Perhaps not directly related to this post: Is anyone interested in looking to use Metaculus or Manifold to predict the future frequency and/or number of occurrences of sexual misconduct within EA?
I am asking as I suspect the EA community might be putting a disproportionate amount of resources into understanding past occurrences with a disproportionally smaller effort into preventing future occurrences. Maybe having a clear-eyed view on what might be in store going forward could go some way to alleviating this.
I haven’t disagree voted this. I’m the #1 trader on Manifold. I doubt prediction markets will be useful here. In fact, I think I could probably use prediction markets to “screw with people”. For example “watch out for X, Manifold thinks there is a 15% chance that they have a sexual assault charge in the next year”.
Yes I wasn’t sure it was a good idea. Anecdotally, I forecasted a question on VC funding to female founders. My forecast was negative and accurate and I was surprised of how well most other forecasters thought things were going for female founders. The process of forecasting also forced me to think through the factors preventing more female founders from getting more funding which I found educational. But perhaps my experience doesn’t translate well. That said I think most negative reactions to my idea was mostly me hijacking this post which I think is fair criticism and some people seemed to think it might be worthwhile.
I agree it would not be good if predictions are made about people—that seems really bad a bit like black mirror!
FWIW there was actually someone making such a prediction for ACE back in the days. It stood at 15% chance which I think is uncomfortably high for chance of a scandal in a single year—it implies an 80% chance of it happening over a 10-year period! They also ran another question on a scandal related to racism.
It stood at 15% chance which I think is uncomfortably high for chance of a scandal in a single year
Note that it wasn’t just ACE, it was ACE or any of it’s (~dozen) top or standout charities.
it implies an 80% chance of it happening over a 10-year period
I don’t think assuming full independence makes sense? I wouldn’t go all the way to “if they have a scandal-likely person/culture then they will definitely have a scandal this year, so 15% is the probability they have a scandal-likely person/culture” but I do think that’s a piece of it. Very roughly I’d guess that 15% in one year gives 50% in ten?
Thanks, those are important points. I just quickly came across it by searching for something else and thought I would share as it seemed very relevant.
Whomever you are who is on a downvoting rampage against me: Really? A −6 vote on all these comments when I just tried to make it easier for people to give me feedback anonymously? I know it is stupid of me to care about karma but that was just plain mean of you.
I didn’t vote on any of the poll comments, but one major thing in my mind is that they should have been posted under the comment saying you’re going to be doing the poll, not under the parent comment
I haven’t voted either way on your comments, but in general making several comments as a way to run a poll is not a good idea unless you’re pretty confident others will find your poll interesting. And then if someone does/doesn’t think the poll belongs here they upvote/downvote each question in the poll, magnifying the karma impact.
Then I expect this comment was downvoted because (a) it’s complaining about downvotes and (b) doesn’t engage with what people should do if they don’t think your poll comments belong on the Forum.
I often downvote enough of the options to zero out the karma gain from posting the options, on the theory that the user shouldn’t reap an automatic +1 or +2 for each option posted. I would not strong-downvote on the basis (if other users had found value in the options being listed, who am I to attempt to overrule them?) and was not the strong-downvoter here.
Fair enough, I will definitely think of better ways to get feedback in the future. It is completely on me that I didn’t realise the karma impact could be magnified like that (or that running polls that way might cost me karma!). And my concerns about karma is insignificant compared to the personal suffering by those affected by the actions of the original post. What’s most important to me is that I can find ways to constructively and meaningfully contribute to reducing the risk of future harms to the community, something I am dedicating more time to than writing here on the forum. Thanks for your patience Jeff!
By doing something to prevent future occurrences, we are risking causing enough trouble so that there is chance people are deterred from engaging with EA and/or important people might leave our community.
Specifically doing something by forecasting future occurrences is likely to be net negative. It could e.g. deter more women from engaging with EA and cause ~panic as the numbers could end up looking pretty high.
But also there seems to be a total lack of upside. Reported sexual misconduct is not the same as actual sexual misconduct (generally it significantly undercounts) so there’s no real indication whether an apparent improvement in the number represents perceptions of improved behaviour or worsened reporting process. Pretty much all the incentives on either side of the bet are perverse (people with relevant knowledge can earn from suppressing information or breaching confidences, and people can earn money from there being lots of sexual harassment or perhaps even buy looser safeguarding policy!), particularly as I can’t imagine it being a liquid market lots of people uninvolved in sexual harassment cases want to bet on.
I think your right that it would be very bad reputationally if the community as a whole was widely perceived as doing this. But it’s also a bit easier to say that if you also believe that this is actually a bad, or at least useless, thing to do on it’s own merits. If you don’t think that, it seems a bit sleazy (even if perhaps correct) to reason ‘this would actually help improve how we deal with sexual misconduct, but we shouldn’t do it because it’ll make us look bad’. Of course, the fact that something looks bad is evidence it IS bad, even if it seem good to you, but not always definitive evidence. Personally I don’t think it would be useful, but I’m not sure how much someone making a Manifold market would actually cause outside people to perceive this as something “EAs” do, or even that they would notice the market at all.
Could you please indicate why you are disagreeing with me? To make it easier, I will make a sub comment each with possible objections so all you have to do is read and click.
I am genuinely keen to understand what, if anything, can be done to reduce the chance of future offences. I should make it clear, in case there was any doubt: I am talking about offences from anyone, and not from anyone in particular. In other words, I have no reason to suspect that Owen in particular is at increased risk from repeat offences. If anything, I would trust the judgement of other commenters indicating such risks are low.
Perhaps not directly related to this post: Is anyone interested in looking to use Metaculus or Manifold to predict the future frequency and/or number of occurrences of sexual misconduct within EA?
I am asking as I suspect the EA community might be putting a disproportionate amount of resources into understanding past occurrences with a disproportionally smaller effort into preventing future occurrences. Maybe having a clear-eyed view on what might be in store going forward could go some way to alleviating this.
I haven’t disagree voted this. I’m the #1 trader on Manifold. I doubt prediction markets will be useful here. In fact, I think I could probably use prediction markets to “screw with people”. For example “watch out for X, Manifold thinks there is a 15% chance that they have a sexual assault charge in the next year”.
Yes I wasn’t sure it was a good idea. Anecdotally, I forecasted a question on VC funding to female founders. My forecast was negative and accurate and I was surprised of how well most other forecasters thought things were going for female founders. The process of forecasting also forced me to think through the factors preventing more female founders from getting more funding which I found educational. But perhaps my experience doesn’t translate well. That said I think most negative reactions to my idea was mostly me hijacking this post which I think is fair criticism and some people seemed to think it might be worthwhile.
I agree it would not be good if predictions are made about people—that seems really bad a bit like black mirror!
FWIW there was actually someone making such a prediction for ACE back in the days. It stood at 15% chance which I think is uncomfortably high for chance of a scandal in a single year—it implies an 80% chance of it happening over a 10-year period! They also ran another question on a scandal related to racism.
Note that it wasn’t just ACE, it was ACE or any of it’s (~dozen) top or standout charities.
I don’t think assuming full independence makes sense? I wouldn’t go all the way to “if they have a scandal-likely person/culture then they will definitely have a scandal this year, so 15% is the probability they have a scandal-likely person/culture” but I do think that’s a piece of it. Very roughly I’d guess that 15% in one year gives 50% in ten?
Thanks, those are important points. I just quickly came across it by searching for something else and thought I would share as it seemed very relevant.
Whomever you are who is on a downvoting rampage against me: Really? A −6 vote on all these comments when I just tried to make it easier for people to give me feedback anonymously? I know it is stupid of me to care about karma but that was just plain mean of you.
I didn’t vote on any of the poll comments, but one major thing in my mind is that they should have been posted under the comment saying you’re going to be doing the poll, not under the parent comment
I haven’t voted either way on your comments, but in general making several comments as a way to run a poll is not a good idea unless you’re pretty confident others will find your poll interesting. And then if someone does/doesn’t think the poll belongs here they upvote/downvote each question in the poll, magnifying the karma impact.
Then I expect this comment was downvoted because (a) it’s complaining about downvotes and (b) doesn’t engage with what people should do if they don’t think your poll comments belong on the Forum.
I often downvote enough of the options to zero out the karma gain from posting the options, on the theory that the user shouldn’t reap an automatic +1 or +2 for each option posted. I would not strong-downvote on the basis (if other users had found value in the options being listed, who am I to attempt to overrule them?) and was not the strong-downvoter here.
I think people do not get karma from the baseline +1 or +2 that comes with making a new comment.
Fair enough, I will definitely think of better ways to get feedback in the future. It is completely on me that I didn’t realise the karma impact could be magnified like that (or that running polls that way might cost me karma!). And my concerns about karma is insignificant compared to the personal suffering by those affected by the actions of the original post. What’s most important to me is that I can find ways to constructively and meaningfully contribute to reducing the risk of future harms to the community, something I am dedicating more time to than writing here on the forum. Thanks for your patience Jeff!
By doing something to prevent future occurrences, we are risking causing enough trouble so that there is chance people are deterred from engaging with EA and/or important people might leave our community.
There is not much that can be done.
EA is already putting sufficient focus on preventing future occurrences.
This is the wrong place to pose that question.
Specifically doing something by forecasting future occurrences is likely to be net negative. It could e.g. deter more women from engaging with EA and cause ~panic as the numbers could end up looking pretty high.
A prediction of future occurrences is unlikely to be effective.
I think it would be net negative, in the “What is your community doing to prevent sexual misconduct? - Oh, we make bets about it” kind of way.
This. It’s awful from a reputational perspective
But also there seems to be a total lack of upside. Reported sexual misconduct is not the same as actual sexual misconduct (generally it significantly undercounts) so there’s no real indication whether an apparent improvement in the number represents perceptions of improved behaviour or worsened reporting process. Pretty much all the incentives on either side of the bet are perverse (people with relevant knowledge can earn from suppressing information or breaching confidences, and people can earn money from there being lots of sexual harassment or perhaps even buy looser safeguarding policy!), particularly as I can’t imagine it being a liquid market lots of people uninvolved in sexual harassment cases want to bet on.
I think your right that it would be very bad reputationally if the community as a whole was widely perceived as doing this. But it’s also a bit easier to say that if you also believe that this is actually a bad, or at least useless, thing to do on it’s own merits. If you don’t think that, it seems a bit sleazy (even if perhaps correct) to reason ‘this would actually help improve how we deal with sexual misconduct, but we shouldn’t do it because it’ll make us look bad’. Of course, the fact that something looks bad is evidence it IS bad, even if it seem good to you, but not always definitive evidence. Personally I don’t think it would be useful, but I’m not sure how much someone making a Manifold market would actually cause outside people to perceive this as something “EAs” do, or even that they would notice the market at all.
I certainly also think it’d be useless, like most prediction markets in EA.
Could you please indicate why you are disagreeing with me? To make it easier, I will make a sub comment each with possible objections so all you have to do is read and click.
I am genuinely keen to understand what, if anything, can be done to reduce the chance of future offences. I should make it clear, in case there was any doubt: I am talking about offences from anyone, and not from anyone in particular. In other words, I have no reason to suspect that Owen in particular is at increased risk from repeat offences. If anything, I would trust the judgement of other commenters indicating such risks are low.