Monday, September 21, 2020

Limits of neuroscience

I think that our best physicalist view right now is a functionalism on which mental states are identified with types of computation in a hardware-agnostic way (i.e., whatever the hardware is, as long as the same type of computation is done, the mental states get tokened).


  1. If functionalism is true, what human discipline, if any, will discover which functional processes (e.g., the execution of what algorithms) constitute consciousness?

There are, I think, three plausible answers:

  1. None. We wouldn’t be able to know the answer.

  2. Philosophy and neuroscience working together.

  3. Neuroscience working alone.

I think the most plausible answer is (2), with (3) being a runner up.

In this post I want to give a quick argument that (4) is not the answer.

Neuroscience is a natural science. The natural sciences do not discover substantive facts about worlds whose laws of nature are radically different from ours. Some possible worlds whose laws of nature are radically different from ours contain beings with functional processes isomorphic to the ones running in us. Thus, if neuroscience discovered which functional processes constitute consciousness, it would discover about those worlds that they contain consciousness. That would be a substantive fact about these worlds, and that would contradict the assumption that neuroscience is a natural science.


Ovidiu Badea said...

Would you give some details on why you think 2 is the most plausible answer? Is it a limit inherent to functionalism?

Alexander R Pruss said...

I think the question of how far one is allowed to abstract from the actual algorithms used by the human brain for something to count as computationally isomorphic to our functioning for purposes of philosophy of mind would be too deeply empirically underdetermined to give us a good answer even when we add a lot of philosophy to the mix.

Imagine, for instance, that somewhere in our brain, crucial to the processing of conscious visual perception of shapes, an arithmetical operation of multiplying two numbers is performed. There are infinitely many algorithms by which two numbers can be multiplied together. (E.g., log-add-exp, long multiplication in any of an infinite number of bases.) Does it matter for consciousness which algorithm is used? Since third-party consciousness is only knowable by observation of behavior, and all the algorithms would produce the same behavior, the question of which range (one, some or all) of multiplication algorithms is compatible with visual consciousness is deeply empirically underdetermined. All we can find out empirically is what algorithms our brains use.