Luc Legay (lucru3.org)  - 3 feb, 2003 | Last modification : 15 may, 2003 - This is a working paper, some translation in english to come. (translation from french : Jacques Ampolini.)

Please seen up to date version here: RU3 Wiki Project (in French)

Discover the RU3 Project (En) | Projet RU3 (Fr) | RU3.com (Blog)

Keywords: collective intelligence , connective intelligence, user networks, open networks, semantic web, fuzzy interfaces, groupware, contribution, interaction, information, intelligence flow, theory, collaborative software, design, conception.

 

 "Nobody knows everything… Everybody knows something…"  
Pierre Lévy

 

 

 

The open networks of collective intelligence

 

The intelligence we are interested in emerges while information is being exchanged amongst a group of individuals.

A large number of individuals gathering in persons, at the same time and in the same place, rarely leads to collective intelligence, because the information exchange fluidity is met with obstacles that are often insurmountable.

Thanks to modern communication systems, the obstacle to exchange fluidity seems to be overcome. However, if message exchange between two persons seems greatly facilitated, these exchanges quickly become difficult to manage as soon as the number of exchanged messages, as well as the number of persons involved, grows.

Today, one realizes that potential access to a large volume of information, or to a large number of individuals, is not sufficient to access to more intelligence.

Access to large volumes of information requires interfaces able to organize, structure and hierarchically set, into elementary pieces of information, those that are too complex, too large, or just too numerous, and that, as a result, cannot be assimilated by one person in a given time.

Descending communication from mass media

In order to communicate to the most important number of individuals, the information society, has made every effort to produce messages that are simplified, normalized and smooth, and that are destined to satisfy a population of individuals but no individual in particular.

Mass information is, most often, delivered as "take it or leave it". This is the only alternative left to the addressee.

Mass communication, or descending communication, is the media application of the communication system established by Claude Shannon with the mathematical theory of information. We owe to Warren Weaver the fundamental diagram of communication systems shown below.

Communication, as it is described in the theory, is not concerned with the semantic aspect of the message being sent. Information is described, in the message, as an increasing function of the uncertainty reduction that it brings; this makes this theory essentially a static one.

From information flow to intelligence flow

In order to set networks able to support flow of intelligent pieces of information, it is necessary to redefine the theory of information.

The theory, which in reality is only concerned with signals, does not take into account the meaning of the information being moved. This is the reason why we propose the design of a less restrictive theory, yet more ambitious: a general theory of intelligence flows.

By integrating in this new theory the sciences of signs, that is semiology, one can see that a fundamental notion, retroaction, was not part of the previous theory.

Roland Barthes gives the following definition of semiology: "a science that studies the life of signs within the social life".

The semiotic triade is a schematic representation of the sign interpretation proposed by the American logician Charles S. Peirce (1839-1914). It makes it possible to understand the place of the interpreter in relation to the object, and the sign that represents it.

The necessity to develop new languages, particularly for robotic applications, made it possible to improve our understanding of the mechanisms involved. The semiotic cycle demonstrates that a representation cannot exist without the sharing of a common space between the one who issues and the one who interprets. (From the kind contribution of Luc Steels, VUB AI Lab, Brussels and Sony Computer Science Lab, Paris)

The two-way communication of open networks

Access to collective intelligence is based upon the information user involvement.

The users access a volume of information that they have structured using their own criteria. Initiative belongs to the users, or their agents, that is to say to the programs able to implement searches and analysis on their behalf.

Within this schematic representation, communication is first ascending since the information contents are made available to the users or to their agents. Information is published under an open and editable form.

This means that users can not only freely access the contents, they can also directly intervene on these contents. Communication is then two-way since user-editor interventions are immediately perceived by the others users.

Yet, experience shows that it exists as many access logic as content authors (for instance, refer to some of the six billions Web pages…). The interface function also consists in adapting this diversity of contents to the diversity of user interests.

The intelligent open network finality is to make intelligible, to a given user, a large volume of information, or intense flux of information exchange.

The production cycle of collective intelligence

The production cycle can be broken down in three fundamental phases, production, usage and information exchange. This general schematic representation includes the schematic of the production and information access through complex communication networks, like Internet; yet, it remains valid within a much larger frame.

From the upper part of the cycle we can distinguish:

  • Contents interpreted by the user and considered as intelligible.
  • Conceptualization of information on knowledge.
  • Partial exploitation of acquired knowledge.
  • Interactions by modification of the existing representations or expression of new ideas/contents by contribution.
  • Representation of ideas or contents on a shared media

From the lower part of the cycle, one can distinguish:

  • Representation captions by an agent program according to criteria defined by the user.
  • Content aggregation in meta-databases structured according to personalized criteria.
  • Data analysis.
  • Data interpretation presented through a fuzzy interface.

- -> Back to the upper part of the cycle.

Interface models adapted to the intelligence phase of information

To collect quickly daily updated contents, new interfaces are necessary.

The content aggregators allow for visualization, through a unique window, of information coming from different sources. The aggregators themselves possess no content. Actualized data is simply presented in a single interface, using similar presentation logic, and not through as many interfaces as information sources anymore.

News wire agregator NetNewsWireLite

Mode of operation of an information aggregator :

 

 

The current aggregators operate only from contents that have adopted a normalized structure, like for instance the RSS norm. Our project proposes the installation of more powerful aggregation systems that operate with unstructured contents, or structured but using different norms. The concept is to generate a structure that does not affect the source content. These systems can be structuration interfaces like that presented below.

 

 

Data structuration interface (ru3.org project)

The interface is directly usable, after a training phase, by a user or a software agent. Each window of the structuration tree relates to a hierarchy. The tree libraries, or nurseries, allow for choices of structuration appropriate to the contents that are to be structured.

Structured contents for the users

The structuration that the authors want does not necessarily correspond to the user's needs. We currently know where the sources of information are localized (web, intranet), but we don't know who are, or who will be the users of this information:

  • Which language do they speak?
  • Are they adults or children?
  • What is their capacity of comprehension?
  • What is the depth of their vocabulary?
  • In which context will they access the information?
  • Are they in a public place or a private place?
  • Are they in a hurry or not?
  • Is the tool of information's access able to reproduce correctly all the information?
  • Are they looking for a particular piece of information?

In such a context, a piece of information cannot be qualified using an established terminology because this terminology can vary from user to user. It cannot be labeled either in an absolute and definitive form since context, time, fashion and environment can change it.

This is the reason why the principle of semantic web, previously developed at W3C, seems to us a utopian quest and a program developed for the machines rather than the users. Opposite to the semantic web, the RU3 project founding principle considers, a priori, that the existing information sources, whether structured or not, cannot be standardized but instead, a posteriori, that they can be qualified, in particular during the interactions with users.

Interaction information is in the center of open networks

What is the nature of these interactions and what are the means at our disposal to intercept them?
Because information contents, like ideas, cannot be categorized in a formal manner, we are going to qualify them in relation to their use and their users. Let's remember that in an open network the users broadcast their own contents that, then, become accessible by any other user with no restriction.

Nowadays, open networks are made up using such means as: wiki, weblog, moblog, slashdot, forum, chat, IRC... Since the published content is in a language shared with other users, propounders presented by one user can feed other users' propounders. Information interaction comes from three essential characteristics of these new tools:

  • The rapidity with which idea and content publication is implemented
  • The remanence, or persistence, of the propounders
  • The magnitude of the potential audience

Within the open networks, an uninteresting discussion builds up no audience, just because no other users pick it up. On the contrary, a propounder that is federalizing, mobilizing, interesting, understood, whether contested or approved, will rapidly build up an audience among the community of open network users.

Accordingly, the value of a piece of information, or that of an idea, is related to the number of interactions induced within a community of users of this information, or this idea. And the more they are mobilizing and shared, the more interactions they create with other communities.

To measure the relevance of information, it is to measure interaction

Thus, we are assisting here at a complete redefinition of the pertinence of a piece of information based no longer on an absolute value, frozen in time and labeled as reference, as true, or as disinformation. Information becomes relative to the way it is being used because it becomes possible, within the open networks, to make use of it, to reformulate it or to contest it.

Within this context, one better understands that any tentative measurement of content pertinence, and more broadly of information, is in fact the measurement of interactions between information and information users.

Making possible the measurement of the interactions is also making possible the detection, the marking and the use of pertinent information. The access interface to this type of meta-information is foremost an interface that makes possible the improvement of the information signal/noise ratio. It is an interface that permits masking part of the non-pertinent contents.

The fuzzy interfaces, in reference to the mathematical logic of fuzzy ensembles, permits manipulation of coherent information ensembles and not any longer of individual and delimited pieces of information.

New access interfaces

This chapter is being debated on wiki Web site (in French). The concepts that are being developed:

The final White Paper will be finalized starting from these pages.

Project Web site : RU3 Project | Wiki Web site RU3 Wiki Project (in French)

This work is dedicated to the Public Domain.