Just an idle quip. But perhaps this could be an interesting research question for Bayesian reasoning. What is the minimum communication necessary for the process of Aumann agreement? Aumann original paper requires that the reasoners have the same priors and common knowledge of their posteriors, without considering the mechanism that gets them there—presumably, they just tell each other. There has been some work on more explicit protocols for them to share their information, some of it referenced here.
How much agreement can a community of perfect Bayesian reasoners reach through the medium of a karma voting system?
You clearly have a weird understanding of Aumann. First of all, why did you bring it up? Because 6 out of 17 people did not agree with you!? People are neither perfect bayesians nor have common priors as you probably know and the agreement theorem is not really applicable here.
Furthermore perfect bayesians are not susceptible to framing effects so it does not matter how they arrive at the information.
Also I do not really see”different protocols for them to share information” on the wiki. Hanson’s idea is basically that bayesians cannot agree to disagree if instead of having common priors they have common information about how they arrive at those priors (because this way they will again end up with common priors if his pre-rationality condition is fulfilled). And I hadn’t seen the other paper but it seems like Hellman is just examining agreement in the case of differing priors.
How much agreement can a community of perfect Bayesian reasoners reach through the medium of a karma voting system?
Perfect Bayesian reasoners are able to reach maximum agreement. The medium does not matter. So I’d say maximum agreement.
You clearly have a weird understanding of Aumann. First of all, why did you bring it up?
Like I said, an idle quip. Perfect Bayesians, perfectly communicating, agree, and there is a frequently expressed attitude on LessWrong that as would-be perfect Bayesians, we should easily come to agreement about everything. Yet here we have an extreme divergence. (The score is now at 11-8. I love the new feature of showing the % positive!)
Also I do not really see”different protocols for them to share information” on the wiki.
Try the Aaronson paper, and the papers it references.
How much agreement can a community of perfect Bayesian reasoners reach through the medium of a karma voting system?
Perfect Bayesian reasoners are able to reach maximum agreement. The medium does not matter. So I’d say maximum agreement.
There must be a medium. The properties of that medium determine what can be communicated and how long it takes.
there is a frequently expressed attitude on LesWrong that as would-be perfect Bayesians, we should easily come to agreement about everything.
Hm, definitely no. I think that most lesswrongers realize that we are really really far from being prefect bayesians and having common priors and also realize how far we are from agreeing. That Helman paper surely argues with me, however I can believe that there are a bunch of lesswrongers deluded into thinking that we are already at the stage when we can easily reach agreements about everything. I am skeptical to them being more than a minority though.
Try the Aaronson paper, and the papers it references.
The Aaronson protocols address the real-world limitations of time and computing power . The mechanism through which perfect bayesians receive information is not important outside of such real-world limitations (as those are limitations for humans and as perfect bayesians exist slightly outside those limitations). The data that they currently posses and are able to obtain is.
There must be a medium. The properties of that medium determine what can be communicated and how long it takes.
I am assuming mediums where they are able to exchange whatever information they want even if they need to send it through binary—mediums such as the one here. And no, it does not matter how long it takes—the same agreement will be reached if it takes a billion years or 4 milliseconds, this is perfect bayesians we are talking about.
Just an idle quip. But perhaps this could be an interesting research question for Bayesian reasoning. What is the minimum communication necessary for the process of Aumann agreement? Aumann original paper requires that the reasoners have the same priors and common knowledge of their posteriors, without considering the mechanism that gets them there—presumably, they just tell each other. There has been some work on more explicit protocols for them to share their information, some of it referenced here.
How much agreement can a community of perfect Bayesian reasoners reach through the medium of a karma voting system?
You clearly have a weird understanding of Aumann. First of all, why did you bring it up? Because 6 out of 17 people did not agree with you!? People are neither perfect bayesians nor have common priors as you probably know and the agreement theorem is not really applicable here.
Furthermore perfect bayesians are not susceptible to framing effects so it does not matter how they arrive at the information.
Also I do not really see”different protocols for them to share information” on the wiki. Hanson’s idea is basically that bayesians cannot agree to disagree if instead of having common priors they have common information about how they arrive at those priors (because this way they will again end up with common priors if his pre-rationality condition is fulfilled). And I hadn’t seen the other paper but it seems like Hellman is just examining agreement in the case of differing priors.
Perfect Bayesian reasoners are able to reach maximum agreement. The medium does not matter. So I’d say maximum agreement.
Like I said, an idle quip. Perfect Bayesians, perfectly communicating, agree, and there is a frequently expressed attitude on LessWrong that as would-be perfect Bayesians, we should easily come to agreement about everything. Yet here we have an extreme divergence. (The score is now at 11-8. I love the new feature of showing the % positive!)
Try the Aaronson paper, and the papers it references.
There must be a medium. The properties of that medium determine what can be communicated and how long it takes.
Hm, definitely no. I think that most lesswrongers realize that we are really really far from being prefect bayesians and having common priors and also realize how far we are from agreeing. That Helman paper surely argues with me, however I can believe that there are a bunch of lesswrongers deluded into thinking that we are already at the stage when we can easily reach agreements about everything. I am skeptical to them being more than a minority though.
The Aaronson protocols address the real-world limitations of time and computing power . The mechanism through which perfect bayesians receive information is not important outside of such real-world limitations (as those are limitations for humans and as perfect bayesians exist slightly outside those limitations). The data that they currently posses and are able to obtain is.
I am assuming mediums where they are able to exchange whatever information they want even if they need to send it through binary—mediums such as the one here. And no, it does not matter how long it takes—the same agreement will be reached if it takes a billion years or 4 milliseconds, this is perfect bayesians we are talking about.