers consistently offer the best quality, thereby enhancing future decision accuracy.
- Citar trabajo
- Frank Schneider (Autor), 2008, Multiple Criteria Decision Making in Application Layer Networks, Múnich, GRIN Verlag, https://www.grin.com/document/162795
