How Likely Is Someone Accused of Assault by Multiple People to Be Innocent?
obsessionwithregression.blogspot.com
obsessionwithregression.blogspot.com
That's not just a problem of ancient superstition; as recently as the 90s, a number of people were convicted of Satanic Ritual Abuse, with multiple witnesses, and some of the real victims (often women working in places like daycare centers) have been in jail for decades on the basis of clearly unbelievable and coached testimony.
Whenever one of these periods occurs, it's not a one-off act by some individual, nor even a matter of any specific religion. It's a clear indication that the justice system has been fundamentally compromised in the way it treats evidence and renders a final judgement.
This article might have good intentions, but its method could be applied to witchcraft or Satanic Ritual Abuse, and obtain the same kinds of conclusions: If you're accused of witchcraft by multiple people, you're probably guilty. This is still mostly true if we don't believe that witches are a separate kind of people. Increasing our certainty that witches are witches is even good for the people who are accused of being witches.
Counting up the number of accusations is clearly a wrong standard for determining that someone is guilty. The available evidence in each case matters, and accusations are not necessarily independent. The general presumption of innocence until proven guilty is still necessary and still applies to rape as to other crimes.
I remember watching the Frontline reports of this. There were several over a few years and I was just flabbergasted as to how amazingly stupid people where to believe this. That no one in the prosecutors office put the brakes on it. Really sad.
hard to read it strait-faced.
Because the Department of Justice says that a female is one third more likely to be raped not attending college than on a college campus I'd say it is at least possible we're in the midst of a moral panic over sexual assaults on college campuses.
The lack of reliable data, which places the number of campus sexual assaults somewhere in the wildly varying range of .08% to 20% of the population, the pseudo-religious "with-us-or-against-us" mentality, the desire to see justice done extralegally rather than in a court, and the vitriolic shaming of dissenting voices all have the flavor of a moral panic.
The mob-based tribal thinking that "if he's been accused more than once, he must be guilty" is an inherent flaw in human reasoning, of which this program is simply a mathematical model.
Mathematically, the fatal flaw of the model is that it assumes all accusations are supported by evidence of an arbitrary positive weight, and that all accusations are independent. This is never the almost never the case in real accusations. Perhaps before dropping out of school he should read up on other misuses of statistics that have manipulated juries and ruined lives, like the famous "Prosecutor's Fallacy." http://en.wikipedia.org/wiki/Prosecutor%27s_fallacy, which quite literally lead to the death of British Solicitor Sally Clark.
What saddens me most is that some casual readers are now walking around with the incredibly mistaken assumption that it is mathematically sound to assume a person accused thrice is virtually certain to be guilty.
I don't say any of this to diminish the very real problem of sexual assault, but merely to point out that people are surrendering their rights and their reason to mob mentality.
That mentality is precisely what two thousand years of legal scholarship evolved to fight. And whether executed by silicon or scythe, mob mentality is wrong. And quite frankly it is frightening to watch.
Moral panics very rarely involve people coming forward and accusing an individual person of attacking them. It's possible for one or two deranged nut jobs to go around saying that (eg) Bill Clinton raped them. Over a dozen people accusing someone like Bill Cosby who isn't widely hated the way Clinton was, and recounting their individual experiences? Is there any precedent for something like that happening?
>"By spring of 1984, it was claimed that 360 children had been abused."
All of these linked allegations later turned out to be complete fabrications, creations by the investigators invented out of whole cloth, and targeted against the same specific individuals.
In fact false accusations from multiple individuals are a common police tactic. Perhaps the individual has slept with several others, and the encounters ended non-violently but non-amicably. Suddenly consensual encounters retroactively become remembered as non-consensual after questioning by investigators, the very purpose of which is to elicit accusations. As a representative at these hearings, and student advisor, I've witnessed this happen to students many times.
See also the DSK trial, where a French author from years before brought charges that were similarly dismissed. The bandwagoning effect occurs regardless of whether the allegations are true or false. People want money, power, special treatment, and attention, and sadly, false allegations are often an easy mechanism used to get them.
So yes, both false allegations and true allegations are unlikely to be made independently. Not to mention the truly crazy people who are completely uninvolved and will jump in for no discernible reason whatsoever, or in criminal cases CIs who are transformed into additional "victims" to secure convictions.
This isn't a story of math to the rescue. It's a store of bad statistics and not paying enough attention in class, where they usually teach some variant of the prosecutor's fallacy, since this type of falsity comes up often enough to warrant consistent reinforcement that statistics aren't magic, and they can ruin people's lives.
Agreed that the statistics don't tie out.
There is a lot of grey in this area, and there are lots of opinions that can sway people one way or the other.
I was raped, but not violently raped. I've heard arguments from both sides. Can you imagine being a rape victim trying to determine whether they were actually raped or not? It sucks. I don't like talking about it with people outside of the internet. I've seen thousands of facial reactions to people reacting to a blank description of it, because I'm still too ashamed to describe it. Counselors aren't supposed to judge, but they do. I was called a pathological liar, in counseling, years after the rape had happened. The people you think are supposed to be there for you the most, can sometimes understand you the least. I have been argued with and screamed at over whether I was raped or not. It just makes me feel kind of dead inside. I don't have anything to say to you aside from the fact that determining what 'real rape' is, can be as complicated as determining whether witches exist or do not exist, or anything else you think is really complicated. You can't really prove it either way, and all you are really doing is convincing a group of people to think like you, without them having to be you. How many times do I have to tell myself whether I was raped or not, until I believe it? I don't really care. I just know that shit is complicated, and there aren't easy answers to anything that is complicated.
That is one of the reasons why the long-established comparison between McCarthyism and the Salem Witch Trials is so apt. It is not beyond reason that there were people in Hollywood who were genuinely communists; there are communists in just about any profession today that you can think of, just as there were in the early 20th century. Hell, some of them may have even been Soviet agents; we know that Soviet agents operating in the US is something that happened.
The fact that communists did in fact exist does absolutely nothing to excuse the flawed methodology that was used to track down and label communists.
In effect, it just gives a mathematical veneer to the author's guesses. Similar to how wearing a lab coat while stating your opinion makes it seem scientific.
(The most obvious issue is non-independence, but there are plenty of other problems.)
I didn't read the study, but how could anyone know this? I suppose you could assume that a conviction means that the accusation is true, and that exoneration means that accusation is false, but there are a lot of cases where it's not known one way or another. I would be surprised if 90%+ of accusations result in a conviction, so the study must be assuming that any accusation is true if the accused is not exonerated.
There are also other, like if one party believes an assault took place and the other does not, or if the accusation is true but the perpetrator is misidentified.
Please correct me if I'm wrong, but my guess is that the statement above (alone, out of context) is not well supported by evidence.
Advocates are correct to note that this "2-10%" includes some mistakes - true rape victims might be very confused in their testimony or might recant because they get frustrated with the justice system - but it still should probably be regarded as a lower bound on the actual number of false accusations, as it's a subset of the full total.
False accusations that wouldn't show up in this measurable "2-10%" include (a) those in which the accusation is plausible, consistent and the accuser does not recant, (b) false accusations that aren't ever brought to the police.
The amount of people who were raped and never went to the police could be much higher than the amount of people who lied to the police about being raped.
Pretty sure. my guess is that the true number is at least 15%, based largely on the sort of accounting found here:
http://slatestarcodex.com/2014/02/17/lies-damned-lies-and-so...
(The lower bound is 2-10%; the upper bound is 20-40%)
> The amount of people who were raped and never went to the police could be much higher than the amount of people who lied to the police about being raped.
This is true, but here is a factor you are overlooking:
"While rape victims have some incentives to report their cases to the police – a desire for justice, a desire for safety, the belief that the evidence will support them – false accusers have very strong incentives not to – too much work, easier revenge through other means, knowledge that the evidence is unlikely to support them, fear of getting in trouble for perjury if their deception gets out. So I consider it a very conservative estimate to say that the ratio of unreported to reported false accusations is 4:1 – the same as it is with rapes. A more realistic estimate might be as high as double or triple that."
The actual figures are an unknown proportion of the vast majority of the middle (further complicated by the large, and variable, proportion of sexual assaults that go unreported) and any study requires a lot of guesswork. The decision to report an assault, accurately or otherwise, is also evidently not an event independent of other reports of similar assaults, especially since many come after the allegations and name of the accused are revealed.
I see the other comments pointing at this potential problem as well, but I'll phrase it as a question - since the article opens with a reference to Bill Cosby, is it even valid to combine multiple accusations of assault when the accusers are aware of each other, possibly through media exposure, and may have come forward as a result?
And while there is a (perhaps unsurprising?) correlation between independent accusations and probability of guilt, how can this statistical approach be responsibly used, and should it, and what for?
Assault is awful and I'm sure the majority of assaults with even one accuser is (or should be) straightforward), but it seems like there are other uncontrolled possibilities here besides the independence of accusations, like whether the accuser and accused know each other.
I'd imagine that would make a difference when compared with a regular person who is not in the public spotlight or doesn't have a large estate.
Does this "intuitive thought" actually represent real world events?
It would be really interesting for someone to actually pull lists of famous celebrities from the last few decades (from mentions in media, for instance), look at rape accusations, and classify results. I don't know how you could compare to "regular" guys as most non-famous people don't make the news even if accused of rape, but a comparison of conviction rates could be illuminating.
This doesn't account for whether or not the accusations are independent of one another, which in this case most were not.
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.
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.
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_wlfwnUoYF...).
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-i...
2) The accuracy of human memory is disappointingly terrible.
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
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...
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.
For example, while Michael Jackson was accused of assault on two occasions, I think it's fair to also consider that the prospect of a financial payoff could be a motivation to press charges against him. In the first such case, he eventually settled for $15 million towards a trust fund for the accuser, despite no evidence and a strong case that the accusations were only levied as a money grab (the accuser was seeking a monetary settlement instead of criminal charges from the very beginning)
In the second public accusation (years later), Michael was found innocent. Additionally, given the way things went in the first case, it is entirely believable that someone could see accusing Jackson of child molestation as a means to an easy payoff.
I'm not saying Michael Jackson is innocent beyond all doubt, but the court clearly determined it was impossible to make a case that he was guilty beyond all doubt. I do think it's important to consider other factors in situations where assault (sexual or otherwise) is accused though.
For what it's worth, I think it's very likely Bill Cosby is guilty of most of the allegations he is currently party to.
What about the substantially different personal cost of making an immediate accusation vs a delayed one? The same applies to someone like Snowden who could have waited until he was in his 70s before blowing the whistle.
Leaving college aside, there has been a lot of discussion of rape in the military. Apparently there are a number of repeat predators, in an environment where bringing an accusation can result in the accuser's being thrown out of the military, shunned, or assaulted. (Male victims abound, by the way.) I would love to see this simulation with numbers based on the military experience. Meanwhile the military are reluctant to allow outside interference in their complicity with rape. The missing item in the reporting is how many of those accused were accused more than once. http://www.militarytimes.com/story/military/pentagon/2014/12...
False reports/accusations of crime in general tend to be low, in the 2-8% range reported here. By this logic, we can dispense with this whole silly and expensive criminal justice system and just convict everyone based on simple accusation. We used to call this sort of thing "lynching", and it is generally considered a good thing that it's become rare.
"It is better that ten guilty persons escape than that one innocent suffer"[1]
Also note that the "low rate" for false accusations is for cases were the false accusation is proven:
"The determination that a report of sexual assault is false can be made only if the evidence establishes that no crime was committed or attempted. This determination can be made only after a thorough investigation. This should not be confused with an investigation that fails to prove a sexual assault occurred. In that case the investigation would be labeled unsubstantiated. The determination that a report is false must be supported by evidence that the assault did not happen."[2][3]
By this standard, cases of rape that are actually proven are also low, depending on where you look in the 7% range of reported[3]. So the "we just don't know" figure is very, very high, we simply do not have strong evidence either way[4]. Other studies have found false report rates much higher [5][6].
The Air Force study[6] is interesting in that admissions of false accusations increased dramatically when it became known that a Polygraph would be administered (though AFAIK, the Polygraph information itself was not used as indication of a false report, only an actual admission). The raw data of admitted false cases was then combined with other characteristics of those cases to develop a model for "likely" false accusations (things like "does the rape cause problems or solve problems for the accuser", "does the report follow widely assumed patterns of rape vs. real patterns", "do the injuries show patterns that are typical of self-infliction" etc.) With that model, applied conservatively, the false report rate was above 50%.
There is also an interesting investigation showing that the 2% figure that pops up in many "studies" books and papers on the subject can all be traced to a single off-the-cuff, non-researched remark.[7]
So the reference figures are at best flawed and way too uncertain to base a model on.
Additionally, the statistics are incorrectly applied to a single event ("How likely is someone ..."). The statistics only work for a sufficiently large population.
Finally, as many others have pointed out, the events are not uncorrelated, and the assumption that they are and that "lots of reports must mean there is something there" is a well-documented fallacy of human cognition. The McMartin case is an obvious one, but there is also the famous case of the French city of Orleans, where an entire city was convinced that girls were being drugged and abducted in fashion stores, which turned out to be completely unfounded.[8][9][10]
So please pack up this model, or let's all just abandon our legal system, due process and meet for a happy lynching.
[1] http://en.wikipedia.org/wiki/Blackstone's_formulation
[2] http://yesmeansyesblog.wordpress.com/2010/09/09/false-rape-a...
[3] https://rainn.org/get-information/statistics/reporting-rates
[4] http://lawprofessors.typepad.com/crimprof_blog/2004/12/2_fal...
[5] http://www.informationisbeautiful.net/visualizations/rape-a-...
[6] http://www.fathersmanifesto.net/mcdowell.htm
[7] http://digitalcommons.lmu.edu/cgi/viewcontent.cgi?article=22...
[8] https://www.kirkusreviews.com/book-reviews/edgar-morin/the-r...
"While it can be difficult to determine guilt in a single case, my intuition is that a person independently accused of assault by multiple people is quite unlikely to be innocent."
Just so we're clear here: intuition is a heuristic, which is incredibly flawed and only intended to be used as a potential risk indicator for split-second decisions. Any time you hear of something risky you are more inclined to believe it because to not believe it would potentially open yourself up to risk. The only way to mitigate it is to have additional information or motive that overrides the unknown risk.
Back to the model: they cite one study on sexual assault. Do you know how many actual studies there have been on sexual assault? Not many. Do you know how many of them had widely varying results? Most of them. Do you know how many had different methodologies which changed the nature of the model? All of them. So I guess let's ignore that someone could probably pick a different study that would change the results.
So what does this particular study say? The study involved asking 2,000 men at one college in one state about their sex lives. (Random sampling? Who needs that?) Of those men, about six percent admitted to sex that fell under the legal definition of rape without consent of the woman, which includes force or alcohol.
Consider for a moment that the legal guidelines basically say that if you have a bottle of wine with your girlfriend and you end up in bed together, it's rape if there wasn't a consent agreement. In this case they're only asking the men and making a judgement based on the legal framework, and not asking the women about their relationship or if there was consent, which could possibly change the results. But anyway,
"I think the UVA episode illustrates precisely why statistics are so important -- because anything can happen in a single story, making it a risky thing to hang a cause on."
So why would you create a model based on a single limited study, and not at the very least recreate your model based on the data from other studies?
"Regardless of what really happened at UVA, the broader trend is clear: the rate of campus sexual assaults is high (20%, says CDC, although better data should be collected); the rate of false accusations is low (this review cites 6 studies which all yield estimates between 2% and 8%, lower than the rate of false reports for car theft)."
Another study said the campus assaults were 90%. The CSA final report put it at 13.7%. And (all) the false accusation studies were collected and analyzed by Philip N.S. Rumney and determined that due to the unreliable research methodologies used, along with unreliable judgements by law enforcement, it was impossible to give a plausible statistic.
I've always found that people rely on statistics when they want a nice simple number to sum up their opinions, and don't want to deal with the fact that there is no simple, clean way to summarize complicated, wide-ranging circumstances, events and behaviors.
"Can't consent" is different than "lack of consent", which is different than "consenting", etc. Consent is a highly varied and complex subject. There's explicit consent, implicit consent, informed consent, implied consent, non-consent, consent agreement violation, consensual non-consent, re-established consent, etc etc. Legally speaking (Note that I am not a lawyer) in America, there's very few laws that determine what is consensual or not. Usually it has to be determined by a jury when it goes to trial. It's because of this that, legally speaking, having alcohol does not alone remove the ability to consent in all cases.
Non-legally-speaking, I think you missed my original point. Nowhere in my commentary, nor in the study, does it say that anyone was actually intoxicated past the point of being able to consent. It merely says that within the legal framework someone could have been held liable for non-consensual sex. All this requires is that the person was intoxicated (about .08 BAC, or 1 beer, in the US).
If you feel like condemning people for this, so be it. All i'm saying is a study that generalizes based on guidelines like this is flawed, amongst other things.
But that definition is ridiculously broad, criminalizing a large part of normal sexual behavior, and also one-sided, in that two drunk people having sex means the guy is at fault.