The essential issue seems to be here that your friend is claiming that because humans aren’t perfect Bayesians that Bayesianism is somehow philosophically wrong. Whether human cognition is flawed even severely doesn’t impact whether or not Bayesianism is a better approach. Note that your friend’s argument if it were valid then it would apply not just to Bayesianism but any attempt to use statistics. It is pretty clear that humans pay a lot more attention to anecdotes than actual stats for example. By this argument, statistics themselves should be ignored.
This seems in essence to be an is v. ought fallacy.
One might be tempted to assume that if the human specie evolved as so blatantly non-Bayesian, yet survived and took over the world, then Bayesianism is probably incorrect. Because if it was, then surely any specie that would have evolved Bayesianism would have taken over the world instead of us. If we have this in mind, that should take care of the “be vs ought” fallacy, because what ought to be, would be.
I reject this argument however, mainly because Bayesian calculations are simply intractable. Even when they are, “Yikes! A tiger!!” is way more effective at Darwinism than the more explicit “Yellow, stripes, feline shaped, looking at me, big, danger so let’s -AHRRGH CRUNCH GULP”. And the fair amount of false positives that the emotional quick guess generates probably wasn’t very harmful in the ancestral environment.
Because if it was, then surely any specie that would have evolved Bayesianism would have taken over the world instead of us.
Humans evolved a step towards being capable of Bayesian reasoning and we completely overthrew the natural order in an evolutionary instant. We should not expect to see close approximations of Bayesian thinking (combined with typical goal seeking behaviour) evolve because when a species gets vaguely close it becomes better at optimising than evolution is!
The essential issue seems to be here that your friend is claiming that because humans aren’t perfect Bayesians that Bayesianism is somehow philosophically wrong. Whether human cognition is flawed even severely doesn’t impact whether or not Bayesianism is a better approach. Note that your friend’s argument if it were valid then it would apply not just to Bayesianism but any attempt to use statistics. It is pretty clear that humans pay a lot more attention to anecdotes than actual stats for example. By this argument, statistics themselves should be ignored.
This seems in essence to be an is v. ought fallacy.
One might be tempted to assume that if the human specie evolved as so blatantly non-Bayesian, yet survived and took over the world, then Bayesianism is probably incorrect. Because if it was, then surely any specie that would have evolved Bayesianism would have taken over the world instead of us. If we have this in mind, that should take care of the “be vs ought” fallacy, because what ought to be, would be.
I reject this argument however, mainly because Bayesian calculations are simply intractable. Even when they are, “Yikes! A tiger!!” is way more effective at Darwinism than the more explicit “Yellow, stripes, feline shaped, looking at me, big, danger so let’s -AHRRGH CRUNCH GULP”. And the fair amount of false positives that the emotional quick guess generates probably wasn’t very harmful in the ancestral environment.
Humans evolved a step towards being capable of Bayesian reasoning and we completely overthrew the natural order in an evolutionary instant. We should not expect to see close approximations of Bayesian thinking (combined with typical goal seeking behaviour) evolve because when a species gets vaguely close it becomes better at optimising than evolution is!