Sunday, November 11, 2012

"What Darwin Got Wrong" Got Wrong

(I've already written about this.)

I really wanted to like "What Darwin Got Wrong." I am skeptical of many commonly-heard versions of natural selection, especially regarding evolutionary psychology, and Fodor is a giant in his field. However, the book is deeply flawed. The critics may be right on this one.

To recap one of the major arguments in the book, it makes the distinction between frogs "snapping at flies" and "snapping at Ambient Black Nuisances (ABNs)." The two concepts are coextensive, i.e., overlap - all flies are ABNs and vice versa, so indeed as Fodor and Piattelli-Palmarini say, there can be no causal process that discriminates between them. So we can't say "natural selection selected for" one or the other. But this is misleading.

First, these two concepts are not related in the same way that arches and spandrels are, even though the book assumes they are parallel. Arches and spandrels are not coextensive, they are different things. The property of having arches or having spandrels are coextensive, though. But we can still see a very concrete difference between arches and spandrels in the actual world. We cannot see such a difference between flies and ABNs, there is no physical difference to see.

The concepts of flies and ABNs are different of course. But does that matter? Perhaps all we need is to say that evolution selected for flies or ABNs or in general the intersection of all relevant descriptions (or perhaps say: "ABN is shorthand for fly, currently"). Descriptions matter, but not here - at least until the concepts do diverge in the actual world, for example if an ABN shows up that is not a fly. Then we can reassess.

We can, however, reassess even earlier. A major flaw in the book is the "magical power" it assumes mental processes to have, that nothing else does: intentionality, the ability to consider counterfactuals, etc. But a naturalistic, computational view of cognition avoids the problems with that. So, even in the current world where ABNs and flies are coextensive, we can look at the processes taking place in the frog's head. If we see that the representations there pick out flies very specifically, or if they look just for the properties of an ABN, we can say something about that.

And we can apply the same to the mental processes of the architect designing for arches but not spandrels. Investigation of the causal cognitive processes taking place will show that arches are being designed for, not spandrels. Architect brains are just physical systems, in a naturalistic/computational world view.

In short, we can - in principle - find out what causal processes take place, and see what is being selected for. If it is hard to find that information out, we may never know (I suspect that is the case for at least some of the evolution of human psychology), but there is still a fact of the matter. Minds can take counterfactuals into account, but when they do so, there is a causal process at work.

And it isn't just minds that can take into account counterfactuals. A chess-playing machine takes counterfactuals into account, and we might not want to say it has a mind. It doesn't matter if it does or not - what matters is what computation it does. A computation on counterfactuals can be done in a mind, or in a mindless (sophisticated-enough) machine.

Which brings us to selection in general. Natural selection for Darwin was not the only mechanism of evolution, another crucial one was sexual selection. Sexual selection happens when one organism chooses another to mate with. The best strategy is to pick the mate that increases the fitness of your potential offspring, and a good estimate of that is to consider the potential mate's fitness. This is exactly a case where counterfactuals can happen: When an organism considers who to mate with, it can run scenarios in its mind such as "X's fast running is a useful trait, our children will evade predators more easily" or even "the weather is getting warmer and warmer, Y's shorter fur will be useful to our children to not overheat." In other words, sexual selection does allow for counterfactuals to be taken into account, and it is surprising to me that this point was not made in the book, which did try to anticipate responses to its arguments. Even if natural selection falls (although as I said, I do not believe it does), sexual selection can remain, and it can have effects that mirror those that natural selection can, see the second example before - it can lead to shorter fur being selected for, which an adaptationist view would argue natural selection could account for. And it does so using the minds of the organisms, which even the authors will grant intentionality to.

Of course sexual selection is not special here. Any sufficiently sophisticated system can take into account counterfactuals in its computations, as mentioned before. In principle it could happen not just in brains but in ant colonies or entire ecosystems. (Perhaps we can define "Mother Nature" using that?) But we don't need to go to such speculative lengths, for even in plain old brains the ability to consider counterfactuals can affect not just sexual selection but other kinds of selection as well: for example, similar things can happen when choosing friends/coalition-members/etc. ("I should work together with Z, his superior strength will be useful", and this improves Z's fitness since Z can now choose from more potential collaborators).

Finally, regarding "Sobel's Sieve", where red large balls remain on top while small blue ones are let through the holes, and the question is, was color selected for or size. First, it is clear as the authors admit that size is being selected for: we can see the causal processes in play. But the point is that the causal processes are there, not if we can see them. It might be hard to see them, but that doesn't mean something was not selected for as a matter of fact. Second, the authors claim a fatal flaw in the analogy is that it is not clear what is "better" - to remain on top (prospector filtering out debris) or fall out the bottom (someone trying to filter out large undesirable objects). This supposedly shows that we do need intentionality in order to know what is being selected for. But it does nothing of the sort. The analogy is only half-done: To make it a proper analogy to natural selection, it would need to continue with our taking the ones that remained on top, mixing up their "genes" to create another generation, and repeating the process on that. In that complete analogy, it is clear that the ones on top are selected for, because the ones that fell through were discarded (and if we completed the analogy the other way, the fact of the matter would be vice versa of course). In natural selection, after all, we do see which ones are actually kept and which are discarded.