Warp wrote:
>   How many times do I have to write "but I understand why it wouldn't work
> in practice" before it sinks in?

Because I'm telling you why it wouldn't work even in theory. Even if you 
*found* that statistical correlation, the answer is still "No, it wouldn't 
work."

>   (And how many times do I have to write that it doesn't change my original
> point?)

And that's why I think you don't understand what I'm saying. It completely 
invalidates your original point, as well as all the rewordings of the 
original point you've expressed.

Because your original point is *incorrect*, and there is *no way* to make it 
correct, even by completely eliminating all prejudice and fear of racism and 
profiling from the entire population. It's not that "it wouldn't work in 
practice." It's that "the math says it wouldn't even work in theory."

>> I completely understand. And I'm telling you "the math shows that there are 
>> no contexts in which crime could be more efficiently stopped by making the 
>> distinction," in any case where you don't already know something about the 
>> specific criminal you seek.  And yet, you still haven't said you understand 
>> this. Do you?
> 
>   No.
> 
>   Firstly, I don't see how the math says that.

Look up Bayesian statistics. It's a common misunderstanding of how things work.

As I've tried to explain, it's a difference between reducing false negatives 
and reducing false positives.

http://betterexplained.com/articles/an-intuitive-and-short-explanation-of-bayes-theorem/

Your "test" is the ethnicity, and your "result" is whether the person 
committed the crime. The number of people who are *not* criminals far 
outweighs the number of people who *are* criminals.

That's why I've been pounding on this. You seem to not understand "it 
mathematically cannot make a difference."

>   Secondly, it was not my point. My point *still* is "if it could make a
> difference, it would make sense to use it".

OK. I don't know what else to say besides "It's impossible for it to make a 
difference, so there's never any way in which it makes sense to use it."

>   You talk like using a profile is mutually exclusive with using other clues.

Having a profile is unnecessary if you have other clues. You match the 
person to the clues, even if that clue is a clue about the race or gender or 
age of the suspect.  That's not profiling. That's using clues to eliminate 
suspects.

>   If your clues have narrowed down the list of suspects to 100 people, 50
> males and 50 females, and your profile says that the criminal is most likely
> a male (for example, in rape cases it's pretty unlikely for the criminal to
> be female), you reduced your list from 100 to 50.

You understand that is eliminating false positives, right?  By the time 
you've already narrowed it from 300,000,000 people down to 100 people, you 
have probable cause to suspect some of those people. You're not eliminating 
the females because "in rape cases it's unlikely to be a female." You're 
eliminating the females because you have genetics spilled all over the 
victim telling you it's a male. That's no more profiling than "the rape 
happened in Boulder, so let's not suspect people who were in Beijing that day."

If the victim got raped with a broomstick, you wouldn't be eliminating the 
females, because there would be no *evidence* outside the profile that would 
allow you to eliminate them.

>   For some reason you seem to think that "using a profile" means "discard
> all other evidence and *only* use the profile" (in some past post you even
> explicitly talked about profiles *increasing* the amount of subjects, which
> at the time I didn't understand at all because it made no sense, but now I
> think you were implying "using a profile *and* discarding all other clues").

I said that using a profile based on the total number of criminals with 
attribute X can increase the number of people you look for without catching 
someone if there's an even *higher* number of people who have X that are 
*not* criminals.

If you say "We should throw away the rotten fruit. 80% of the rotten fruit 
is strawberries. Only 20% of the rotten fruit is blueberries. Hence, we 
should examine the strawberries more closely."  But if you have more than 
five times as many total strawberries as total blueberries, you'll be 
*increasing* your work by concentrating on strawberries rather than picking 
fruit at random to examine because there's a lower percentage *of 
strawberries* that are rotten.  And it has nothing to do with racism or 
prejudice.  It's just math.

I know you'll say "but what if it helps?"  I'm giving this as an example of 
how my statement could make sense, not a reason why your statement was 
incorrect.

> I don't understand where you are getting that. It's certainly not something
> I have said nor implied.

But if you already suspect someone of a crime, why do you need the profile?

-- 
Darren New, San Diego CA, USA (PST)
   Linux: Now bringing the quality and usability of
   open source desktop apps to your personal electronics.

Post a reply to this message