The word "agent" can be applied at a range of scales. You might use it to describe individual biological cells, human beings (made up of cells), or corporations (made up of human beings). If you think there might be a theory of agency that can apply at all these scales, you might describe that theory as hierarchical or scale-free. This blog post is not directly about those approaches, which are primarily about searching for formal methods which apply across those scales, but it uses the word "agent" in the same broad scope that they do. The paradigm case of cells->humans->corporations is also lifted directly from that work. At face value this departs from the way the word is used in 20th century analytic philosophy - where to be an agent usually assumes the ability not just to act, but to act intentionally1. One way to get around this is through application of Daniel Dennett's concept of an intentional stance.
At some level of complexity, agents become smart enough to model themselves2. For simple kinds of agent this is of no concern. Cells do not care that they are cells. Furthermore, if our best theories of cell behaviour do not correspond to the basic natural processes which govern cell behaviour, cells will still behave according to the basic natural processes and we can update our mistaken theories accordingly.
Humans are not like this. The writings of Michel Foucault are full of examples where human behaviour at a particular time in history ends up being at least in part explained by the discourse about human behaviour at that time. His examples include education, criminality, sexual nonconformity, and psychiatry. In each case, the creation of a certain discipline for the study of human behaviour results in the imperfect sorting of people into categories (e.g., the delinquent, the slow learner, the sexual deviant). Once the categories have been created, a series of social forces begin to act which result in people coming to understand themselves through these categories. This results in a reflexive effect where the subject and the discipline are each creating one another3.
These kinds of effect make the study of human behaviour especially complex, how are we to come up with explanations of human behaviour in a way which accounts for the reflexivity between those explanations and human behaviour itself?4
One rare approach which I think makes some progress on this question is Ian Hacking's Dynamic Nominalism. It is an approach which tries to explain how:
a causal understanding, if known by those who are understood, can change their character, can change the kind of people that they are. That can lead to a change in the causal understanding itself.
I am looking forward to writing more posts about what I learn.
-
My speculative view is that Donald Davidson's account of agency probably turns out to be one that applies at approximately the levels complex enough for Foucauldian power dynamics to take place. I think this is especially due to the role of the principle of charity, and its requirement that agents are able to model one another as rational. ↩
-
An open question on my mind is whether this is also the level of complexity at which agents become smart enough to model superagents they are a part of. ↩
-
One thing I haven't seen explicitly mentioned in Foucault's writing, but which is also on my mind, is that this effect could take place even in the absence of an explicit discourse about the categorisations. In some cases agents might only need to be aware of the possibility of certain categorisations in order to be affected by them. However, I accept that the presence of an active discourse between agents makes the strength of this effect much greater. ↩
-
As large language models become increasingly situationally aware, answering such questions is also essential for the study of their behaviour. ↩