It is a short version of the original paper in Russian - http://rvles.ieie.nsc.ru/~parinov/papers/repec-robots.htm

December 1999

Thomas Krichel, Department of Economics, University of Surrey
Guildford GU2 5XH, United Kingdom
T.Krichel@surrey.ac.uk, phone: +44-(0)1483-876958, fax +44-(0)1483-303775

David Levin, Russian Research Institute of Artificial Intelligence
http://www.rriai.org.ru/

Sergei Parinov, Institute of Economics and Industrial Engineering
pr. Lavrentieva 17, Novosibirsk, 630090, Russia
parinov@ieie.nsc.ru, phone: +7-(3832) 303640, fax +7-(3832) 302580

A tendency to decentralized creation of information resources on the Internet changes the usual context of the work of traditional libraries and, particularly, the archives of research publications. Present online means (for example, RePEc database and related online services) create a meta databases of the electronic research publications and other information resources dispersed over big number of sites. Such software system can make everyday contents control and collect updates from a big number of resource sites. The positive aspect of this situation is an opportunity of prompt access to all recent achievements and new resources of a specified research field. The negative aspect is an increase of information flows’ intensity. As a consequence of the negative aspect, the users face the problem of information overloading. The paper considers an opportunity of how to solve this problem with the help of designed online representatives (software agents) of a researcher. These agents will undertake the partial work on filtration input information flows and interacting with the similar agents of other users. The significant element of such community’s operation is a design of artificial "living" environment of these agents. A structure of this environment is also discussed in the paper.

The establishment of common information space for online-active and geographically dispersed groups of people who are connected by some joint interests, should take into account certain tendencies in the present development of online resources in the Internet:

1)decentralization of information resources which shows itself in the creation of the own online information resources by private persons and organizations instead of using “foreign” public resources;

2)personalization of users services which involves the development of your own means for searching and filtration information for your own private data bases.

In general case, these present peculiarities of Internet resources development generate specific problems: 1)set of data in which economic (for example) community may be interested, is distributed among a big number of cites and has quite different formats and completeness of presentation; 2)administrators of some web cites often create highly effective and convenient navigation and filtration means, but these services can be used, as a rule, only for the limited local databases stored on a single server.

International project RePEc [T. Krichel and others] focuses on the creation of common information space for the community of researchers in economic fields. Thanks to the realization of this project, researchers in economics have now at their disposal the unique online information resource, which includes:

1)data base of meta information, daily updated through the net of connected decentralized online sources of information on economic disciplines (electronic publications, announcements of published issues, researchers’ individual information, etc., more than 100 cites totally);

2)set of users’ interfaces to this common database, which are created independently on the sites in different countries (Canada, Great Britain, USA, Russia and others).

RePEc database has significant total amount of documents (more than 55 thousand descriptions of electronic publications by August 1999) as well as constantly increasing flow of input documents (new sections appear in the database every month). Traditional means for facilitating the users’ work with this data base, which have been developed on several RePEc mirror cites, partially solve the problem of users’ information orientation in the increasing flow of data on research results and other aspects of activity of international economic researchers community.

The estimation of coming prospect shows that the widening of research activities, which can be realized online, can soon produce information overloading for the database users. Such situation requires new ideas on the development of a set of tools to support a proper functioning of common information environment of economics scientists' international community that currently form about the RePEc database. This case requires the development of either: a)new methodological approaches (researcher information behavior models) adequate to the given conditions, and b)suitable technical maintenance means (online services) which can help in solving the problem of information overloading.

The problem of information overloading in the environment with intense exchange and information flows exists not only for online community of researchers in economics. One of the consequences of information resources decentralization’s tendency is a gradual growth of online information activity of members of all kinds of online community. Under these conditions, any online service meeting the demands for information exchange among a great number of users, will inevitably goes through the “point of growth” where its attractiveness falls since the users begin run into difficulties with information overloading.

As applied to the community of researchers in economics this fact is manifested in the increase of their online activity. This activity results both in the growth of online resources opened by researchers and in the increase of intensity of information exchanges between them. These facts should lead to the situation when the intensity of input information flow for a certain researcher will exceed his abilities to react adequately on the content of this flow. Without using additional technical means, which can make a part of technical work, researcher should restrict the number of the partners with whom he can maintain contacts and/or reduce the scope of his interests.

The task of creating additional technical tools mean the development of concept and software which enable researchers to tune his/her active information robot (software agent) on his interests profile and orient this robot on performing some online activities. Online activity of researcher's robot includes the information filtering, exchange and other interactions with similar robots belonging to other researchers. Within this context two main methodological sub problems can be selected: 1)development of the conception of information robots’ "living" (“vital activity”) environment which allows them to imitate quite effectively the interactions of real users and perform their functions; 2)development of a model of a single information robot which has simple rules for tuning, openness of standards and extendable structure for its further decentralized development by users community. To simplify the term “information robot” we will use thereafter a term “i-robot”.

The problem of designing software agents like the “i-robot” is not new and it has been the objective of quite a lot of both theoretical and applied works. A survey of some works from this area see, for example, on a Leigh Tesfatsion's cite (see [Leigh Tesfatsion] and some other links from the Reference section, in this short version of the paper the Survey section is omitted). However the topic of living environment of i-robots has not been developed essentially. The next sections include the approach of how to establish environment enabling i-robots of researchers' community efficiently interact with each other and containing the means for maintaining self-organization and community development as well. The paper covers the description of i-robot which structure fits the principles of organization of their living environment. The conclusion states the most important tasks for practical realization of presented concept as applied to information flows and information resources of the RePEc databases.

I-robots inevitably should imitate information interactions between people, as robots are their actual online representatives (agents) in virtual space of computer nets. Resulted from this fact, our description of the robots’ living environment is based on the conception of information interaction between people, which was developed first for social and economic systems (see in details about this concept [S. Parinov^A]). We admit such approach, as online community of researchers in economics, that is the object of our study, is a typical social system.

To our opinion, by now there is some disproportion between researchers’ focusing on the design of i-robots and design of environment structure where these robots have to act. Shift on robots design task may be explained by the fact that robots interaction is a coming problem. Living environment, which is a place where i-robots interact, is also an important element of general system where this artificial community is functioning. The specific features of the organization of this system determine, to a large extent, what designed functions of i-robots may be realized and will be successful. The effectiveness of organization of living environment restricts the maximum possible level of effectiveness of robots themselves. Thus most advanced i-robots cannot realize their capacities in case the environment where they will interact will not give them necessary conditions.

According to the research of people’s interaction in social systems (see [S. Parinov^B]), communities always have direct or indirect system of rules for behavior. These rules, which, to the researchers’ opinion, are given as if by nature itself, can be divided into three main categories: 1) rules determining the membership in a group; 2) rules regulating specific behavior of isolated individual inside a group; 3) rules regulating interactions between individuals inside a group. We think that the developers of artificial communities of i-robots should take up the function of the “nature” and build in the living environment the rules of similar type. This work considers possible principles of organization and structure of living environment, which refer mainly to the realization of two last categories, mentioned above.

When analyzing and designing social systems, a procedure where main active elements are some “agents” with their own behavior model finds an increasing use. Here is how, for example, the term “active agent” is defined in the instruction for program complex TAO, designed for the creation of such models (see [Technology of Active Objects]): "What is it an "active agent"? This is an object that changes its purpose (behavior) by itself on the basis of analysis of the meanings of other objects “seen” by it; its previous state and events came from external surrounding. Inside it (active agent) not only rules for its own changing are located, but the control over inducing these rules as well.

Basing on the procedure of active agents and taking into account the principles of information interactions conception in social systems, living environment can be described as joint functioning of the following types of active agents (for the details see in [S. Parinov^A]):

а) researchers should have some common ideas which determine their interest towards each other and the necessity of interaction between them. Let us call this factor as the existence of sub system of “technological” or “work” links between researchers. This sub system, from one side, determines the current state of each specific researcher in the topology of links net, from the other side, enables the researchers to evaluate desired (favorable) position in this system of links, which is the final aim of researchers’ activity.

b) "Technological" activity should be regulated by the system of conventional (for this community) behavior rules. This factor may be called as the existence of sub system “behavior rules” (in social sciences this is called “institutional structure”). This sub system determines the framework of researchers community, general parameters of their own activity and interactions between each other. From the general reasoning, it can be assumed that there may be several such sub systems in the actual researchers community. An isolated researcher can belong to several such sub systems simultaneously. These sub systems help create “groups of same interest”, which can have different membership rules, different rules for regulating members’ interaction, different motivations of members. Participation in such group requires from a researcher to perform its rules and standards.

c) In order to make above mentioned factors work, there should be a mechanism for participants’ interacting. This is performed by “collective model of environment”, which is designed with the help of reflection of information about possibilities and intentions of each participant into common information space. Collective model of environment allows to organize information exchange on the “all with all” principle, accumulate participants’ proposals on the change of their current “technological” activity, imitate on information level different possible variants of how technological communications between participants may be changed in order to choose the most attractive variant for the whole community under current conditions.

d) The last type of active agent – i-robots themselves – is the direct net representatives of community members. They act within the behavior rules for the whole community and interact with each other through the collective model of environment in an effort to maximize the performance from their “technological” activity. In more details their structure is considered in the next section.

Thus, living environment, in addition to i-robots, involves three types of active agents. The behavior of each type influences the functioning of other types, as all of them interact with each other in some way. Each type of active agents can be presented in a general scheme we are describing either by small number of units (for instance, from one to several numbers of technological and institutional sub systems) and by a big number (for example, i-robots and their collective models may number in dozens, hundreds and more units).

In order to apply this active agents scheme to the community of researchers in economics we should refine the content of several above described types of these agents.

In economic systems the meaning of this type of agent is quite clear. Technological communications given by the system of social division of labor, determine from whom an individual gets resources in production process and to whom gives them after processing made by him. In general case, the more number of people gains from communication namely with them, the higher will be individual’s performance from participation in social division of labor. Thus the goal of individual in the production process is such a position on social division of labor, where he will yield the maximum possible profit for the other members of community.

In scientific sphere there is also a specific “system of labor division” which shows itself in the specialization of researchers in different directions of their discipline. One type of researchers create a new knowledge, based on the results of their predecessors, and then this new knowledge is used by other researchers, etc. We can see here quite close similarity with technological system of labor division in economy. Without a great exaggeration, we can consider that a researcher makes the more gain from his activity, the more number of community’s members use (cite) his results in their works. Thus, the goal of a researcher’s activity is considered to be the achievement of such a position in division of scientific labor when his results are quoted in a maximum possible number of other researchers’ works.

If we pose a task to design the system of i-robots for scientific community then the main goal of each separate i-robot’s activity is a search for maximum favorable places in division of scientific labor for its “master” and the propaganda of its master’s results among other researchers involved in similar activity. Simultaneous activity of a set of i-robots in order to reach this goal for all members of some scientific community (for example, researchers in economics) should generate a stochastic process of establishing topic structure of a specific field (another name – “ontology”, see [Tom Gruber]). In addition there should be a relation with separate elements of ontology of researcher names’ hierarchy structure. A researcher has the more high position the more number of works made within this community are used (quoted) his results.

In economic disciplines the ontology already has a visual presentation in the view of various topic heading dividers and classifiers. Presently, the mostly used among international community of researchers in economics is JEL classifier, which has been developed and is maintained by Journal of Economic Literature (see for instance http://rvles.ieie.nsc.ru/r-archive/by_JEL.html). Thus, practical realization of a design of active agent like “sub system of technological communications” can be based on this classifier.

Summarizing described above, we can assume that major work characteristics of this type of active agent should be as follows: а)it keeps and visualizes the current configuration of ontological hierarchy either for viewing by researchers in usual mode and by i-robots;

б)it corrects and develops basic ontology as well as researchers names’ hierarchy linked with this ontology on the basis of the results obtained from an agent “collective model of environment” (see below).

It is also necessary to provide a possibility for researchers to create completely new sub systems of technological communications, which may give to a researcher community new principles of organization of technological communications and of calculation of what this community gains from the activity of its members. Thus a set of agents of one type competes between each other for attracting to its side a maximum big number of community’s members.

Active agent of this type is a keeper of behavior rules established in this community (special-interest group). As soon as a new researcher joins this community, his i-robot receives a model of behavior from this active agent. I-robots, on the order of their masters, can put in a collective model of environment proposals on changing a current system of behavior rules.

It is also necessary to provide a possibility for researchers to create new sub systems of behavior rules, which will be able to compete with each other and with the current behavior rules within a collective model of environment. However, at each instant of time there is actually the only one agent of this type who acts. There always should be the one sub system of behavior rules while there could be several sub systems of technological communications acting simultaneously. Current sub system of behavior rules can be changed by new one if and only if an agent “collective model of environment” gives such an instruction.

I-robots of separate researchers are monitoring references/quotations of results/publications of their masters over the input flow of new documents into RePEc database. Based on these results, robots may establish in semi-automatic mode a new structure of ontology and/or hierarchy of names linked with this ontology. They put these new images of ontology into collective model of environment with the aim to make a proposal how to change the current structure of ontology. Such proposals made by i-robots can have references not only on the achievements of their masters but also on any number of researchers from this community. All set of proposals can be ordered in a collective model of environment, for example, over the results of “voting” of the community members. Under certain conditions, a proposal, which has the highest rating, will be adopted and then it is transferred to an agent, responsible for the ontology. This agent rebuilds ontological hierarchy on the basis of obtained information.

On a similar scheme can be organized a procedure of discussing and changing of the current state of “behavior” agent.

To our opinion, a separate i-robot should consist of two main sub systems, to say the least:

1)sub system of tuning to users’ “interests profile”, convenient interface for setting a behavior models of i-robot, and the means for visualization the results of this robot’s work;

2)sub system of online performing the user’s orders including the means for interacting with community of i-robots, which belong to other researchers as well as with other types of active agents creating a living environment for this community of i-robots as a whole.

Conception part of a task on creating an i-robot is, first of all, the development and coordination of standards for functional capacities of i-robot and also the development of protocols and procedures of robots’ interaction (the latest refers to the standards for a living environment). This stage is needed for providing a basic compatibility of i-robots, which can be developed independently in different organizations.

Functional part of the task is the development of minimum set of i-robot functions and definition of a protocol on adding new functions to a basic list without losing compatibility over crossing positions from this list.

Basic set of functions may involve the following:

а)visualization of the content of input information flow, including the filtration of input documents flows according to interests profile set by a user, then displaying them in a suitable view (see the description of a conception of web-portal for RePEc data base in [T. Krichel and others]);

b)”communication” with an agent “collective model of environment”, including the construction of input documents flow on the basis of given set of user’s “interests”, also the development of proposals either on linking separate elements of input information flow to the current ontology and on developing main ontology on the basis of widening the subjects of input documents flow; in addition, the development of proposals on changing the agent “behavior rules” and transmitting all recommendations to the agent “collective model of environment” (in general case, recommendations can have a wide nature);

c)direct communication with i-robots of other researchers, for example questioning of other researchers’ i-robots for following the instructions of robot’s master on collecting specific data;

d)”negotiations” with the i-robots of other researchers for preparing and developing preliminary agreements among the members of researchers group and so on.

The speedy increase of intensity of input information flows in RePEc data base, which originated from a widening online activity of international community of researchers in economics, is the main impetus to the search for approaches of how to solve a problem of expected information overloading. It is required to develop such a program and technical means that will enable to keep direct and prompt interactions (traditional for research groups) among scientific community in the situation when the total number of participants of common information space is sharply increasing, like the average information activity of each participant is also increasing. In this case the major task is to develop program agents which may take up both the part of daily work of a researcher on filtration input information flow, and some efforts on communicating with scientific community, namely, information exchanges with colleagues, preliminary coordination of mutual activities etc.

The development of such program agents or i-robots is already in the process, but this work has some gaps connected with the aspects of interaction among a large number of i-robots, as well as with the organization of a pattern of their artificial community as an environment for maintaining activity of i-robots.

One can assume that a living environment for i-robots can be created by online activity of active agents belonging to some specific type. Among them agents that visualize “technological” links between the members of community. The use of these links is the main activity of researchers for the benefit of each participant and community as a whole. In addition to above described agents, there are another agents acting in the living environment which ensure the observance of behavior rules, and agents establishing the conditions for direct interactions between i-robots (collective model of environment).

Simultaneous functioning of agents of all types should create favorable conditions for self-organization and development of a community of researchers in economics under the current situation when information pressure on each member of this community takes place.

T. Krichel, V. Lyapunov, S. Parinov. Online Scholarly Information for Economics: The RePEc database and the RuPEc web portal, http://rvles.ieie.nsc.ru/~parinov/papers/repec-rupec.htm   (in Russian)

S. Parinov^A. Toward a Theory and Agent-Based Model of the Networked Economy, http://rvles.ieie.nsc.ru/parinov/theory/eng/

S. Parinov^B. The third form of a regulation mechanism for the networked economy, http://rvles.ieie.nsc.ru/parinov/net-form.htm   (in Russian)

Leigh Tesfatsion. Pointers to ACE/CAS-Related Research Groups and Sites, http://www.econ.iastate.edu/tesfatsi/allist.htm

Philip E. Agre. Computational Research on Interaction and Agency, http://weber.ucsd.edu/~pagre/aij-intro.html

COGNET, GINA and iGEN™ - The Cognitive Agent Software Toolkit, http://www.chiinc.com/ginahome.shtml

Aaron Sloman. The Cognition and Affect Project THE SIM_AGENT TOOLKIT, http://www.cs.bham.ac.uk/~axs/cog_affect/sim_agent.html

The Web Robots Pages, http://info.webcrawler.com/mak/projects/robots/robots.html

The AmEC Initiative, http://ecommerce.media.mit.edu/top.html

BOTSPOT, http://www.botspot.com/

Agents Learning about Agents, http://jmvidal.ece.sc.edu/alaa/

Peter Danielson. Evolving Artificial Moral Ecologies, http://eame.ethics.ubc.ca/

Collective Intelligence, http://ic.arc.nasa.gov/ic/projects/collective-intelligence.html

MultiAgent systems, http://multiagent.com/

Technology of Active Objects (TAO), http://www.rriai.org.ru/TAO/

Tom Gruber. What is an Ontology? http://www-ksl.stanford.edu/kst/what-is-an-ontology.html