Bayesian Minds?

[luckyme]

How would you describe the decision making we've evolved to have. We meet one-of-a-kind situations constantly and we typically use background probabilities and adjust for current variables to resolve a lot of them. It has a bayesian look to it ( we don't assume 'equally valid' solutions) and we are reasonably good at it when the situation is 'typical' or perhaps 'experienced many times over our evolutionary history'. It seems to work terribly in more unusual situations. Are we using some cookie-cutter approach in those cases or bayesian misfiring?

I suspect that's because we're doing it in non-verbal and non-mathematical formats. That could explain the difference in performance when the same problem is posed with people involved rather than more unusual and less concrete variables, for example.

Dennett describes a Multiple Draft model, is it some bayesian-formed "most likely" that is the final draft?

Any decent layman reading on the topic to suggest?

luckyme

[bigmonkey]

It makes sense that we are all utility-optimizers, even if there are examples that supposedly show this is not true. Apparently there are recorded cases of people preferring A to B and B to C and then C to A (although I haven't seen any), but I think we can explain these in terms of people having bad belief-forming methods resulting in contradictory beliefs, or perhaps xhanging their beliefs frequently. I think that if we allow ourselves to say that some agents are rational and some aren't, that creates ethical problems. I would like to define utility as being whatever an agent intends to cause, meaning that all agents are optimizing utility just by acting.

I just read about the Multiple Draft model. It sounds like he is suggesting that each time we make a draft interpretation of reality from our perception of it we are conditionalizing on our past evidence (memory) and also new incoming evidence. Perhaps we have so much information that it is impossible to analyse all of it at any one time and our judgments are like a process. Perhaps when we are given new evidence X we have to conjoin the information with what we previously know and adjust whichever contadicting beliefs have the least plausibility, and that we can only add proportions of our pre-existing information (say 20%) to the new information each time, meaning that we would make 5 drafts until we make the best judgment, which is then being continually updated by new evidence.

I'm not sure what kind of reading you want. I know more about the normative guides to rationality than philosophy of mind. Have a look at this work by Michael Strevens to see if there's anything useful to you.

http://www.nyu.edu/gsas/dept/philo/u...Conf06/BCT.pdf

[luckyme]

Thanks for the link, I'm looking at it now.
To clarify my interest, here is a post of Jason1990 in DS's 'mathematicians please" thread -

[ QUOTE ]
But how should I fill in the unknown elements? Clearly, this is up to me to do subjectively and it is exactly what you are asking me to do. But it is so much easier for me to just tell you how much I would be willing to bet, rather than to try to cook up an infinite sequence of alternate imaginary universes.

[/ QUOTE ]

What process is going on in our minds to accept or reject certain odds on the bet? Even if we just sample various ones until one 'feels' right, "4-1? nah, 9-2 and it's a bet." it's not likely because 9 to 2 rhymes with scooby-doo, although there may be a bit of that type of thing going on. Rather how we don't like Betsy when first introduced because our mean aunts name was Betsy.

Even if we do it as a Rainman, "531 matchsticks" there still is some process going on somewhere. What is being processed and how is it being processed. Is there enough research so when asked the chances of life on Mars and we answer ".01% since I'd accept 10,000 to one as a fair bet" we know it comes from a different process than "I like sugar". or not?

Jason didn't expand other than to say the multiple universe method doesn't appeal to him, and my question is a tangent off that thread.

luckyme

[bigmonkey]

It could be said that we quantify over possible worlds when we accept or reject bets. Proponents of classical probability, like Carnap, argued that there is a logical relation between the evidence and the hypothesis, and that the probability of a hypothesis given the evidence for it is an objective fact, and part of "logical space".

The idea of possible worlds is consistent with determinism. If we knew every single fact about the actual world then we would never accept bets on anything apart from what we knew was guaranteed to happen. We would know every statement about the future to have either proability 0 or 1.

Because we only know limited facts about the actual world, we don't know which of the possible worlds is the one we inhabit. Obviously I can see we don't live in the world where Kerry is president. But I can't see whether we live in the world where Fidel Castro is alive right now because as far as my information tells me he could be dead. To work out what my subjective probability for Castro being alive is I quantify over all of the worlds my information tells me are possible, and work out the proportion of those worlds he is alive in compared to the proportion he is either alive or dead in. It's a bit hard to conceive of how we would make this calculation because there seems to be infinite possible worlds in which Castro is alive, and a greater degree of infinite worlds in which he is dead or alive.

I don't know anything about what part of the brain entertains these thoughts of conditionalization upon evidence. It must be the same area that entertains counterfactual reasoning, where we ask questions like "If Kerry had become president would the world be such that I got promoted in the year after his election?" In that situation we feed ourselves false information just to see what the world would be like had that false information been true.

All of this talk of possible worlds and conditionalization is philosophical talk and not observed as an actual way we think by neuroscientists or even psychologists. I don't know if the necessary work in those fields has been done or not. But let us say that calulating probabilities is done by the X fibre in the brain doing process Z. We would say that X fibre doing process Z supervenes on conditionalization. One is using talk about a physical brain, and one is talk about ideas (which aren't physical). Equally an elitist physicist might come along and say "well there's no real thing as X fibre or process Z. What's really happening is this bunch of molcules is reacting with this other bunch of molecules with some energy going over here and some reaction over there...blah, blah..". He would just be saying that the brain molecules supervene on the X fibre and the X fibre supervenes on the rational thought-processes. All of these subjects and theories are corect and are just different ways of saying the same thing.