4 ms·

(I'm the author of the post) the fact that accusations may not be independent is of course important (although I'd question the aptness of the above commenter's

(I'm the author of the post) the fact that accusations may not be independent is of course important (although I'd question the aptness of the above commenter's comparison to witch trials); I'd welcome your suggestions for incorporating correlations into the model. This is one of the reasons the tool discussed in the NYT's article could be powerful -- because it allows us to be more confident that accusations were levied independently, since people submit to the tool without knowing about other accusers.

logicallee 12y ago
The comparison with witch trials is simple. People can be totally independent, and not know each other, and independently conclude that you're a witch. Why would they do this? Obviously due to some factor about you. The mistake you make is using Occam's razor to conclude that that factor is "being a witch". The other poster used the example of witches because we all know there is no such thing. There is 0% chance of someone being a witch, whom nobody has ever so much as suspected of being a witch. But there is also 0% of someone being a witch, about whom dozens of totally independent people, in different locations, throughout the person's life and career, have independently come to the firm and undeniable conclusion that they are a witch. This is why this was the example that was used.
- darkhorn 12y ago
  Everyone says that [put some religion here] is true. All these people can't be wrong, can they be? Therefore this religion is true. Applies not only to religion but also to heating home, sickness, astrology, cutting penices of children etc.
  - logicallee 12y ago
    Not bad :) Author should post "How likely is someone to be the son of God based on number of disciples?" and we can see where 12 fits in.
lectrick 12y ago
You are not accounting for a few typical problems with these cases. 1) Eyewitness testimony ("personal accounts done a certain time after the event in question") is disappointingly terrible. http://www.scientificamerican.com/article/do-the-eyes-have-it/ http://www.scientificamerican.com/article/do-the-eyes-have-i... 2) The accuracy of human memory is disappointingly terrible. http://www.theatlantic.com/health/archive/2013/11/how-many-of-your-memories-are-fake/281558/ http://www.theatlantic.com/health/archive/2013/11/how-many-o... ... Especially under stress/duress/emotionally-charged events. https://en.wikipedia.org/wiki/Effects_of_stress_on_memory https://en.wikipedia.org/wiki/Effects_of_stress_on_memory Lastly, regarding the anonymous reporting tool cited in the NYTimes article- that is no different from an anonymous witchcraft reporting tool. Any single individual who accrues too much community ire (whether deserved or not) can get anonymously accused of something (and at minimum, publicly humiliated or reputation-damaged) by a collusive group of individuals (who can act with complete impunity due to anonymity). It is for this reason that I believe, difficult as it is to do, that all accusers should be known... unless we also allow anonymity to the accused while they fight off the accusations! A little bit more controversially (and I freely admit this may be ENTIRELY up for debate, YMMV), there may exist a false-accusation bandwagon effect due to the economic forces at play https://omegavirginrevolt.wordpress.com/2014/11/23/the-false-accusation-bandwagon-effect/ https://omegavirginrevolt.wordpress.com/2014/11/23/the-false... Essentially, each new woman who steps forward and does not file a criminal suit but a civil one makes it more likely that (for example) Cosby's liability insurer will force him to pay out to alleged victims regardless of actual guilt. In addition, the accuser, who is likely someone that the "fame" spotlight has left in the past, can enjoy some more time under it. You fail to account for the additional financial/fame ("perverse") incentive of joining a bandwagon in this regard, especially when the accused is a wealthy individual. I sincerely hope that the above does not come off as anti-victim, misogynist, or otherwise. I just think we need to remember the lessons from the Witchcraft Era as well as from psychology research of the past few years.
lotharbot 12y ago
> "people submit to the tool without knowing about other accusers" Not necessarily. People can conspire to submit multiple accusations to the tool against the same person. Even without actual conspiracy, someone who is a target for other reasons could have multiple false accusations independently levied against them. Example: I have a friend who runs a religious ministry for ex-members-of-another-religion. This draws the ire of current-members-of-the-other-religion, and leads to a lot of things they post being flagged as abusive (even things like photographs of themselves.) In some cases, it's clear that there's a group conspiring -- something posted years ago will get flagged multiple times over the course of a day (FB once investigated and located a private group that was being used to coordinate mass-flagging efforts.) In other cases, it's just the most recent thing they posted, likely being independently flagged.
jamessb 12y ago
Perhaps a better example of how a naive assumption of statistical independence has led to a severe miscarriage of justice is provided by the case of Sally Clark: https://en.wikipedia.org/wiki/Sally_Clark https://en.wikipedia.org/wiki/Sally_Clark I would be extremely wary of your claim that "multiple accusations enable us to get enough evidence to convict even in parameter regimes where a single accusation is never enough" (https://www.dropbox.com/sh/drrvcd5zzqclhbt/AACdWnx_wlfwnUoYFme9UgX9a/exercises_with_binomial_posterior_probabilities.pdf?dl=0 https://www.dropbox.com/sh/drrvcd5zzqclhbt/AACdWnx_wlfwnUoYF...).

logicallee 12y ago

The comparison with witch trials is simple. People can be totally independent, and not know each other, and independently conclude that you're a witch. Why would they do this? Obviously due to some factor about you. The mistake you make is using Occam's razor to conclude that that factor is "being a witch". The other poster used the example of witches because we all know there is no such thing. There is 0% chance of someone being a witch, whom nobody has ever so much as suspected of being a witch. But there is also 0% of someone being a witch, about whom dozens of totally independent people, in different locations, throughout the person's life and career, have independently come to the firm and undeniable conclusion that they are a witch. This is why this was the example that was used.

darkhorn 12y ago
Everyone says that [put some religion here] is true. All these people can't be wrong, can they be? Therefore this religion is true. Applies not only to religion but also to heating home, sickness, astrology, cutting penices of children etc.
- logicallee 12y ago
  Not bad :) Author should post "How likely is someone to be the son of God based on number of disciples?" and we can see where 12 fits in.

lectrick 12y ago

You are not accounting for a few typical problems with these cases. 1) Eyewitness testimony ("personal accounts done a certain time after the event in question") is disappointingly terrible. http://www.scientificamerican.com/article/do-the-eyes-have-it/ http://www.scientificamerican.com/article/do-the-eyes-have-i... 2) The accuracy of human memory is disappointingly terrible. http://www.theatlantic.com/health/archive/2013/11/how-many-of-your-memories-are-fake/281558/ http://www.theatlantic.com/health/archive/2013/11/how-many-o... ... Especially under stress/duress/emotionally-charged events. https://en.wikipedia.org/wiki/Effects_of_stress_on_memory https://en.wikipedia.org/wiki/Effects_of_stress_on_memory Lastly, regarding the anonymous reporting tool cited in the NYTimes article- that is no different from an anonymous witchcraft reporting tool. Any single individual who accrues too much community ire (whether deserved or not) can get anonymously accused of something (and at minimum, publicly humiliated or reputation-damaged) by a collusive group of individuals (who can act with complete impunity due to anonymity). It is for this reason that I believe, difficult as it is to do, that all accusers should be known... unless we also allow anonymity to the accused while they fight off the accusations! A little bit more controversially (and I freely admit this may be ENTIRELY up for debate, YMMV), there may exist a false-accusation bandwagon effect due to the economic forces at play https://omegavirginrevolt.wordpress.com/2014/11/23/the-false-accusation-bandwagon-effect/ https://omegavirginrevolt.wordpress.com/2014/11/23/the-false... Essentially, each new woman who steps forward and does not file a criminal suit but a civil one makes it more likely that (for example) Cosby's liability insurer will force him to pay out to alleged victims regardless of actual guilt. In addition, the accuser, who is likely someone that the "fame" spotlight has left in the past, can enjoy some more time under it. You fail to account for the additional financial/fame ("perverse") incentive of joining a bandwagon in this regard, especially when the accused is a wealthy individual. I sincerely hope that the above does not come off as anti-victim, misogynist, or otherwise. I just think we need to remember the lessons from the Witchcraft Era as well as from psychology research of the past few years.

lotharbot 12y ago

> "people submit to the tool without knowing about other accusers" Not necessarily. People can conspire to submit multiple accusations to the tool against the same person. Even without actual conspiracy, someone who is a target for other reasons could have multiple false accusations independently levied against them. Example: I have a friend who runs a religious ministry for ex-members-of-another-religion. This draws the ire of current-members-of-the-other-religion, and leads to a lot of things they post being flagged as abusive (even things like photographs of themselves.) In some cases, it's clear that there's a group conspiring -- something posted years ago will get flagged multiple times over the course of a day (FB once investigated and located a private group that was being used to coordinate mass-flagging efforts.) In other cases, it's just the most recent thing they posted, likely being independently flagged.

jamessb 12y ago

Perhaps a better example of how a naive assumption of statistical independence has led to a severe miscarriage of justice is provided by the case of Sally Clark: https://en.wikipedia.org/wiki/Sally_Clark https://en.wikipedia.org/wiki/Sally_Clark I would be extremely wary of your claim that "multiple accusations enable us to get enough evidence to convict even in parameter regimes where a single accusation is never enough" (https://www.dropbox.com/sh/drrvcd5zzqclhbt/AACdWnx_wlfwnUoYFme9UgX9a/exercises_with_binomial_posterior_probabilities.pdf?dl=0 https://www.dropbox.com/sh/drrvcd5zzqclhbt/AACdWnx_wlfwnUoYF...).