We think in teleological terms all the time. We can't help it.
It's obvious why this is; the ability to model a future state of the world and plan out actions for the purpose of bringing it about is a massive survival advantage, as is the ability to model observed events in the context of the possible plans of others. The bias that this introduces into both thought and language is overwhelming; it is very difficult to write without teleological aspects creeping in, and usually we don't even bother to try. Hence statements like “the heart is for pumping blood” seem natural and pass without comment, whereas “the action of the heart pumps blood” seems contrived, passive.
But the future does not cause the past. When I walk down to the shop to buy milk, my actions are not caused by future-me’s possession of milk, but by present-me’s mental model of future-me with milk. This can be most easily seen if some event intervenes; if the shop is unexpectedly closed, there is no future-me with milk, but my actions were no less purposeful as a result.
Similarly, the development of my heart was not caused by a ‘purpose’ of pumping my blood, but rather by the developmental program in my genes that resulted in my ancestors having hearts that pumped their blood.
In the case of the milk, the term ‘purpose’ reflects a state of mind; in the case of the heart, there is no legitimate ground to use the term at all.
False vs. vacuous concepts
A point that Boyden makes in the review of The Last Superstition linked on the references page is that concepts like Aristotelian forms are ambiguous enough that they can be understood as either a true but vacuous concept, or as a false one.
Aristotelian apologists will sometimes respond to criticism of final causes by pointing to physical tendencies, such as objects falling under gravity. This is a vacuous concept because it adds nothing to our understanding of the situation: once we know the laws of motion and gravitation, we can predict where a falling object will end up with more reliability than any argument about final causes could provide. To use Aristotelian terminology, the final cause in this sense is nothing more than the sum of all the efficient causes in operation.
As an example, consider the case of an object thrown upwards. The Aristotelian has no way to make any prediction other than “the object will eventually fall back down because it tends towards its proper place”, whereas the Newtonian can say “if I impart more kinetic energy to the object than its gravitational potential, it will never return”. Thus the Newtonian, despite never having actually launched anything into space, can predict the existence of a qualitative difference in final destination in a way that the Aristotelian cannot.
Ultimately, then, if we start with ‘final cause’ as a mere physical tendency or even an attractor in a physical system, then any attempt to assign greater significance or additional properties to it will tend to make our predictions less reliable; as good a definition as any of a false concept. So far better to just junk the concept entirely, both the vacuous version and the false version.
(Equivocating between the vacuous and the false version is an example of a motte-and-bailey argument.)
Evolution and Development
The development of organisms looks so obviously directed towards an end that it's very hard to untangle the erroneous teleological thinking.
A starting point is to establish that DNA is not a ‘blueprint’, but more like an instruction book, and that therefore the form of the adult organism is not some pre-existing target that it develops towards, but rather an outcome that it reaches as a result of following the instructions—as long as nothing too serious goes wrong.
Again, we see the difference in the event that something does go wrong. The result of mutations in DNA or unusual environmental impacts during development are predictable only in terms of the processes followed, not in terms of the destination.
Natural selection ensures that the developmental program is reasonably robust, since that's a survival benefit; this creates the strong illusion of development towards a specific target. But it's still only an illusion.
Compare the analogy of following a map to a designated place, with following a set of driving directions. If you have a map, and something goes wrong with your route, you can figure out a new one. If all you have are directions with no idea of the destination, you're probably not going to be able to do more than guess in the event of a problem. But following the map requires that you have access to a planner—a computational device, whether part of your brain or implemented in other hardware, that can convert the map and destination information into a new list of directions.
Our brains have little trouble with this. But that makes it seem easier than it is; when one tries to write planning algorithms, it turns out to be harder than one typically expects. When we look at things happening in the outside world, however, we naturally tend to interpret them by (possibly unconsciously) putting ourselves into the action and asking our brain “what is happening here” — which means we tend to get the answer that our brain thinks is easy, not the answer that really is easy.
Convergence and Contingency in Evolution
We talk about evolution in terms of ‘fitness landscapes’, and just as there may be tendencies for water to accumulate at some points of an actual landscape rather than others, so there are tendencies for organisms to accumulate at some regions of relatively optimal fitness of some characteristic. An obvious example is free-swimming creatures evolving a hydrodynamic shape.
But the directions in which a particular lineage of organisms can evolve are constrained by its past. Whales can't evolve the ability to breathe water, regardless of how useful it would be to them; on the other hand, they did evolve the ability to echolocate, which I rather suspect could not have evolved without spending significant evolutionary time out of the water.
So we see again that attractors in the fitness landscape do not ‘cause’ or ‘explain’ anything other than vacuously. Only the knowledge of how evolutionary processes actually work can tell us—given sufficient knowledge of the past history and likely future selection pressures—whether such an attractor could possibly even be reached by descendents of a given species.(to be continued, maybe)
Three Fallacies of Teleology (Less Wrong)