Problems of Automation in the 1960s
In the 1960s, what did Luhmann know about digitisation and algorithms? It may sound like a rhetorical question, but Luhmann already knew a surprising amount of information about these subjects, which I should like to demonstrate by way of analysing the 1966 text Recht und Automation in der öffentlichen Verwaltung (Law and Automation in Public Administration). I use quite a few direct quotations from Luhmann’s text, which is unusual for me, but I hope that the quality of the citations will speak for themselves. Of course the argument in the end is my own, yet it has been inspired by and constructed with material from Luhmann’s text.
Big data and self-learning algorithms obviously did not yet exist, but Luhmann was already anticipating them, as various passages show. For example, he suggested that the idea that machines cannot solve system problems through other than purely logical means:
[…] will probably one day be unhinged with the counterargument that you can teach a machine leaps of logic and that it can clarify the admissibility of such leaps better than humans. The Carnegie Institute of Technology, Pittsburgh, is already working on computer programmes for problems which are unclearly defined and [these programmes are] designed to imitate and perhaps even surpass the human approach […] And every time that those advocating the use of human beings precisely formulate their reasons, at the same time they create the basis for the formulation of new equivalent machine programmes.
The current competition/cooperation between humans and machines in the development of game algorithms (and machine-learning in general) could hardly be better described. For example, there is the famous case of DeepMind’s ‘AlphaGo’ programme that plays the Chinese board game Go, in which the machine learned from humans how to beat the best human players, who then learned from the machine how to improve their strategies, leading to a further improvement in the programmes. Luhmann was also informed about attempts to build ‘a general problem-solving computer programme with learning possibilities’, which is what we have today with self-learning algorithms.
The acuity of Luhmann’s text, however, lies not only in his foresight, but above all in his approach.