We can’t excuse our disagreement by claiming that it would be impossible or unreasonable since we would have to dissolve our sense of self.

We can’t excuse our disagreement by claiming that it would be impossible or unreasonable since we would have to dissolve our sense of self.

The second sentence does not necessarily follow from the first If you have enough data so that your prior has a particularly sharp shape, then you’re not going to end up changing your mind much simply from receiving one more piece of information, so long as you continue to also encounter information which reinforces your existing prior. Bayes’ Theorem itself predicts that. And of course the paper you link to discusses the question of whether multiple possible priors from the same initial evidence are rational, another important point here.

Indeed, there is a decent amount of evidence that people do react in at least somewhat a Bayesian way.

It is important to distinguish between a Bayesian (or a few Bayesians) in our society and a society full of Bayesians. In the latter case, I think that our concept of “belief” would cease to exist.

Ah, but what about a society where everyone is a Bayesian except about a few matters about which he or she feels qualified to have an independent judgement? Or, to put it another way, a society where being an “expert” or professional in a field is essentially synonymous with not performing Bayesian updates? Actually, what’s more accurate, though, is that simply being an expert means having a sharper prior distribution, due to having more prior data. Thus updates occur much more slowly.

I disagree with this statement from the paper:
“Mary’s best public estimate of John’s next estimate must instead equal Mary’s current best estimate.”

Unlikely, given how people argue. In the experiment as set up, it is extremely likely that, after John first gives his estimate of the car’s age, Mary will initially offer not her current best estimate, but her original estimate of the car’s age before hearing John’s next estimate. Not only does this seem intrinsically fairer, but each can then offer a subsequent estimate that will give the other an estimation of the initial confidence in the original first guess, by seeing how far the new best estimate moved. This is naturally because the two respondents do not have identical priors; one or the other may have more experience or better judgement in estimating the age of a car.

The second sentence does not necessarily follow from the first If you have enough data so that your prior has a particularly sharp shape, then you’re not going to end up changing your mind much simply from receiving one more piece of information, so long as you continue to also encounter information which reinforces your existing prior. Bayes’ Theorem itself predicts that. And of course the paper you link to discusses the question of whether multiple possible priors from the same initial evidence are rational, another important point here.

Indeed, there is a decent amount of evidence that people do react in at least somewhat a Bayesian way.

It is important to distinguish between a Bayesian (or a few Bayesians) in our society and a society full of Bayesians. In the latter case, I think that our concept of “belief” would cease to exist.

Ah, but what about a society where everyone is a Bayesian except about a few matters about which he or she feels qualified to have an independent judgement? Or, to put it another way, a society where being an “expert” or professional in a field is essentially synonymous with not performing Bayesian updates? Actually, what’s more accurate, though, is that simply being an expert means having a sharper prior distribution, due to having more prior data. Thus updates occur much more slowly.

I disagree with this statement from the paper:
“Mary’s best public estimate of John’s next estimate must instead equal Mary’s current best estimate.”

Unlikely, given how people argue. In the experiment as set up, it is extremely likely that, after John first gives his estimate of the car’s age, Mary will initially offer not her current best estimate, but her original estimate of the car’s age before hearing John’s next estimate. Not only does this seem intrinsically fairer, but each can then offer a subsequent estimate that will give the other an estimation of the initial confidence in the original first guess, by seeing how far the new best estimate moved. This is naturally because the two respondents do not have identical priors; one or the other may have more experience or better judgement in estimating the age of a car.

I’m not sure that I agree with you, though, about how often people would change their minds if they were Bayesians. To take the example that you give, once you knew that X% of the population were Democrats and Y% were Republicans, and you knew something about the demographics of the two parties (how intelligent, how well-informed, etc.), there wouldn’t be much that you could learn just from talking to someone. You might learn a little from good arguments, with arguments about other people’s credibility as judges of the truth (and as judges of credibility, and as judges of meta-credibility, and, …) often playing a larger role than arguments about the facts, but your beliefs wouldn’t be all that much more malleable than they are today. The exception would be when you’re first learning about a new topic and you’re figuring out what positions are out there – a true Bayesian wouldn’t be immediately drawn to certain congenial positions the way that people tend to be in the real world.

You’re right that Bayesianism would involve drastic changes to the self. It is important to distinguish between a Bayesian (or a few Bayesians) in our society and a society full of Bayesians. In the latter case, I think that our concept of “belief” would cease to exist. In the actual world, if the information possessed by humanity implies that proposition A has a 90% chance of being true and that there is a 10% chance of alternative proposition B being true (and A being false), then there will still be some zealous advocates and diligent researchers who believe in B and are committed to making the best possible case for B (and against A). In Bayesian Land, if A vs. B was an important question, we’d probably still want some people to specialize and become B advocates, turning pro-B investigation and advocacy into a part of their identity, but these people (like everyone else) would still say that p(B)=.1 & p(A)=.9. Under one concept of belief (“think true”), these people would believe in A, but in another sense (which is more pragmatic and more closely related to the self) they would be B-believers.

Javier, when people are one way (e.g. shy) and they develop a narrative that says that they are a different kind of person (e.g. outgoing), there’s evidence that some of their behaviors will tend to become consistent with their narrative (e.g. they’ll choose to go to a museum with others rather than alone) while many aspects of their behavior (especially those involving less deliberation) remain consistent with how they had been (e.g. they won’t spend that high a percentage of their time talking with others). Social psychologist Timothy Wilson has a lot to say about this in his book Strangers to Ourselves, which probably covers a lot of the same ground as Gazzaniga’s book, thoughWilson thinks that the unconscious system is a part of the self along with the “interpreter”.

I’m not sure that I agree with you, though, about how often people would change their minds if they were Bayesians. To take the example that you give, once you knew that X% of the population were Democrats and Y% were Republicans, and you knew something about the demographics of the two parties (how intelligent, how well-informed, etc.), there wouldn’t be much that you could learn just from talking to someone. You might learn a little from good arguments, with arguments about other people’s credibility as judges of the truth (and as judges of credibility, and as judges of meta-credibility, and, …) often playing a larger role than arguments about the facts, but your beliefs wouldn’t be all that much more malleable than they are today. The exception would be when you’re first learning about a new topic and you’re figuring out what positions are out there – a true Bayesian wouldn’t be immediately drawn to certain congenial positions the way that people tend to be in the real world.

You’re right that Bayesianism would involve drastic changes to the self. It is important to distinguish between a Bayesian (or a few Bayesians) in our society and a society full of Bayesians. In the latter case, I think that our concept of “belief” would cease to exist. In the actual world, if the information possessed by humanity implies that proposition A has a 90% chance of being true and that there is a 10% chance of alternative proposition B being true (and A being false), then there will still be some zealous advocates and diligent researchers who believe in B and are committed to making the best possible case for B (and against A). In Bayesian Land, if A vs. B was an important question, we’d probably still want some people to specialize and become B advocates, turning pro-B investigation and advocacy into a part of their identity, but these people (like everyone else) would still say that p(B)=.1 & p(A)=.9. Under one concept of belief (“think true”), these people would believe in A, but in another sense (which is more pragmatic and more closely related to the self) they would be B-believers.

Javier, when people are one way (e.g. shy) and they develop a narrative that says that they are a different kind of person (e.g. outgoing), there’s evidence that some of their behaviors will tend to become consistent with their narrative (e.g. they’ll choose to go to a museum with others rather than alone) while many aspects of their behavior (especially those involving less deliberation) remain consistent with how they had been (e.g. they won’t spend that high a percentage of their time talking with others). Social psychologist Timothy Wilson has a lot to say about this in his book Strangers to Ourselves, which probably covers a lot of the same ground as Gazzaniga’s book, thoughWilson thinks that the unconscious system is a part of the self along with the “interpreter”.

(e) capital markets allocate resources efficiently,
(f) marriage is the natural way to organize human families, or
(g) people are naturally egocentric (or benevolent)

is IN ITSELF a common form of self-deception.

Happily, I think that we can act and reflect quite well without affirming or denying these kind of propositions. Indeed, the idea that we need to adhere to them (or their negations) in order to construct our selves is a very destructive myth. Trying to get people to come around to your views on these kinds of propositions sows discord, produces factions, and distracts people from the real practical problems at hand.

Not that I don’t think ideologies are interesting, but
we have to see them for what they are.

(e) capital markets allocate resources efficiently,
(f) marriage is the natural way to organize human families, or
(g) people are naturally egocentric (or benevolent)

is IN ITSELF a common form of self-deception.

Happily, I think that we can act and reflect quite well without affirming or denying these kind of propositions. Indeed, the idea that we need to adhere to them (or their negations) in order to construct our selves is a very destructive myth. Trying to get people to come around to your views on these kinds of propositions sows discord, produces factions, and distracts people from the real practical problems at hand.

Not that I don’t think ideologies are interesting, but
we have to see them for what they are.