Warp wrote:
>   Originally I was comparing "concentrate the resources on people who fit
> the profile better" vs. "distribute the resources evenly among all people".
> That last part is important.

I *still* understand what you're saying. You *still* aren't saying what 
"profile" you're talking about. Using a profile based on race is *still* not 
as efficient as a profile based on probable cause, regardless of the 
political correctness.

And people aren't complaining about profiles based on actual causal 
relationships to the crimes being investigated. Nobody is complaining about 
"profiling" people who are driving badly for sobriety tests, and nobody is 
complaining about "profiling" people who don't pay their taxes for 
immigration status testing.

And people aren't complaining about random testing, but you repeatedly seem 
to conflate "random" testing with "random testing of people who fit a 
profile", which are two entirely different things. Plus, random testing 
might have a deterrent value in the case of sobriety tests, but it's not 
really possible to deter someone from having their ethnicity.

Yet you still haven't indicated that you understand that.

> when you narrow your input by some factor, the same resources will be more
> efficiently used compared to if you distributed the resource evenly to the
> entire input.

Yes. And again, nobody is disputing that. Nobody is saying that if you know 
something about a criminal, you should use that information to not 
investigate people who could not be the criminal.

A profile of the type you seem to be recommending is not that sort of thing, 
for reasons I have repeatedly tried to explain and you *seem* to be ignoring.

>   Your argument was that "using ethnicity as a basis of narrowing down the
> possible suspects is not the *best* factor". In other words, you were not
> disagreeing with what I was saying (in other words, that narrowing the
> input by using some known factor helps utilize resources better), you were
> simply disagreeing with the notion that ethnicity would be the best factor
> to do that.

No, I was disagreeing on several fronts:

1) You *cannot* use the statement "Most of the criminals have trait X, hence 
we should focus our questioning on people with trait X." The math doesn't 
work that way. You have yet to acknowledge that you understand why the math 
doesn't work that way.

2) You *can* use the statement "Most of the criminals have trait Y, hence we 
should *not* focus our question on people who *lack* trait Y."  Again, this 
is how the math works.  Again, you have yet to acknowledge that you 
understand this point.

This is a tremendously fundamental part of the whole "profiling" 
conversation. You waved it off with "don't nit-pick the math." This implied 
to me that you didn't understand why it's important.

It's like having a debate with someone who thinks "A implies B" implies "Not 
A implies Not B".  And then the correction is waved off with "Oh, you know 
what I mean."  No, really, I don't.

Basically, the problem is related to the rate of false negatives not telling 
you anything about the rate of false positives. Maybe that phrasing helps.

>   That's also fine, and I said that many times. I didn't claim that
> ethnicity is the *best* factor. I simply said that narrowing factors can
> be used to distribute resources better and increase the likelihood of
> success, and that *if* ethnicity were such a factor 

And I gave examples where ethnicity *can* be used that way (e.g., medicine), 
and explained why linking genetics to criminal behavior is *not* an 
effective means of profiling (because of (1) and (2) above).  Which, as far 
as I could tell, you didn't read or didn't understand.

> I even gave a simple example where this indeed gave a positive result.

Other than the rapist bit?  See (1) vs (2) above. You're not going to get 
many false negatives be eliminating female rapists. You're going to get a 
lot of false positives by considering all or random subsets of males as 
rapists.

I didn't see any example where you wound up with a positive result other 
than the rapist bit. Note that the stopping of SUV drivers has the same 
problem as (1). Your example of eliminating the bus drivers is an example of 
(2). The sparsity of false negatives in a group does not imply a sparsity of 
false positives in the compliment of that group.

> The example was not intended to depict a real situation, but as a simple
> demonstration of the basic principle of narrowing down the samples.
> 
>   The problem is that you kept sticking onto the "ethnicity" part and
> wouldn't understand what I was *really* saying.

Except when I tried substituting other examples with the same problem, such 
as profiling based on religion, you first ignored it, then called it 
outlandish. In this country, when you say someone's statement is 
"outlandish", it's generally of the connotation "What, are you crazy?"

Hence, it first looked like you were trying to ignore the example and 
continue the debate as if it hadn't been given (which is a mark of debating 
with someone unwilling to change their mind). Then it looked like you 
thought it was a terrible example because it *wasn't* based on ethnicity, 
and indeed the very thought of profiling based on something other than 
ethnicity was crazy talk.

Then you gave a bunch of examples like eye color and hair color, all of 
which are part of race/ethnicity/etc.

Then you got defensive when people were saying all your examples were 
racist.  (Not you, your examples.)

That's part of why I thought you were suggesting using ethnic profiling 
rather than some other type of profiling.


Here's a suggestion, tho, just in terms of personal-communication:

If you've made a mistake, and someone corrects you, don't berate them for 
correcting you.  If you give an example, and someone says "The math doesn't 
work that way," don't say "don't nit-pick the math."  Instead, say "sorry, 
that was a bad example, let me try again."

Otherwise, you're saying "You're clearly misinterpreting what I meant to 
say, because what I meant to say was something that's correct. Don't nitpick 
on the fact that what I actually said was wrong, but instead assume instead 
that I said something correct and go from there."

You sometimes get into this mode in a conversation and it makes it hard to 
get out, because you start thinking people are intentionally 
misunderstanding you.  You give bad examples, then complain when people use 
those examples even when you say "that's not what I meant." But you never 
actually express what you *do* mean, you just say "apply the appropriate 
corrections, then discuss that instead."

HTH.

And honestly, it was fun.  I learned a bunch how to express my thoughts on 
this subject. I hope I didn't offend you. That wasn't my intention. I don't 
think you're racist. I simply think in this case you either haven't thought 
through the math of the "statistics" you're proposing or you don't actually 
care enough about the topic (since you see it as irrelevant and it probably 
doesn't happen much around you) to really think about whether profiling 
really would work. I think because of this you might *sound* racist to some.

-- 
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