Juval Portugali

Department of Geography and the Human Environment
Tel Aviv University
Tel Aviv 69978, Israel
Email: juval@post.tau.ca.il

Self-organized Inter-representation networks

A proposal for the Varenius workshop on cognitive models of dynamic phenomena and their representation

1. The contribution

The ability to manipulate, interpret and store information about changing environment is a critical skill for human survival and also is very important for geographic information science (from the Veronius Initiative).

According to the IPA (information processing approach), which currently dominates cognitive sciences, this is so because we use the cognitive system as a metaphor to build hardware and software, and the latter as metaphors to understand the operation of the human cognitive system.

The basic assumption of the IPA is that the cognitive system and the environment are external to, and thus fully independent of, each other. In the context of the Varenius workshop this implies that human cognitive abilities to manipulate, interpret and store, information, and, computers and geographical information systems, are external to each other and co-exist in causal relations. The notion of IRN suggests a somewhat different view, the basic tenets of which are the following:

1. The human mind and the human environment are only relatively independent of each other.
2. Humans have an innate capability for representation that comes in two forms: external and internal.
3. The boundaries of their cognitive system extend beyond the brain/skull and includes, first, the body and second, stand-alone artifacts in the environment.
4. The cognitive system is composed of internal representations in the mind/brain as well as of biological, and artificial, external representations in the environment.
5. The dynamics of the system involves an on-going interaction between internal and external representations.
6. External representations provide the link between individual and collective (cultural and social) cognition.
7. The cognitive system in general, and the one associated with cognitive mapping, is a self-organizing system, the dynamics of which is described by Haken's theory of synergetics.
One of the major implications of IRN is that artifacts such as tools, buildings, cities, maps, computerized information nets and the like, are external representations and as such integral components of the cognitive system. That is to say, not only that the ability to manipulate... information . is a critical skill and so on, but that many information systems, GIS included, are externally represented artifacts that often function as integral component in the cognitive systems of individuals and collectivities.

From this perspective, the central question of the initiative, is not just a question about the relations between human cognition, geographic space and different kinds of computerised representations and display, but of the various elements of the cognitive system itself.

My contribution to the Varenius workshop will elaborate on the philosophical issues raised above and their entailed operational and modeling implications.

2. Research Abstract

The research abstract that I describe below falls into two parts. The first, philosophical, discusses the general nature of IRN and its implication to GIS, and the second, operational, shows how the abstract ideas can be cast into formal algorithms and simulation models. Some of the issues of this research abstract, (namely, sections 1, part of 2, and 6) have already been published, others will be presented here for the first time. The emphasis of my contribution to the Varenius initiative will be on the latter. The various bibliographical references made below can be found in Portugali (1996) and in Haken and Portugali (1996)

Part I: General discussion

1. A short remainder of IRN. This will be based on Portugali (1996). In this introduction the concept of IRN is constructed, first, by reference to its main inspirational sources, namely, Bohmís philosophy of implicate and explicate order, and Hakenís synergetic approach to self-organization. Second, by reference to the writings of several cognitive scientists who implicitly or explicitly recognized the role of external representations within the overall process of cognition. I'm refering to people like Vygotsky, Gibson, Bartlet檉, Rumelhart et al., Johnson and Lakoff, Adelman, and others. Of the latter, Barttlet븢scenarios of serial reproduction are used as metaphors to convey and model processes of IRN.

2. Some empirical examples for its operation. The latter include, among others, (i) the Bartlett scenarios of serial reproduction, as devised by him in his study of remembering, (ii) city-games, which are public-collective serial reproductions that we have devised within the context of our IRN research. (iii) Golani et al. experiments with rats, which according to our interpretation show how the external and internal spaces are simultaneously constructed, and, (iv) several experiments we are currently conducting on emotional effects during navigation. The latter examine the hypothesis that externally represented body effects participate in the process of learning and navigation. These (and other) experiments illustrate the interplay between internal and external representations, and the self-organizing nature of the process.

3. On the relations between IRN and the foundations of cognitive mapping. In this section I show that the notion of IRN is already implicit in the elementary ideas of the founding fathers of cognitive mapping. On the one hand, in Tollman, who pioneered the concept cognitive map as an internal representation within the frame of externally represented behaviorism. On the other, in Lynch who, in his The Image of the City, has elevated the role of five elementary artifacts (paths, areas, junctions, nodes, landmarks) as legible external representations with which one builds the image (i.e. internal representation) of the city.

4. IRN and the computer metaphor. Iíll show that the very Turing Machine, which is so central to the IPA and computationism, is essentially an IRN Machine. This claim is based on a new reading of Turing.

5. On the biological dimension of IRN. Iíll show that the relations between internal and external representations that is central to IRN, typify also other biological systems. Iím refering to the relations between genotypes and phenotypes as presented by Dawkins in connection with his concepts memes and extended phenotypes.

Part II: Operational discussion

6. SIRN (synergetic inter-representation network). The first step toward operational implementation is the model SIRN developed by Haken and Portugali (1996). This model casts the notion of IRN into the formalism of synergetics Ė Hakenís (1996) theory of self-organization. The building of SIRN starts with the so-called synergetic computer and adds to it externally represented inputs and outputs. The synergetic computer is a fully parallel computer, the neural network algorithm of which represents an alternative to the conventional neural network model.

7. The proper method of representation. Like all neural networks the structure of Hakenís synergetic computer and consequently of our SIRN model, metaphorically mimics the neural structure of the brain. As long as one deals with internal representations only, this is indeed an advantage. However, when external representations are added as integral elements of the model, we are facing a problem. On the one hand, we have a neural net that enfolds information, while on the other, artifacts (buildings, cities computer systems, etc.) that enfold information. The challenge is to go beyond neural nets and artifacts and define a medium of representation that is appropriate for both.

8. The cultural code. The variable cultural code attempts to go beyond neural nets and artifacts and can thus be a medium of representation that is appropriate to both. It defines each individual cognitive system (each human individual) by means of internal and external representation, as in our SIRN model, and each representation, internal or external, by means of a cultural code, reminiscence of a genetic code. Technically, the cultural code can be defined by a Boolean vector. The theoretical foundation to this analogy between genetic and cultural codes is based on section 5 above. A graphical representation of this model is given below.

This page is currently maintained by Varenius Workshop Webmaster
Last Updated: Sep. 18, 1998