The network argument

Rainer Fischbach

Contributed to the workshop on Multitude by Hardt/Negri at RLS, December 14, 2004

A central argumentative figure in the book refers to networks: It is based on the assumption that there would exist a certain category of things - networks - which are intrinsically endowed with certain features: having no centre, being flexible, nonhierarchical, delocalized, able to grow indefinitely, and therefore being almost undefeatable, open and democratic as well, etc. If something is said to be a network, it is implied that it should also have these features. Sometimes something is said to be a distributed network, in order to add emphasis to this implication.

The problem with this argumentative figure is that it rests on unwarranted assumptions: Networks don't form an ontological category. Of course, here exists a class of physical things which are genuine nets or networks: some of them human-made like certain textiles used as dressing, in the socker goal or for catching fish, some of them animal-made like the net produced by a spider to catch flies. But besides these, it is just figurative talk, if technological artefacts, organizations or the social relations individuals are involved in are called networks.

As a metaphor, the term network doesn't have precise implications. It's actually quite unclear what it means to be a network in the figurative meaning, and many of the features associated with the network metaphor are not to be found in physical netorks: A spider web has a centre, no pysical net can grow indefinitely without tearing under its own weight, physical nets have only limited flexibility, they are localized, etc. Besides that, it is no sound scientific practice to base an argument on the accidental connotations of a metaphor.

Beyond the figurative use, the talk of networks can be made more precise by giving it a modellistic semantic. The mathematical theory of graphs and networks, the latter being graphs with arcs labeled by nonnegative numbers, can be employed to construct models of anything. Anything may be modelled as a network! Even a solid body - which is actually done if the dynamics of engineering artefacts like aircraft, cars, etc. is studied by finite element methods (FEM). The modellistic use of networks doesn't carry any implications of the kind assumed by Hardt/Negri, or Manuel Castells, either. The properties of a network model depend on the way the model is constructed. If anything can be modelled as a network, then there are, by implication, also hierarchical, centralized, inflexible, limited, etc. networks.

If the modellistic approach is taken, its not possible to confront networks with non-networks any more, e. g. al-Queda and the nation states, as Hardt/Negri do. That would mean just messing up categories. Taking the modellistic approach, both, al-Queda and the nation states should be modelled as networks using consistent criteria and mappings. When modelling by networks, it should be made clear what we are modelling: E. g. if we are talking about the Internet, we should make clear, if we are talking about the technical infrastructure - and, if that is the case, about which of its physical and software layers - or about social entities communicating via the internet.

It's all possible: a hierarchical organization communicating via a nonhierarchical infrastructure and, vice versa, a nonhierarchical organization communicating via a hierarchical infrastructure. And, by the way: as a technical entitity, the Internet is far more hierarchical, centralized, and, of course, localized, as is commonly believed. If it would look like the image almost everyone has of it, it would be rather inefficient, inflexible and costly, and, of course, have breaken down since long time ...