@dbarack @NicoleCRust @ekmiller

Hi David -- interesting! I always read your paper with Krakauer as saying that the our explananda for cognition are our internal representations and the transformations we perform over them (a mental level description), with substrate-independent electrical activity as the most fruitful explanans on a neural level (rather than circuits).

But here I think you're saying that the explanatory target is electrical activity itself, which normally sits at the neural level. Could you clarify? Or just tell me to wait for your book chapter ;-)

@sandervanbree @NicoleCRust @ekmiller Great question. I'm a reductionist--in fact, I'm worse: I'm an identity theorist. I think that mental representations will be type identical to physical types. What types? Those are the physical-substrate-independent electrical activities. So the mental level description = a physical level description (despite being substrate-independent! You have to target the right physical level, and you still get substrate independence). And, the explanatory target (the explanandum) is the cognitive phenomena, not the electrical activity (which is the explanans, what does the explaining), but I take it that that is what you meant.

Now, explanation might be intensional. I can explain that the cat is on the mat without explaining that Granny's favorite critter is on the mat (to provide a Fodor-inspired example). Note though that this is intensionality for the explanandum. There is also intensionality for the explanans. I can explain that the cat is on the mat (the explanandum) because the door was left open, or because Cousin went out to play. These are also distinct. In the first case, the explananda are not the same, and explanans cited to explain the one may not explain the other. In the second case, the explanans are not the same, and they might explain different explananda. (1/2)

@sandervanbree @NicoleCRust @ekmiller
So, to be annoyingly philosophical, it depends on what you mean by "our explananda for cognition are our internal representations". The explananda for cognition is intelligent behavior (or near enough). The explanans are physical things--those are the internal representations (namely, spaces of electrical activity). Applying the foregoing, the electrical activity and the mental representations are type identical explanans that explain the explanandum of intelligent behavior, but they might explain different explananda as well.

I think this is actually important. A lot of explanans (such as neural electrical activity) are type identical to other explanans (such as mental representations) but simply aren't suited to some explanations that their identical counterparts are suited to. Referring to biological neural electrical activity qua biological activity won't help explain the cognitive phenomena evident in artificial neural networks, for example, even if referring to that biological neural activity qua mental representations might just do.

Also, because of the intensionality of explanation, you can give an explanation of spaces of electrical activity without giving an explanation of mental representations (or vice versa). But note, the explanans for cognition are our internal representations, whether neural or mental; when the explananda are those representations, the explanatory target has changed. (2/2)

@dbarack @NicoleCRust @ekmiller

Thank you for taking the time to write out this wealth of information. I think I get what you're saying: intelligent behavior is the explanandum, and substrate-independent e-fields are the explanans, and so too are internal representations because they are actually type-identical to those fields. But their distinction is nevertheless important because in other explanatory contexts it might not be the case that the e-fields and internal representations can be exchanged for each other, which we see for instance if we move to the explanatory context of ANN behavior. Clickidy?

Another question that interests me -- and I know it interests you too -- is whether manifolds are real objects. It was either a talk or a tweet where you made a passing comment that you think they are, and it seems not too far off from the point you're making today. I always think of manifolds as good old descriptions only -- which of course may well figure in scientific explanations. Will questions of realism be covered in your book also? I hope it will.