Niklas Luhmann (1927-1998) was a German sociologist who aimed to understand society in terms of systems theory.
I am reading Luhmann’s Social Systems (1995) because I have a hunch that this theory is relevant to my research. This post contains notes about Chapter 1, section I-II.
Often, scientists need to sacrifice intelligibility for accuracy. Luhmann is a scientist. He is unapologetic about this. He opens his “Instead of a Preface to the English Edition” (actual title) with:
“This is not an easy book. It does not accommodate those who prefer a quick and easy read, yet do not want to die without a taste of systems theory. This holds for the German text, too. If one seriously undertakes to work out a comprehensive theory of the social and strives for sufficient conceptual precision, abstraction and complexity in the conceptual infrastructure are unavoidable.”
Why bother reading such a difficult book? Why be a scientist and study social systems?
One reason to study society scientifically is to design and build better smart digital infrastructure.
Most people designing and building smart digital infrastructure today are not studying Luhmann. They are studying computer science. That makes sense: computer science is a science of smart digital artifacts. What has become increasingly apparent in recent years is that smart digital infrastructure is having an impact on society, and that the infrastructure is often mismatched to its social context. These mismatches are often considered to be a problem. Hence, a science of society might inform better technical designs.
Chapter 1 opens with:
The following considerations assume that there are systems. Thus they do not begin with epistemological doubt. They also do not advocate a “purely analytical relevance” for systems theory. The most narrow interpretation of systems theory as a mere method of analyzing reality is deliberately avoided. Of course, one must never confuse statements with their objects; one must realize that statements are only statements and that scientific statements are only scientific statements. But, as least in systems theory, they refer to the real world. Thus the concept of system refers to something that is in reality a system and thereby incurs the responsibility of testing its statements against reality.
This is a great opening. It is highly uncommon for work in the social sciences to begin this way. Today, social science is almost always taught in theoretically pluralistic way. The student is taught several different theories of the same phenomenon. As they specialize into a social scientific discipline, they are taught to reproduce that discipline by citing its canonical thinkers and apply its analytical tools to whatever new phenomenon presents itself.
Not so with Luhmann. Luhmann is trying to start from a general scientific theory — systems theory — that in principle applies to physical, biological, and other systems, and to apply it to social systems. He cites Talcott Parsons, but also Herbert Simon, Ludwig von Bertalanffy, and Humberto Maturana. Luhmnn is not interested in reproducing a social scientific field; he is interested in reproducing the scientific field of systems theory in the domain of social science.
So the book is going to:
- Be about systems theory in general
- Address how social systems are a kind of system
- Address how social systems relate to other kinds of system
There is a major challenge to studying this book in 2021. That challenge is that “systems theory” is not a mainstream scientific field today, and that people that do talk about “systems” normally do so in the context of “systems engineering”, to study and design industrial processes for example. They have their own quantitative discipline and methodologies that has little to do with sociology. Computer scientists, meanwhile, will talk about software systems and information systems, but normally in a way that has nothing to do with “systems theory” or systems engineering in a mechanical sense. Hazarding a guess, I would say that this has something to do with the cybernetics/AI split in the second half of the 20th century.
There is now a great deal of convergence in mathematical notation and concepts between different STEM fields, in part because much of the computational tooling has become ubiquitous. Computational social science has made great strides in recent years as a result. But many computational social science studies apply machine learning techniques to data generated by a social process, despite the fact that nobody believes the model spaces used in machine learning contain a veridical model of society.
This has led to many of the ethical and social problems with “AI”. Just for brief example, it is well known that estimating fitness via regression for employment or parole from personal information is, even when sensitive categories are excluded, likely to reproduce existing societal biases extant in the data through proxy variables in the feature set. A more subtle causal analysis can perhaps do better, but the way causality works at a societal level is not straightforward. See Lily Hu’s discussion of this topic, for some deeper analysis. Understanding the possible causal structures of society, including the possibility of “bottom-up” emergent effects and “downward causation” effects from social structures, would potentially improve the process of infrastructure design, whether manual or automated (via machine learning).
With this motive in mind, we will continue to slowly analyze and distill Luhmann in search for relevant insights.
For Luhmann, “systems theory … claims universal validity for everything that is a system.” Implicitly, systems theory has perfect internal validity. Luhmann expresses this theory in German, originally. But it really feels like there should be a mathematization of this work. He does not cite one yet, but the spoiler is that he’s eventually going to use George Spencer-Brown’s Laws of Form. For reasons I may get into later if I continue with this project, I believe that’s an unfortunate choice. I may have to find a different way to do the mathematization.
Rather, he follows through on his commitment to the existence of real systems by inferring some necessary consequences of that first principle. He is not content with a mathematical representation; systems theory must have “a real reference to the world”; “it is forced to treat itself as one of its objects in order to compare itself with others among those objects”. The crux is that systems theory, being a system itself, has to be able to take itself into account from the start. Hence, the commitment to real systems entails the realness of self-referential systems. “This means … there are systems that have the ability to establish relations with themselves and to differentiate these relations from relations with their environment.”
We are still in §I, which is itself a sort of preamble situating systems theory as a scientific theory, but already Luhmann is exposing the substance of the theory; in doing so, he demonstrates how truly self-referential — and consistently so — systems theory is. As he’ll say more definitively later, one essential feature of a system is that it is different from its environment. A system has, in effect, an “inside”, and also an “outside”. Outside the system is the environment. The part of the system that separates the inside of the system and its environment is the boundary. This binary aspect of the system (the system, and the not-the-system (environment)) clarifies the logic of ‘self-reference’. Self-referential systems differentiate between themselves and not-themselves.
So far, you have perhaps noted that Luhmann is a terribly literal writing. It is no surprise that the focus of his book, Social Systems, is that subset of systems that are “social”. What are these systems like? What makes them different from organisms (also systems), or systems of machines? Luhmann eschews metaphor — a bold choice. “[W]e do not choose the shortcut of analogy, but rather the longer path of generalization and respecification.” We don’t want to be misled by analogies.
“Above all, we will have to emphasize the nonpsychic character of social systems.”
That’s something Luhmann says right after saying he doesn’t want to use metaphors when talking about social systems. What can this possibly mean? It means, among other things, that Luhmann is not interested in anybody’s subjective experience of a society as an account of what a social system is. A “psychic system”, like my lived experience, or yours, is not the same thing as the social system — though, as we will later read, psychic systems are “structurally coupled” with the social system in important ways. Rather, the social system is constituted, objectively, by the communications between people. This makes it a more ready object of science.
It is striking to me that Luhmann is not more popular among analysts of social media data, because at least superficially he seems to be arguing, in effect, the social system of Twitter is not the system of Twitter’s users. Rather, it’s the system of the tweets. That’s one way of looking at things, for sure. Somewhat abashedly, I will say that Luhmann is an interesting lens through which to view Weird Twitter, which you may recall as a joke-telling subculture of Twitter that was popular before Former President Trump made Twitter much, much weirder. I think there’s some interesting comparisons to be drawn between Anthony Cohen’s theory of the symbolic construction of community, complete with symbolic boundary, and Luhmann’s notion of the boundary of a social system. But I digress.
Luhmann hasn’t actually used the word “communication” yet. He instead says “social contact”. “Every social contact is understood as a system, up to and including society as the inclusion of all possible contacts.” Possible contacts. Meaning that the system is defined in part by its unrealized but potential states. It can be stochastic; it can be changing its internal states to adapt to the external environment. “In other words, the general theory of social systems claims to encompass all sociology’s potential topics and, in this sense, to be a universal sociological theory.” Universal sociological theories are terribly unpopular these days. But Luhmann attempted it. Did he succeed?
“Yet, a claim to universality is not a claim to exclusive correctness, to the exclusive validity, and thus necessity (noncontingency), of one’s own account.” Nobody claiming to have a universal theory does this. Indeed, a theory learns about its own contingency through self-reference. So, social systems theory discovers it European origins, for example, as soon as it considers itself. What then? At that point, one “distinguish[es] between claims of universality and claims to exclusivity”, which makes utter sense, or “by recognizing that structural contingencies must be employed as an operative necessity, with the consequence that there is a constant contingency absorbtion through the successes, practices, and commitments in the scientific system.”
Contingency absorbtion is a nice idea. It is perhaps associated with the idea of abstraction: as one accumulates contingent experiences and abstracts from them, one discovers necessary generalities which are true for all contingent experiences. This has been the core German philosophical method for centuries, and it is quite powerful. We seem to have completely forgotten it in the American academic system. That is why the computer scientists have taken over everything. They have a better universalizing science than the sociologists do. Precisely for that reason, we are seeing computational systems in constant and irksome friction with society. American sociologists need to stop insisting on theoretical pluralism and start developing a universal sociology that is competitive, in terms of its universality, with computer science, or else we will never get smart infrastructure and AI ethics right.
Luhmann, N. (1995). Social systems. Stanford University Press.