Another key feature of [edit] group rationality is the ability to not be swayed by what the social group thinks.
There are simple experiments (though I cannot think of the relevant keywords) where a test subject is put in a room full of confederates, all of whom estimate one line segment to be longer than another when the two lines are in fact the same length.
EDIT: Conforming to the group opinion (on average) increases the probability that you are right, thus improving individual truth-tracking. But adding more conformers to the LW community just screws it over. In the limit of infinitely many perfect conformers, the community would display irreversible belief hysteresis; as soon as >50% of the group believed X, all the conformers would switch to believing X, and they would stay that way.
EDIT: The Asch experiments found that only 25% of people would consistently report the truth their own eyes were telling them. Thus most people don’t make good group rationalists.
This comment thread makes clear the need to distinguish between creating good individual rationalists and good group rationalists. Optimizing for group rationality and individual rationality are different tasks.
A gold-plated ignorance prior will be awarded to the first commenter who finds the link for me…
You are the proud recipient of a gold-plated uniform distribution on a finite set. Congrats.
Since my comment has been downvoted to 0, I assume that the LW community likes people who go along with the group opinion even when they know it is wrong? Perhaps people are unsatisfied with this as a rationality test because they think that the test should focus on getting as close to the truth as possible (in which case conforming is good in most cases for most people) rather than adding value to a rationalist community (in which case conforming just because everyone else does is actively hurting the community).
Also, having skimmed the pace.edu link, I am unconvinced that Asch’s results are being misinterpreted, at least by me. Asch found that, in the situation of overwhelming evidence, only 25% of subjects could be trusted to consistently call things the way they really were, i.e. 25% of the subjects pass what I would call the absolute minimum standard of rationality over social conformity.
Note that Carl’s link to the OB article gives us a more nuanced version of this debate, which I recommend.
“Paul Crowley reminds me to note that when subjects can respond in a way that will not be seen by the group, conformity also drops, which also argues against an Aumann interpretation.”
yeah, OK, it’s only 1 person’s opinion, I’ll wait and see what happens when more time passes and more people get the chance to vote.
In defense of my interpretation… few comments get downvoted to zero, so even a small amount of time at zero is fairly significant evidence that people don’t like what you’re saying.
Another key feature of [edit] group rationality is the ability to not be swayed by what the social group thinks.
There are simple experiments (though I cannot think of the relevant keywords) where a test subject is put in a room full of confederates, all of whom estimate one line segment to be longer than another when the two lines are in fact the same length.
EDIT: Conforming to the group opinion (on average) increases the probability that you are right, thus improving individual truth-tracking. But adding more conformers to the LW community just screws it over. In the limit of infinitely many perfect conformers, the community would display irreversible belief hysteresis; as soon as >50% of the group believed X, all the conformers would switch to believing X, and they would stay that way.
EDIT: The Asch experiments found that only 25% of people would consistently report the truth their own eyes were telling them. Thus most people don’t make good group rationalists.
This comment thread makes clear the need to distinguish between creating good individual rationalists and good group rationalists. Optimizing for group rationality and individual rationality are different tasks.
A gold-plated ignorance prior will be awarded to the first commenter who finds the link for me…
You’re thinking of Asch’s experiments. Apparently, they are widely misrepresented: http://webpage.pace.edu/yrafferty/Yvonne/AschConformityStudy.pdf See also: http://www.hss.caltech.edu/~jkg/Conformity.pdf (I don’t remember where I found these… possibly through OB)
You are the proud recipient of a gold-plated uniform distribution on a finite set. Congrats.
Since my comment has been downvoted to 0, I assume that the LW community likes people who go along with the group opinion even when they know it is wrong? Perhaps people are unsatisfied with this as a rationality test because they think that the test should focus on getting as close to the truth as possible (in which case conforming is good in most cases for most people) rather than adding value to a rationalist community (in which case conforming just because everyone else does is actively hurting the community).
Also, having skimmed the pace.edu link, I am unconvinced that Asch’s results are being misinterpreted, at least by me. Asch found that, in the situation of overwhelming evidence, only 25% of subjects could be trusted to consistently call things the way they really were, i.e. 25% of the subjects pass what I would call the absolute minimum standard of rationality over social conformity.
Note that Carl’s link to the OB article gives us a more nuanced version of this debate, which I recommend.
“Paul Crowley reminds me to note that when subjects can respond in a way that will not be seen by the group, conformity also drops, which also argues against an Aumann interpretation.”
Hasty generalization/Belief in the law of small numbers
yeah, OK, it’s only 1 person’s opinion, I’ll wait and see what happens when more time passes and more people get the chance to vote.
In defense of my interpretation… few comments get downvoted to zero, so even a small amount of time at zero is fairly significant evidence that people don’t like what you’re saying.
...and here’s the video (the one in the OB link is dead).
http://www.overcomingbias.com/2007/12/aschs-conformit.html