Anomalous Functionalism

In a couple of previous blog posts I talked about functionalist accounts of phenomenal consciousness. This blog post also looks at computational functionalism, but is about mental causation and semantic content rather than about phenomenal consciousness. Functionalist accounts have been used for both of these purposes, that is: on the one hand to explain mental phenomena like beliefs, desires, and intentional actions; and on the other hand to either explain¹ (or at least make predictions about) phenomenal consciousness. Some writers take the position that a full account of one will give us a full account of the other, I am skeptical of this, but am still interested in each independently, as I think full and compatible accounts of both are required to present an overall position in the philosophy of mind.

Anomalous monism is the position that even very basic mental phenomena like beliefs, desires and intentional actions cannot be subject to strict laws - even though they are instantiated by the strict laws of physics. Since physics (as conceived by this approach) is subject to strict laws, an implication of this position is that there are no strict psychophysical laws. The way this position remains a monist one is that maintains that mental events are token identical with physical events, but not type identical with physical events.

The natural question is - is there something else that mental events are type identical with? A related problem in the philosophy of mind is the problem of multiple-realizability, that is: that states like "beliefs" can be realised in varieties of organisms with quite different physical properties. One solution to the multiple-realizability problem is to adopt a functionalist position, specifically that there is a type identity between mental states and functional states.

Functionalism, especially construed as a type identity theory, is normally seen as incompatible with anomalist monism. One small matter is extending it from being a theory about mental states to being a theory about mental events. The main challenge is rather the anomalism thesis - as a functionalist account seems like it would establish the existence of strict psychophysical laws after all.

The anomalism thesis doesn't seem to have had many strong defenders come out in its favour since it was first presented in these terms by Donald Davidson. John McDowell was one, Douglas Hofstadter may have been another. The strongest arguments I have been able to find for it rely on the principles that apply to normative beliefs and their semantic content, for example revisability or the principle of charity. I won't try to reconstruct these arguments here, but instead I want entertain what kind of position might hold if both the anomalism thesis and functionalism as a type identity theory were true.

I think if this were this case it would have to imply that functional instantiation is, itself, an anomalous relation. More specifically it would imply that there are not strict laws about how functions are realised in the physical world. This seems to go against the core intuition about what "functions" are supposed to be, however - I think there are some reasons to entertain this view:

1. Computational functions as defined have infinite domains and, and require machines with infinite memory to operate on the whole of these domains, but physical machines are finite.
2. Agentic beings of a certain level of complexity, on account of having beliefs and desires, are able to reference, model and strive for certain idealised behaviours - even though their "hardware" in general cannot realise these behaviours - they can still behave in ways which are locally consistent with them.
3. This might require some degree of self correction, an agent might realise it is not in fact instantiating the function it wants to instantiate, and update its behaviour accordingly.
4. This also explains how, in systems with multiple agents, agents can interact and revise their beliefs to define new behaviours, and more generally new types of agent, along the lines of Ian Hacking's dynamic nominalism.

Daniel Dennett appears to express similar views in Consciousness Explained, placing an especially strong emphasis on a point similar to (4), making use of Richard Dawkins' concept of "memes":

The haven all memes depend on reaching is the human mind, but a human mind is itself an artifact created when memes restructure a human brain in order to make it a better habitat for memes.

Assuming it could be made to work in computationally sound terms, and assuming it responds correctly to real worries raised by defenders of the anomalism thesis, I think I would find this a relatively compelling account. However there remains one strong objection to it: namely, that even we only endorse a "weak" version which does not try to account for phenomenal consciousness, and only for other mental phenomena, the role that an account like this leaves for phenomenal consciousness may be too narrow. If minds are ultimately identical with these idealized mimetic functional structures, the boundaries of our identities should not be easily confined to our skulls, because these structures themselves are not. However experience and introspection strongly indicate that beliefs, desires, actions and so on have a residual and irreducible phenomenal content which does in fact tie them to our nervous systems and sense organs, in a way that challenges this degree of abstraction. Dennett's approach was explain that this content is in fact confused or illusory, I am curious if, instead, it can be made compatible with a non-eliminative account of phenomenal consciousness.