Well, I agree that once you get to court, it’s best if it’s evidence or nothing. Prior to, I dunno. Some recent threads on race-related issues have me so discombobulated I don’t much trust my own experience on some matters, and have practically no trust for anyone else’s. I readily confess I default quite often to “common sense” and I’m saying, ideally, if I can restrain myself and do the hard work (and such work is even possible), it’s best if I don’t do that. My “sense” and your “sense” can differ so wildly I’m inclined to toss sense altogether. It’s a recent recurring trauma of mine, I have to say, that sensibilities can really seem to be for crap. Inescapable, but crap all the same.
“Common sense” in this context is really nothing more than Bayes’ Theorem. It is hardly unreasonable to expect people to update their beliefs in accordance with this rule, even though it involves a subjective estimation of conditional probability. The problem is not that people have beliefs, or “common sense,” but that they do not update them very well when faced with new data.
So the question of this thread is really whether people can use their beliefs they have formed about the state of the world to estimate specific conditional probabilities of guilt or innocence. The answer is, in my opinion, of course they can.
Doesn’t it have to do with limited resources?
Suppose you have one cop, with a thousand hours of time available to him. He is confronted with a hundred investigations, each of which will require fifteen hours to resolve if they are real crimes.
Fifty of the investigations are of redheads, and fifty are of blondes. Nine out of ten investigations of redheads turn out to be false reports; nine out of ten investigations of blondes are real crimes. Wouldn’t it make sense to concentrate on the investigations of blondes? If you have some spare time, or if the pressure from the RedHead Liberation Front gets too intense, you can thrown some hours at the redhead investigations, but you are reasonably sure that most of that time is going to be wasted. Yes, this means the genuine reports about red heads are going to be short changed. But overall, this is a more efficient strategy.
Yes?
Regards,
Shodan
Only if you can compellingly explain how hair color determines veracity. Your scenario is too implausible for us to apply to the real world.
So you don’t understand the nature of a theoretical? You’re going to have trouble with this discussion, ISTM.
Regards,
Shodan
Notice how the hypothetical that Loopy presented was based on a world similar to the one we know. A place where deer are expected to be in the woods and not in the streets of Manhattan. He used it to illustrate how he’d evaluate a particular claim in a setting that is realistic and familiar to most people.
Your hypothetical takes place in a world in which hair color, of all things, is associated with veracity. We have no idea why this association would exist, or whether hair color is supposed to determine veracity or is merely correlated with factors that do, and yet we are supposed to accept this as a given. Okay. But wait! On top of that craziness, you create a situation in which only one cop is supposed to investigate 100 cases and has to decide which ones to focus on first.
How is any of this at all comparable to what goes on in the real world? You’re not really elucidating anything with your hypothetical. You might as well be dreaming up a world where unicorned buffloes roam.
How’s that? I being a smart criminal dyes my hair red. Your efficient policeman looks at my hair, and puts me on the bottom on the list. I wave good-bye and go back to cutting off the head of the teenager that everyone’s looking for.
All the while, your effecient policeman is looking for a blonde, because we all know what they’re like.
Or you do propose in order to be efficient, your policemen drives around with a certified beautician in the crusier, in order to verifiy the roots of all redheads? To be sure, no one’s trying to pull a fast one. Perhaps a quick flash of the pubes, would be more efficient?
Yes?
That’s an efficient strategy to you?
Let’s put it this way.
When a 20-something woman is murdered, who’s the first person the cops want to talk to? The boyfriend. Why? Because based on past murder cases for this demographic, the most likely suspect is the boyfriend. Doesn’t mean the boyfriend is guilty in this case. It just means that the boyfriend is going to be checked out a lot more thoroughly than anyone else.
And I don’t think it’s a big stretch for the general public to be suspicious of the boyfriend either. And to evaluate the evidence in light of the knowledge that a young woman is more likely to be murdered by her boyfriend than anyone else. So when Mr. Smith the pharmacist who saw the victim when she came into his store three days before she was murdered doesn’t have an alibi, we give it a certain level of suspicion. When the boyfriend doesn’t have an alibi, we treat that more suspiciously.
Not that we can use such demographic statistics in court. We still have to prove beyond a reasonable doubt that the defendent commited the crime, and showing that when 20-something women are murdered 63% of the time the killer is the boyfriend doesn’t remove reasonable doubt. What about the other 37% of the time? But it does provide a rule of thumb to guide the police in their initial investigation. Of course you investigate the boyfriend in a case like this.
Now, statistical correlations for other things don’t work so well. Suppose you have statistics that show that whites are per capita twice as likely to rape other whites than they are to rape blacks, while blacks are twice as likely to rape blacks. Just making up numbers. That would tell you that if you have a rape victim who couldn’t identify the race of her attackers, you’d be more likely to suspect whites if the victim is white, and blacks if the victim is black.
Trouble comes when you try to use the statistics in the reverse way. If the victim identifies her attacker’s race, does that mean she’s more likely to be filing a false report if she reports a cross-racial rape? I don’t think that can be supported. See, unlikely events are unlikely. But events happen. It’s unlikely that the roulette wheel will land on 17. But it has to land on some number.
So I would say it is not correct to guess that an alleged victim of a cross-racial rape is therefore twice as likely to be lying as someone who alleges a same-race rape, even if same-race rape is twice as likely. We already know where the dice fell, and we can’t argue backwards from that. 17 is an unlikely number, but we can’t argue that someone who claimed to have hit a 17 is more likely to be lying than someone who (say) claimed to have hit black. It seems tempting to believe that the 17-claimant has a 39 in 40 chance of lying, while the black-claimant has a 1 in 2 chance of lying, but we can’t say that.
I think you explained all that very well, Lemur. I’d go on to say, however, that the reason why most rape involves members of the same race is due to most victims being acquaintances to their attackers. Since most people’s acquaintances have race in common, then you’d expect the rape stats to reflect that.
If you have a rape victim who couldn’t identify the race of her attackers, but you know that almost all her friends and associations are of another race, it may not make sense to assume that her attacker was of the same race.
But other than that, I think you summed up the whole thing very well.
Basically what Lemur said. Statistics are definitely useful in the initial investigation (as they were used in the thread about the Duke case). This does not mean that they should be used as evidence; it is only a guideline, not a law. That this may influence further criminality is irrelevent, as its current usage is not to determine guilt or innocence but rather to try the case in the court of public opinion.
As noted, in that thread there was also a statistic that white men are more likely to engage in gang rapes. So which statistic guide the initial investigation?
a. Statistically speaking there are few cases of white men gang-raping black women or
b. When gang rapes do occur, they are more often committed by white men?
Why doesn’t the ‘burden’ go, "this woman says she was ganged raped by white men, statistical we know most gang rapes are by white men, therefor statistically we have a good case, against these white men?
Yeah, but in real life, this doesn’t make any sense from the perspective of a police officer. What are they supposed to do: start rounding up white men or black men and quizzing them if they raped some poor woman?
I maintain that if anyone is going to investigate something – whether a crime, natural phenomena, or historical event – a mass of statistical data does not actually inform someone of what really happened in one particular case, no more than collecting the past rolls on a roulette wheel is an effective means of predicting the next number to come up. If an investigator doesn’t know something, they should admit to themselves that there is a known unknown, and not pretend that statistics constitute some flimsy kind of evidence that can be relied upon until something better comes up.
Most importantly (to me at least) why should race even color our view of what is alleged? Is it because there happens to be stats out there on white-on-black rape and that’s why we would think they are applicable?
Consider the caveat that I raised in response to Lemur’s explanation. The main reason why it is reasonable to assume that a rapist is probably the same race as a person who can not ID their attacker is because usually it is an acquaintance race. So the person’s race is an incidental; we shouldn’t look at race without regard to the underlying factors involved.
Similarly, in this case, we shouldn’t look at race without regard to the underlying factors. The fact that the woman involved was introduced into the home of the men and paid to put on a sexual performance tells me that whatever influence race has on the equation can not be estimated using limited stats.
I agree that “suspect was probably a white male” is even less helpful that “suspect was a white male”. So you’ve narrowed it down to 80 million people there. Except police work isn’t usually a matter of Eratosthenes’s seive, where we start with a suspect pool of the entire world, then gradually eliminate categories, until we are left with the only person in the world who can’t be eliminated.
And to bring up another point, if we evaluated such cases by probability, the more specific an alleged victim could be in identifying her alleged attacker would therefore have to make us less and less likely to believe her story. So if a white woman could identify her attacker as a white man, we believe her because there are so many white men. But if she identifies her attacker as particular person, we would have to say that it’s almost certain that she’s lying, because what are the odds out of all the people in the world to rape her, it’s this one guy? We can believe she was raped, but raped by this guy? 100 million to one!
I can’t think of a scenario where statistics would be useful in the initial investigation. I’d appreciate someone providing an illustrative hypothetical (one applicable to the real world).
Perhaps profiling is an example? We know the qualities of a “typical” serial killer, so we can use this information to winnow down the pool of suspects, or at least give the investigators a general idea of who to look for. But the profile isn’t built off of one or two traits, like race and gender. There’s a bunch of ones.
I don’t see how such an exercise can be helpful in assessing whether or not a suspect is a good “candidate”, once he’s been caught. The DC Sniper shattered the expert’s expectations across the board. Could his defense team have used this as evidence that the police had snagged the wrong guy? Is it possible to estimate the probability that a specific individual is responsible for a crime, given data drawn from the general population?
I think that’s an excellent example of an area of criminology that could certainly benefit from accumulated data, and statistical analyses might provide good leads, if enough data were available. I think what the DC sniper case taught the experts is they don’t know enough about serial killers yet. It’s possible the type is simply too diverse in nature, and too rare, to make that a tractable problem.
There are very many situations where you just cant practically check out every possible scenario.
We pretty much always eliminate some possibilities due to unlikelihood rather than investigating each scenario as having equal weight - at best we only check them when the most likely scenarios have been exhausted first and even then we’ll tend to put it into the ‘cant be solved’ category than expend resources on them.
Without concrete exampels I dont think this debate can really be settled because its more a question of how its done in practise rather than whether it happens at all.
Otara
If a young woman is killed, statistics tell us that her boyfriend is the most likely suspect.
If a drug dealer is killed, statistics tell us that the killer is likely either another drug dealer.
If a child is murdered, statistics tell us that a family member is the most likely suspect.
When petty vandalism is committed, statistics tell us that the perps are likely to be teenagers.
And so on.
What this can do is steer us to fruitful lines of investigation, and prevent us from closing down fruitful lines of investigation. So when a child ends up dead, we don’t exclude the parents, because “who would kill their own child?”. Statistically, we know that if a child is killed they are much more likely to be killed by one of their parents than a stranger. The parents are the top suspects due to the nature of the crime.
The idea is that people don’t generally murder for no reason, they murder for a reason. And they’re more likely to have a reason to murder someone they know than a stranger. And this is why serial-killer cases are so difficult to solve, because the serial-killer doesn’t have a motive to kill, like most killers do. Usually murders have a pretty simple story. Adultery. Jealousy. A drug deal gone bad. A robbery gone bad. Revenge. Generally when someone ends up dead you look at their friends, family, neighbors, coworkers, partners in crime, and so forth, and eventually a story starts to emerge. So when you find bloody fingerprints on the murder weapon you don’t compare the prints to the entire population of North America, you first compare to the boyfriend, then the last person she was seen with, and so on.
Which is why, when Chandra Levy disappeared, her congressman boyfriend was the top suspect, even though he was never charged with anything and no evidence linking him to the murder was ever made public. He was the top suspect because boyfriends are statistically the most likely suspects when young women are murdered. Not to say that young women can’t be murdered by random strangers. It happens. But motiveless crime is pretty rare.
Yep. It’s not like the media hyped the possibility of a congressman having committed murder most foul. What a fantastic example of some stone cold, dispassionate, scientific analysis of statistical data to name the prime suspect in the absence of hard evidence.
Since the alleged victim in the Duke University case was black, then statistically it would be better to assume that she was lying when she said that the only black person at the party was not the rapist.
Regards,
Shodan