Intelligent Agents in the Societies of Control
In his article "Meshworks, Hierarchies and Interfaces", Manuel De Landa
depicts a bright future of autonomous software agents, which will lead to
an individualization of working places through their ability to adapt to the
individual users demands. Although in the last paragraph he points out that
"meshworks", or anti-hierarchical, heterogeneous networks, are not to be
glorified, in telling the story of the standardization and homogenization of
social actions throughout Modernism, he assumes that Behavioral AI,
agents that learn from the actions of an individual user, provide a way of
decreasing (hierarchical) control.
The problem with this presumption is, as Gilles Deleuze pointed out in
"Postscript on the Societies of Control," that today it is no longer
enclosure and homogenization which upholds the systems functioning.
By characterizing the difference between the mechanisms of discipline and
control, Deleuze states that "enclosures are molds, distinct castings, but
controls are a modulation, like a self-deforming cast that will continuously
change from one moment to the other, or like a sieve whose mesh will
transmute from point to point." Here we meet De Landa's "meshwork"
again, this time not as a mechanism of liberation, but as control.
In the societies of control, the signature that designates the individual and
the number which indicates his or her position in the mass is neglected in
favour of the code, which regulates access to information. The subject is
caught in a continuous network, where individuals are no longer constituted
as one body, but where rivalry and continuous modulation - also within the
production force - serve as a perfect guarantee for the functioning of late
The notion of control is inherent in De Landa's argumentation, when he says
that "the agent in question would be constantly looking over the users
shoulder keeping track of whatever regular or repetitive patterns it
observes." He argues that the learning process of the agents will result in
an individualization of the working process. This will prevent homogeneity
evoked by the profile of the "ideal user" contained in the Symbolic AI
But Critical Art Ensemble pointed out that if an individual is convinced he
or she is being observed, it is less likely to behave in an a-normal way (see:
"Utopian Promises - Net Realities"). Today, companies observe their employees by
counting - for example - the frequency of keystrokes. The distinction between
efficient and non-efficient behavior, a certain homogenization of the working
process, is essential, at least in the field of production. By analyzing patterns
of behavior a pre-classification of individuals can become part of everyday life:
in health-care systems, risks for getting certain illnesses are used for
treatments and insurance calculations and risks for a-normal behavior and
deviation can be used for calculating job profiles or salaries.
De Landa's light-hearted enthusiasm about the liberating force of autonomous
software agents ignores the possibility that today the power structure
itself may already be nomadic and rhizomatic. No matter what the user does,
her or his action "gives feedback to the agent" (De Landa). This may support
the society of control in a better way than De Landa thinks.
Critical Art Ensemble: Utopian Promises - Net Realities.
De Landa, Manuel: Meshworks, Hierarchies and Interfaces.
Deleuze, Gilles: Postscript on the Society of Control.
+ + +