...that an Obama supporter acting violently was in some sense evidence against Obama or justification for opposition to Obama; or, that a McCain supporter acting dishonestly was in some sense evidence against McCain or confirmation that Obama supporters were better people. To a Bayesian, this would be balderdash.
Uh, to a Bayesian this would be relevant evidence. Or do I misunderstand something?
A single Hitler supporter acting violently isn’t much evidence against Hitler. Thousands of apparently sane individuals committing horrors is pretty damning though.
I haven’t done the math, but I would have thought that a hundred incidents would be more than a hundred times as much evidence as one, because it says that it’s not just the unsurprising lunatic fringe of your supporters who are up for violence.
I don’t think that’s possible, unless the first incident makes it conditionally less likely that the second incident will occur unless Hitler is ungood.
Unless you mean, “the total information that a sum of one incident has occurred is less than a hundredth the evidence than the total information that a sum of a hundred incidents have occurred”, in which case I agree, because in the former case you’re also getting the information on all the people who didn’t commit violent acts.
unless the first incident makes it conditionally less likely that the second incident will occur unless Hitler is ungood.
That wasn’t what I had in mind (and what I did have in mind is pretty straightforward to express and test mathematically, so I’ll do that later today) but it’s a possibility worth taking seriously: are you the sort of organisation that responds to reports of violence with a memo saying “don’t go carving a backwards B on people”?
Assuming the prior probability of politically-motivated violent incidents to be greater than zero, X incidents where X/(number of supporters) is roughly equal to the incidence for the entire population offers very little evidence of anything, so X*100 is trivially more than a hundred times the evidence.
I guess the question being asked here is whether those Hitler supporters acting so violently should affect your decision on whether to support Hitler or not. Rationally speaking, it should not, because his supporters and the man himself are two separate things, but the initial response will likely be to assign both things to the same category and have both be affected by the negative perception of the supporters.
I think if you use examples that are less confrontational or biased you can get the message across better. Hitler is usually not a useful subject for examples or comparisons.
To a Bayesian, all evidence is relevant. These two pieces of evidence would seem to have very low weights though. Do you think the weights would be significant?
Assigning significant weight to this event (on either side) is likely a combination of sensationalist national mass media and availability heuristic bias.
Uncoordinated behavior of individual followers reflects very weakly on organizations or their leaders. Without any indication of wider trends or mass behavior, the evidence would be weighted so little as to demand disregarding by a bounded rationalist.
I see you’ve been upvoted anyways so I’m likely not the only one, but I want to personally thank you for this. People being more willing to admit that they made a mistake and carry on is an excellent feature of Less Wrong and extremely rare in most online communities.
I disagree that it is extremely rare. I’ve seen a good number of apologies reading reddit, and I think it might be bad to upvote them because it could lead to the motives of any apologizer becoming suspect.
Uh, to a Bayesian this would be relevant evidence. Or do I misunderstand something?
A Hitler supporter acting violently is evidence against Hitler. But it takes a lot of them to reach significance.
A single Hitler supporter acting violently isn’t much evidence against Hitler. Thousands of apparently sane individuals committing horrors is pretty damning though.
I haven’t done the math, but I would have thought that a hundred incidents would be more than a hundred times as much evidence as one, because it says that it’s not just the unsurprising lunatic fringe of your supporters who are up for violence.
I don’t think that’s possible, unless the first incident makes it conditionally less likely that the second incident will occur unless Hitler is ungood.
Unless you mean, “the total information that a sum of one incident has occurred is less than a hundredth the evidence than the total information that a sum of a hundred incidents have occurred”, in which case I agree, because in the former case you’re also getting the information on all the people who didn’t commit violent acts.
That wasn’t what I had in mind (and what I did have in mind is pretty straightforward to express and test mathematically, so I’ll do that later today) but it’s a possibility worth taking seriously: are you the sort of organisation that responds to reports of violence with a memo saying “don’t go carving a backwards B on people”?
Assuming the prior probability of politically-motivated violent incidents to be greater than zero, X incidents where X/(number of supporters) is roughly equal to the incidence for the entire population offers very little evidence of anything, so X*100 is trivially more than a hundred times the evidence.
I guess the question being asked here is whether those Hitler supporters acting so violently should affect your decision on whether to support Hitler or not. Rationally speaking, it should not, because his supporters and the man himself are two separate things, but the initial response will likely be to assign both things to the same category and have both be affected by the negative perception of the supporters.
I think if you use examples that are less confrontational or biased you can get the message across better. Hitler is usually not a useful subject for examples or comparisons.
To a Bayesian, all evidence is relevant. These two pieces of evidence would seem to have very low weights though. Do you think the weights would be significant?
If I were a McCain supporter, the rumor’s turning out to be false would’ve carried significant weight for me. You?
Assigning significant weight to this event (on either side) is likely a combination of sensationalist national mass media and availability heuristic bias.
Uncoordinated behavior of individual followers reflects very weakly on organizations or their leaders. Without any indication of wider trends or mass behavior, the evidence would be weighted so little as to demand disregarding by a bounded rationalist.
Yes, I was being dumb. Sorry.
Edit: stop with the upvotes already!
I see you’ve been upvoted anyways so I’m likely not the only one, but I want to personally thank you for this. People being more willing to admit that they made a mistake and carry on is an excellent feature of Less Wrong and extremely rare in most online communities.
I disagree that it is extremely rare. I’ve seen a good number of apologies reading reddit, and I think it might be bad to upvote them because it could lead to the motives of any apologizer becoming suspect.
Voted up because it asked not to be upvoted.
Hey, that’s my line.
Voted randomly because it references a vote cast on the basis of vote-reference.
I’ll probably get downvoted for this, but please don’t upvote this comment.
EDIT: OK, looks like that wasn’t as funny as I thought, lesson learned!
I left the parent alone.