Evolution, Adaptive Beliefs and Tigers

Natural selection only cares about what's adaptive. Both true and false beliefs can be adaptive as long as they increase an animal's chance of survival. Why should we expect natural selection to favor true adaptive beliefs over false ones? Alvin Plantinga gives the example of Paul, who loves the idea of being eaten by tigers.

Each time he sees a tiger, he runs away thinking that this particular one is unlikely to eat him. So, he wants to find one who will eat him. His false belief that a tiger won't eat him results in adaptive behavior, which is running away from tigers. But his false belief is not as adaptive as the true belief that tigers are dangerous because they're likely to eat you. The best thing to do is to avoid close proximity to tigers in all cases.

To see this is true, we can put Paul in various situations where his options are limited and see what happens. One example is based on something I read on blog.sparkplots.net. Paul is stuck in a cave with a ferocious tiger blocking the exit. There are a number of deep crevasses at the back of the cave, and Paul is hiding in one of them.

If Paul believes that this tiger is unlikely to eat him but would like to be eaten, then he should try and run past the tiger to find a hungrier one. But if he tries, of course, he'll certainly be caught and eaten. Plantinga suggests Paul wouldn't pass on his genes to the next generation if he were stuck in a cave with the tiger.

In order to survive in this situation while still holding these false beliefs, Paul would have to have them supplemented by extra, entirely new false beliefs related specifically to the situation. For instance, a tiger in a cave is sure to eat you, and tigers in caves can very slowly shrink and fit themselves into small cracks.

With certain extra beliefs like these, he'd have the correct motivation to wait in the crevasse and would end up avoiding harm. Actually, a similar thing happens in many specific situations we can think of involving tigers. For Paul to survive in these situations, his false beliefs would have to be supplemented by even more false beliefs, and the new false beliefs would be specific to the situations in which he found himself.

Now consider Paul trapped in the cave, except this time he has true beliefs. If Paul wants to live and holds the true belief that tigers are dangerous, and that the best way to keep safe is to maximize your distance from tigers at all times, then he will stay in the crevice. If he can stay there long enough, the tiger may leave.

He's not certain to survive, but this behavior driven by true beliefs will not lead to certain death. So, his chance of survival is already very much better than the Paul who has false beliefs. Notice that if Paul has true beliefs about tigers, he doesn't need any supplemental beliefs to deal with the specific situation of a tiger in a cave.

A single true adaptive belief is applicable across a very wide range of situations in a way that a single false adaptive belief is not. True beliefs are flexible because they're modular; they're pieced together from other true beliefs. This modularity makes a person with true beliefs much more able to adapt to novel situations than a person with false beliefs.

It's best to keep your distance from tigers because a tiger is a big, strong, meat-eating animal. Big, strong, meat-eating animals can move fast and damage you. If you see a new animal that matches this profile, it's also best to keep your distance from this new animal. In other words, lions are dangerous is not an entirely independent belief; tigers are dangerous.

There's a very large conceptual overlap between the two, and this represents a very efficient way to organize data.

One note before the conclusion: we can understand how a network of mostly true beliefs would grow. We compare new sense data to our existing beliefs, and we modify our existing beliefs to keep the network consistent. But how does a network of mostly false beliefs come to be? How does it develop without harming the survival prospects of the person holding them? It's not clear how this could happen, and I'm not sure that Plantinga has offered any suggestion of how this might come to be.

Adaptive true beliefs have very important advantages that adaptive false beliefs don't have. They're generalizable and can often be applied successfully to a wide range of situations without the need for new modifying beliefs. The way true beliefs are interrelated makes it possible to deal successfully with novel situations.

So, these are very strong reasons to expect natural selection to generally favor true adaptive beliefs over false adaptive beliefs.