Printable View
First, I want to apologize if this isn't really a "math" question.
The background: This question ties into the Duke rape case, but it's not really about that particular case. It's about using statistics related to crime.
The question: One debater on another site says that a black woman who claims to be raped by a white man should be viewed with skepticism because the rate of black women being raped by white men is so low.
Is this a reasonable way to use crime rate data?
Julie
Of course not. Maybe the rate of black women being raped by white men is so low because no bodybelieves them when they make the claim or because the black women do not think they will be believed so they do not even make the claim most of the time.
My point is by using such statistics in a legal setting you invalidate the statistics. In law you must always consider the effects of the precedent set by what you do.
Here is the thing about probability and statistics:
They ONLY apply in the limit of large numbers. That is, they can tell you what to expect after a large number of trials.
Example: In the long run, the casinos are going to take more money from the patrons than the patrons will take from the casinos. That's how they make their money.
BUT... it is not uncommon to hear about someone winning big every once in a while.
Statistics can not be applied to individual cases such as the one you mentioned.
w
Dear JulieQuote:
Originally Posted by jsgoddess
Well, it's certainly a potent method of ensuring that any unjustifiably low conviction rate gets perpetuated and amplified. The resultant feedback loop would make the rate of white-on-black rape convictions dwindle asymptotically towards zero over time. If there were any bias in the system, then feeding back the bias into the system like this would definitely make the system worse at delivering justice.
Funnily enough, a similar problem occurs in my field. There have been few successful prosecutions or law suits of a large class of deniable crimes or torts with which I am concerned. This makes it easy to stigmatise, on a statistical basis, any alleged victim or plaintiff of a deniable crime or tort, as likely (by application of losely Bayes'Theorem-based reasoning), to be "deluded", much as all those alleged black rape victims could so easily get stigmatised as malicious false rape complainants.
Another example is "profiling". If they think, perhaps mistakenly, that black people are more likley to be shoplifters, store detectives will catch more black shoplifters, reinforcing a belief that might have been mistaken in the first place.
In short, feedback loops of this kind amplify errors. For ths reason, they aren't acceptable.
John
PS Only by analysing this as a "math question" can one reach the obviously correct answer! :-)