INTELLIGENT
SOFTWARE AGENTS IN E-BUSINESS SELECTED ISSUES
Szymon Kościów[1]
And Andrzej Małachowski[2],
Wroclaw University Of Economics, Poland
Abstract
In times of rapid growth of information and knowledge sources, what we are searching for tends to be more and more concealed. Thus, finding crucial information is extremely important from the point of view of many different people, companies and institutions (particularly those, which are related to e-commerce and e-business).
With each year, Intelligent Software Agents (ISAs) gain more and more popularity and significance as the right tool to acquire, store and manage the information and knowledge needed for the proper functioning of active units in e-business.
Among them, we can distinguish brokers, managers, customers and so on. ISAs can act on behalf of any of the groups mentioned above.
Information, transaction and cooperative agents constitute the three categories of ISAs, Furthermore, one should also mention Multi-Agent Systems (MAS) comprising several single agents.
From the point of view of e-business, probably the most crucial and important are information agents and MASes based on such agents.
The aim of the following article is to present the potential use of Intelligent Software Agents in e-business as a tool supporting proper management and acquisition of information. In this paper, the authors focus mainly on a role of information agents, but meanwhile assume that cooperative and transactional ones can be useful as well. However, the issue of use of agents other than information ones is dedicated to the future studies, not included in this paper.
Keywords: information agents, Multi-Agent Systems, e-business.
1. Introduction
Contemporary societies and companies evolve inevitably towards information ones. Using electronic networks (mainly Internet) becomes more and more common and obvious, which causes fast and efficient data and information exchange among companies and single people.
One observes a constantly growing number of spheres of activity, where one can successfully implement software agents. Beginning with tools to retrieve the information, through the agents that take actions based on the knowledge about the companys strategy.
The necessity to use software agents arose when the most urgent need of companies was to improve actions like the information retrieval or commerce in Internet. In these cases, the agents work enables to quicken significantly the process of searching and filtration of a huge number of documents found in the Internet.
Of all of the types of agents, the most important and useful - from the point of view of e-business are software agents. They are much more significant than human-agents or hardware-agents (robots).
2. The essence and basic types of software agents
Due to fulfilled duties, a software agent is built mainly to accomplish tasks it is commissioned to by users, who have not enough time and knowledge to make these tasks on their own. However, together with their intelligence and autonomy, software agents must possess such abilities as reactivity, mobility, goal orientation, and so on.
To secure the intended aim, agents have to be in a constant interaction with the environment. They are to be able to acquire the information from numerous sources, and on the ground of it to take proper decisions; and to initialise required actions as a consequence of these decisions. Moreover, one expects software agents to be able to use such a communication language that will enable the cooperation with other objects to resolve more complex problems.
By definition, the intelligent software agent is a software program that can perform specific tasks for a user and possesses a degree of intelligence that permits it to perform parts of its task autonomously and to interact with its environment in a useful manner [BREN98].
We can distinguish between three categories of the software agents: information, transaction and cooperation agents [BREN98, MAES94, SYCA96]. Another group comprises Multi-Agent Systems.
2.1. Information Agents
The basic task of the information agent is to support the user with the information retrieval in networks and distributed systems. This agent has to be able - based on the users profile - to find, extract and filter on his behalf the relevant information, and to present the results of this search in a proper way and form. Furthermore, the information agent has to provide the access to the distributed information sources. To fulfil this task, it possesses the strategies of choosing the information sources and its access, these sources models and ways to merge theme. Information agents tasks include: finding answers for single queries to the information sources (or for the queries which occur frequently) or monitoring the sources of the information from the point of view of changes in their content. The example of such an agent is the agent, which monitors the stock exchange.
According to Klusch [KLUS99, KLUS01], an information agent is a goal-orientated entity and enables the access to distributed information sources. It should be able to perform one or more of the following actions: (1) information acquisition and management; (2) information presentation; (3) intelligent users support.
3. Information Agents in e-business
The general task of E-business Information Agent (EIA) is to acquire, gather, execute, modify (update, restructure) and make access to the relevant information and knowledge on behalf of active units in e-business enterprises. The main objective of EIAs is to penetrate distributed sources of information in form of text, graphics or multimedia. To obtain desired results, one often uses here methods and tools of the artificial intelligence (genetic algorithms, neural networks, fuzzy logic, etc.).
Dynamically changing content of WWW pages constitutes the main source of the information and knowledge for EIAs. Other sources include e-mails and services like Usenet, newsgroups or IRC [PADG02, IMPU03]. Curiously, EIAs can also spy on the work of other agents to take advantages of the information acquired by them. One calls these agents parasite EIAs.
Main fields of e-business where EIAs can be applicable include:
strategic management (supporting decision processes, monitoring the decisions execution) [CARD99, FERB99],
marketing (supporting marketing activities, cooperation with customers) [SADE02, WEIS01],
logistics (activities on the field of distribution, stocks, etc.),
production and sale (improvement of technologies, production and products, negotiations with purchasers/customers) [MAGM03].
The complexity of the problems that agents are meant to solve makes impossible to create one versatile agent. Thus, one uses agents from different groups specialised in one sort of a task, which forces system designers to create proper coordination mechanisms (information exchange, communication, cooperation). Fig.1 presents the levels of cooperation between different agents in the system.
Systems, which rely on numerous agents are called Multi-Agent System MAS. Undoubtedly, the nearest future will confirm the necessity for that domain-specialised, multitask and complex agent systems.
While talking about MASes one cannot forget about the special-purpose coordination agents called mediators, which task is to solve the conflicts and inaccuracies among agents [JACI02, WEIS01].
Figure.1. Cooperation of the agents in e-business.
Source: Own elaboration.
A distinct problem that appears while analysing agents systems is the safety and protection of EIAs from the destructive actions of so-called cannibal-agents.
3.1. Information Agents in e-business example applications
Recently, intelligent information agents and Multi-Agent Systems built of such agents have been gaining more and more popularity as a solution for numerous problems concerning e-business.
Out of many confirmations of the efficiency of such applications, there are information agents systems meant to solve the problem of the production scheduling. Parunak [PARU94] - as one of the firsts proposed to use Multi-Agent Systems in order to divide one sophisticated problem into several much less complex single tasks executed by individual agents.
Few years later, based on another MAS, there was designed and implemented an application for the dynamic production scheduling by Shen and Norie [SHNO98]. Here, the task of scheduling is performed by three different groups of agents - Worker Agents, Machine Agents and Tool Agents. Meanwhile, there are also Mediator Agents, who have to negotiate the final shape of the production schedule between Worker Agents and Tool Agents. Furthermore, some of the Mediators allocate tasks among the Machine Agents. And Worker Agents are to choose the resources and create the production schedule. The most crucial advantage of this system is the constantly up-to-date information about the current states, what makes the process of scheduling much easier.
There are also several other Multi-Agent Systems designed to solve problems of scheduling, management and allocation [PABC97, BOVV00].
Another problem, where information agents fit well is the e-mails management. Segal and Kephart [SEKE99, SEKE00] created two systems for managing incoming mails - MailCat and SwiftFile. Here, the agent is an intelligent assistant, which using adaptive classifier chooses three most probable folders for any incoming mail. But, the final decision of which folder is the most suitable belongs to the user. Fig. 2 presents the user interface in MailCat.
Figure.2. MailCat Agent system managing incoming mails.
Source: [SEKE99].
Information retrieval constitutes one more area where information agents can be successfully implemented as a user-supporting tool.
Information retrieval applications based on the intelligent information agents become helpful from the point of view of management, because they constitute a useful tool for acquiring information for the purpose of this management.
One of such applications is a system called Anticipator [SEVA02].
Anticipators main task is to provide users with the relevant information adequate to their dynamically changing information needs. To succeed, Anticipator creates user profiles describing their preferences and needs and how to adapt them to the changing environment.
There are two types of agents in the system: Profile Agents (PA) and Event Monitoring Agents (EMA).
System works as follows:
(1) For each new user, based on the existing template, system initialises the profile describing the information requests and event rules (how the system should react when there appear changes in the environment crucial from the point of view of the users information needs). Each profile exists in the system as long as its user is in the system.
(2) PA agents are created in the dispatch engine,
(3) When the profile is initialised, EMA agents are associated with it. Each information request and event creates one EMA agent,
(4) In the Anticipator there are two types of EMA agents. First - time-driven EMAs - are the agents that direct users queries to the data sources in certain time intervals. Second group is the data-driven EMAs. Their task is to constantly check whether the changes in the data sources may result in a modification of users requests. Data-driven EMAs pass the information about these changes to PA agents, so they can modify users profile. When the relevant information is obtained, it can be used by the user in management, logistics and other branches of companys activities.
Fig. 3 presents the architecture of the Anticipator system.
Information about adding or removing a user is stored in the Registry. The Dispatch Engine creates/removes and manages PA Agents. Next, EMA agents are associated with every profile. Data-driven EMAs pass the information about the changes in the data sources to PA agents. And time-driven EMAs direct users queries to the data sources in certain time intervals. When the relevant information is obtained, it is sent in packets to users. This information may be used in management, logistics and other branches of companys activities.
Source: [SEVA02].
4. Conclusions
The aim of this paper is to present the idea of the role of intelligent information agents in e-business. Undoubtedly, next few years will prove the usefulness of agent systems. There will be more and more applications, which aim at optimising and improving e-business. This paper shows some examples of already existing systems that organise e-mails, manage the production scheduling or retrieve information.
The growing number of areas, where one can use agents to solve sophisticated problems will definitely increase with each year.. The intention of the authors is further study of the problems of using information agents in e-business.
References:
[BOVV00] Bochmann O., Valckenaers P., Van Brussel H. Negotiation-Based Manufacturing Control in Holonic Manufacturing Systems, Proc. of the ASI2000 Conference, 206-213, 2000
[BREN98] Brenner W., Zarnekow R., Wittig H.
Intelligent Software Agents, Springer-Verlag, 1998
[CARD99] Cardoso H. L.
A multi-agent system for electronic commerce including adaptive strategic behaviours, EPIA99 Portuguese Conference on Artificial Intelligence, September 1999
[FERB99] Ferber J.
Multi-Agent Systems: An Introduction to Distributed Artificial Intelligence, Addison-Wesley, London, 1999
[GUMM98] Guttman R. H., Moukas A. G., Maes P.
Agent-Mediated Electronic Commerce: A Survey, Knowledge Engineering Review, June 1998
[IMPU03] Impulse Agents, http://www.media.mit.edu/groups/agents/projects/impulse, (August 2003),
[JACI02] Jain L. C., Chen Z., Ichalkaranje N. (eds.)
Intelligent Agents and Their Applications, Springer, 2002
[KLUS99] Klusch M.
Intelligent Information Agents, Springer, 1999
[KLUS01] Klusch M.,
Information Agent Technology for the Internet: A Survey, Data and Knowledge Engineering, 36, 337-372, 2001
[MAES94] Maes P.
Agents that Reduce Work and Information Overload, Communication of the ACM, 37, 31-40, 1994
[MAGM03] Maes P., Guttman R. H., Moukas A. G.
Agents That Buy and Sell, http://www.delivery.acm.org/10.1145/300000/295716/p81-maes.pdf, (August 2003)
[PABC97] Parunak H., Baker A., Clark S.
AARIA Agent Architecture: An example of Requirements-Driven Agent-Based System Design, Proc. of the International Conference on Autonomous Agents, 482-483, 1997
[PADG02] Padget J. (et al.)
Agent-Mediated Electronic Commerce IV. Designing Mechanisms and Systems, AAMAS Workshop, Bologna, July 2002
[PARU94] Parunak H.
Applications of Distributed Artificial Intelligence in Industry, w: OHare G. M. P., Jennings N. R. (eds.): Foundations of Distributed Artificial Intelligence, Wiley Inter-Science, 1994
[SADE02] Sadeh N. (et al.)
Open-Agent Environment for Context-Aware M-Commerce, Agentcities Workshop, Bologna, July 2002
[SEKE99] Segal R., Kephart J.
MailCat: An Intelligent Assistant for Organizing E-Mail, Proc. of the Third Int'l Conf. on Autonomous Agents, 1999
[SEKE00] Segal R., Kephart J.
Swiftfile: An intelligent assistant for organizing e-mail, AAAI 2000 Spring Symposium on Adaptive User Interfaces, 2000
[SEVA02] Sevay H., Tsatsoulis C.
Agent-based intelligent information dissemination in dynamically changing environment, in: Jain L. C., Chen Z., Ichalkaranje N.(eds.), Intelligent Agents and Their Applications, Physica-Verlag, 2002
[SHNO98] Shen W., Norrie N.
An Agent-Based Approach for Dynamic Manufacturing Scheduling, Working Notes of the Agent-based Manufacturing Workshop, 117-128, 1998
[SYCA96] Sycara K., Decker K., Pannu A., Williamson M., Zeng D.
Distributed Intelligent Agents, IEEE Expert, 11(6), 36-46, 1996
[WEIS01] Weiss M.
Patterns for e-Commerce Agent Architectures: Using Agents as Delegates, Ottawa, Carleton Univ., 2001
