ISKO est une association internationale pour l'étude des organisations. L'adresse est : http://www.isko.org
Nous reproduisons ici le projet de recherche développé depuis 2004 par des membres italiens de l'ISKO. Il reprend les idées de Needham et de Hartmann sur les niveaux d'intégration et la classification des savoirs qui s'ensuit.The aim of the project
"Reality is the basis for the texts of documents; that is what authors try to describe, and what searchers are investigating. Classification, the investigation of structure, is the foundation of all our knowledge, and therefore should be the foundation of all our systems for controlling information."(Classification for a general index language. p 45 / D.J. Foskett — the Library association : London : 1970)
Aim of the project is to test application of the theory of integrative levels to knowledge organization (KO). This implies a naturalistic-ontological approach to KO, which is obviously not the only possible approach – actually it even looks to be unfashionable nowadays, although it agrees with current trends towards interdisciplinarity and interrelation between many research fields.
Theory of integrative levels claims that the natural world is organized in a series of levels of increasing complexity: from physical particles and molecules, through biological structures, to the most sophisticated products of human thought. Each level cannot exist without the lower ones (EG there are no organisms not being formed with atoms), but at the same time it has additional emergent properties not found at the lower levels (EG organisms can be said to be alive or dead, while atoms cannot). This view goes beyond the traditional opposition between reductionism and vitalism, both of which have important limitations.
The ideas of integrative levels and emergence can be found in some form in Western philosophy since at least 19th century, and generally agree with the naturalistic point of view adopted by many modern scientists. However, it was only later that they were formulated in a more explicit way, by philosophers James K Feibleman and Nicolai Hartmann . Hartmann's work is especially relevant for the foundations of ontology, the branch of philosophy dealing with the structure of reality.
The theory of integrative levels was first considered as a basis for knowledge organization during the 1960s by the Classification Research Group (CRG), under influence of biologist Joseph Needham. CRG members reorganized Feibleman's laws of levels, and analyzed their possible application to the construction of a new general classification scheme. They produced a draft scheme, but did not develop it further.
However, integrative levels have been used as a principle for establishing main class order in other general classification systems, namely CRG's Bliss classification 2nd edition (BC2) and Broad system of ordering (BSO), and Dahlberg's Information coding classification (ICC). Older schemes such as Brown's Subject classification (SC) and Bliss's Bibliographic classification (BC1) also have their main class order based on quite the same idea.
Some schemes provide for non-disciplinary classes devoted to single phenomena (EG, horses instead of zoology or veterinary medicine), and it seems a logical development that these classes would be structured by integrative levels; classes for phenomena and classes for disciplines could be connected through an "accordion device" . Some authors also wish the development of completely new schemes independent on disciplines, allowing for better multidisciplinary search.
Once the conceptual framework has been drafted, it should be formalized into principles regulating relations between classes and notation expressing them. This implies several questions.
A way to classify objects is to compare the number of their shared characters: this is the approach of numerical taxonomy. So dolphins and tunas will go in the same group as they look similar. However, this technique alone does not lead to generalizations as useful as if common origin is taken into account: the most effective taxonomy is one which takes into account both common origin and similarity. Roughly, when a novel level originates, it can be represented as a vertical branching (splitting) from an existing level, while diversity among phenomena belonging to a same level can be represented in the horizontal dimension.
Many higher level phenomena actually originate from more than one lower level – EG, technological artifacts depends both on human activity and on physical matter used as material. Hence our "branching" actually is a network of relationships including rhomboidal figure, with different paths leading to the same node, IE a directed graph with absorption. We need to define explicit rules allowing to translate the multidimensional network model into a linear model expressed by notation, as typical of bibliographic classification schemes .
Another problem is how to quantify diversity among phenomena, in order to place them in the appropriate positions in the graph. To compare phenomena of a same kind, like tigers vs. lions, they can be described by an ordered set of common measurable characters, so that the distance between vectors representing them in a multidimensional space be a measure of their diversity. However, in order to compare phenomena at different integrative levels, like tigers vs. tropical ecosystems, we can't describe them completely by a common set of characters, as by definition the higher integrative level has new properties which are nonsense at the lower levels. This leads to the search for an absolute measure of complexity, a typical problem in contemporary complexity theory. The concepts of algorithmic complexity and of logical depth look as promising tools in this direction.