[ExI] Terrorist? Who can tell?

[Lee Corbin]

Damien had written

> At 08:16 AM 8/29/2008 +0000, BillK wrote:

> > If you go into the large areas of London where mostly non-white people
> > live, you will see about about 3 million people that look like Al
> > Qaeda members to the profiling eye. This information is totally
> > useless.
> 
> This is so self-evident I can't imagine what Lee is on about. Unless--

Apologies to all. I misread Damien's remark, sowing a great deal
of needless confusion. Evidently no one has made any concession
to my newer claim

> > Hence my guess that at the 95% level of confidence,
> > six such individuals would be chosen as "the probable
> > convicts" and the other six average Londoners of the
> > same age and sex chosen as the "non-terrorist" group.
> > And only 5% of the time would the group of Londoners
> > picked at random---which surely includes a very wide
> > variety of people, nothing so nearly homogeneous
> > (I thought) as the bombers' group---be identified as
> > "the terrorists" by an ordinary set of citizens seeing each
> > group in a line-up.

I have been assured by several people here that it is unlikely
that this could be done. Yet we are proffered the statistic
that approximately 40% of London is white, perhaps another,
what?, ten percent black, and what?, yet another sizable
percentage who are Asian? (See my further explication of
my proposed experiment below, to attempt to remove any
last uncertainty as to what I've been asking.)

BillK writes

> The problem mentioned in the original post is 'How do we detect Al
> Qaeda terrorists?'

Yes, but I was proposing the above quoted question in order to
clarify my understanding. (I do thank you, Bill, for your patience
here, given my muddying of waters by having misread Damien's
post.)

On the literal reading of my question above, which was my intention
and which has no necessary direct bearing on profiling, does a six
member *group* of known members of Al-Qaeda differ significantly
in appearance from a random sample of six age-appropriate men
chosen at random from London? Sorry, but I still don't understand
how you and Damien can persist in saying that the answer is "no".
For one thing, aren't the odds of anyone who is Asian or black 
being in the UK component of Al Qaeda quite small? If so, then
the presence of such an individual among the six Londoners chosen
at random would be a very strong giveaway that this was the group B.
This is, of course, not to be read as a claim that there are *no*
Asian or black members of Al Qaeda, but just that the percentages
must be very very small---or so I believe.

I tried to re-explain as follows, in order to dispel confusion:

  > > I wonder if you understood the actual experiment that I
  > > proposed. It was *not* in any way a kind of actual test
  > > that could be put into practice. Quite literally, it worked
  > > as follows. We take for group A the last six convicted
  > > terrorist bombers in London, and as a control group B
  > > take six males of the same age randomly from the city.

  > > In short, you are claiming that the relative homogeneity of my
  > > group A (or at least I thought so), would be insufficient to allow
  > > for a 19 in 20 test.  Now if we *raise* the number from my
  > > "six" to "twenty" (provided that there have been that many
  > > Middle Eastern terrorist bombers so convicted), then the
  > > laws of mathematical probability insist that group A and group
  > > B would become more and more reliably distinguishable.
  > > You don't agree?

So *that's* the question on the table for me (except that I was
forced to revise the statement to Al Qaeda because of Tom's
comments) because if I'm wrong then my knowledge of London
is even worse than I thought, and perhaps even my understanding
of probability and sampling distributions.

> The problem mentioned in the original post is 'How do we detect Al
> Qaeda terrorists?'
> To say that they are more likely to come from the 40% non-white
> section of the London population (about 3 million people) does not
> help in the slightest to detect the tiny number of terrorists among
> them.

I concede that point again, entirely, so long as it is acknowledged
that indeed they are more likely to come from the 40% non-white
proportion of the three million London population.

> Especially as there are a few white converts among them also.

Can you guess what fraction this might be? If it's at all high,
then this is the source of my confusion. But if it's on the order
of a few percent, then your remark seems irrelevant to *my*
question, regardless of how useful it may or may not be to
those attempting the (evidently hopeless) task of profiling
those who, say, board aircraft.  "A few" doesn't help me:
a few in a group of less than 10 is a lot, but "a few" in a
hundred is not.

> This type of profiling is useless as a guide to using police resources.

I cannot yet quite agree with the broadness of this statement;
you would need to qualify it to certain airports and certain
groups, I think. For example, if the Bangkok police were
requested to locate a Mafia member who'd been born in
Sicily, raised in the U.S., and was known to be in a town
located one hundred miles from Bangkok which is almost
wholly native Thai, wouldn't it be quite idiotic not to profile?

> Victimizing this 40% of the population would be counter-productive.
> Apart from all the civil rights issues, (non-white people are citizens
> also, you know)

Yes, I have been given to understand that this is true:
a number of non-white citizens actually are citizens.
I can even testify that it's true in the United States.

> the police are dependent on co-operation and informants from
> within this group.

Yes.

> Even if we did have a really good profiling method, this would just
> lead to the terrorists changing their procedures to avoid the profile.
> (They're not stupid, some have degrees!).

It might not be so easy; e.g., someone born in Sicily who's in a small
Thai city somewhere.

> The only method in use at present in the US and UK is to treat
> everyone as a possible suspect.
> Cases like the 5-year old who gets hassled because his name is on the
> TSA watch list database.

Could you further elaborate on why this makes any sense? (I do understand
if I've exhausted your patience, but I'd appreciate it if you would, thanks.)
Of course, for airline safety, the carry-ons of all five-year-olds must be
checked, because one could wittingly or not be part of a plan to smuggle
weapons aboard. But why would it not make sense to instantly perceive
that the five-year-old could not *possibly* be the sought-after terrorist?
To get hung up on the coincidence of names actually sounds quite stupid,
and obsessively literal-minded, (provided that the kid's name really is so-and-so).

Lee