FC: More on data mining, TIA, and how to ID terrorists

[Declan McCullagh <declan () well com>]

Other Politech messages:
http://www.politechbot.com/cgi-bin/politech.cgi?name=poindexter

---

From: "Werberg, Sam" <swerberg () findsvp com>
To: "'declan () well com'" <declan () well com>
Cc: Abraunberg () currentanalysis com, editor () kdnuggets com
Subject: RE: Data miner replies to Politech, says TIA can ID terrorists
Date: Tue, 10 Dec 2002 15:34:33 -0500

Having followed the data mining space, I agree that it has its usefulness, 
but the question that Gregory hints at but leaves unanswered is critical:

"How many times you will need to flip each coin to find at least one crooked
coin?"

In other words, how many law-abiding citizens will need to be "flipped", or 
have their lives turned over, in order to find the terrorist?  How many is 
too many?

Yes, the purpose of the TIA experiment is to find the answer to this, but 
considering that the person in charge of it (Poindexter) was "convicted of 
conspiracy, lying to Congress, defrauding the government, and destroying 
evidence", can we trust that any conclusions he provides will be valid or 
factual?

Sam Werberg

---

Date: Tue, 10 Dec 2002 14:10:05 -0700
From: "John W. Durham" <johnwdurham () acm org>
Subject: Re: FC: Data miner replies to Politech, says TIA can ID terrorists
To: declan () well com

Of course, the problem with Mr. Piatetsky-Shapiro's argument is that
there's no real assurance that the government will actually employ data
mining or will, instead, simply sift the "data" for whatever seems
expedient or interesting at the time of the search. Since "intelligence"
work is usually done under some time pressure, I find it hard to believe
that rigorous methodology will be used.

Remember, too, that we now have a government controlled by people who
argued successfully against the foundations of statistics in the court
cases relating to the use of sampling by the Census Bureau.

Not, of course, that politicians would ever have access to the TIA
program.

---

Please anonymously post my reply to Gregory Piatetsky-Shapiro.


xxx

Yes, this is an oversimplification!

Let's start with the fact that there are 280 million people, not a
thousand. Two out of a thousand turns out to be 560,000 bad guys, they say
we are only looking for 3000 or so, but how would they KNOW? And what are
these bits of information anyway? They are subjective things like cash
withdrawals, credit card purchases, not Boolean bits, and you want a
million transactions for 280 million people? Oh, and you need to do all
this in real time if you want to catch "pre-crime." Oh, you infer they can
catch every one? Ha!

It seems that the FBI et al had enough information so that a lowly agent
had the foresight to predict that a plane might be used to crash into a
building. Is there going to be any bureaucrat control in the TIA, or will
it be run like the Gestapo?

In this group there was a report that face recognition cameras couldn't
accurately spot the employees at a test airport using willing test
subjects. What happens when most of the evidence is analog? What happens
when the "bad guys" never committed any crime? If pre-perp prays to Allah 3
times a day for 50 years would he get 55,000 bad guy demerits? If he reads
Arabian nights instead of Mien Kompf does he get a Big Brother knock on he
door?

How can the government not sift through -everyone's- personal data
real-time to find the alleged bad guy, who has never done anything bad, and
how do you KNOW he/they were going to do something anyway? If someone has
too many matches is he an arsonist, too many guns a revolutionary? What
programmers are going to write this insightful program to predict crime
that hasn't happened unless they know the bad guy's crime already. No you
are going to spy on farmers who buy too much fertilizer and people who
decide to turn in their credit cards and withdraw an unusual amount of cash
because they are not going to participate in the sacking of the Bill of
Rights.

I am afraid if this continues we will have a default tyranny, one that is
capable of doing bad things even if it hasn't done it yet, does that sound
familiar? Thomas Jefferson said its our DUTY to overthrow tyranny. But what
if we can't because every bit of communication data is linked and
scrutinized so WE THE PEOPLE would be completely unable to form peaceful
protests out of fear, or effective rebellion because of the sheer power of
the state's spy network?

I think you data miners are dangerous. Does that get me put on the watch
list? If you successfully create this universal spy network, the result
will be the creation of a totalitarian ant pile. In the meantime, you will
be creating the terrorists you are looking for, only once upon a time we
called them patriots.

---

Date: Tue, 10 Dec 2002 15:33:49 -0500
From: jeff <jeff () newcitynet com>
Reply-To: jeff () newcity com
Organization: newcity
To: declan () well com
Subject: Re: FC: Data miner replies to Politech, says TIA can ID terrorists

I am happy to read the reasoned response below by Ben Brunk and Gregory
Piatetsky-Shapiro

There are 2 issues outside of its scope:

1. TIA used for partisan purposes, such as investigating adulturous
politicians of only one party, or to intimidate and track small
environmental organisations who are opposing an oil well project owned
by a close political ally of an authority.

2.  a cash-only lifestyle.... if they're smart terrorists are
stockpiling stuff now.... thus rendering a false sense of security and a
maginot-line-mentality among authorities.

3. Industrial intelligence gathering... patterns may emerge from sifting
through data that might benefit company X to know about the
import/export/financial records of foreign supplier Y so that company X
can push harder for a price reduction, etc.... this sort of info might
be tempting for authorities to supply to politically connected
businesses in return for lots of support.

-Jeff C
chicago

---

To: declan () well com
Cc: brunkb () ils unc edu
Date: Mon, 9 Dec 2002 22:44:41 +0000
Subject: Re: FC: What's so bad about Total Information Awareness? by Ben Brunk

What Mr. Brunk doesn't make explicit, but
probably knows, is that TIA is simply an excuse
to get their hands on every database in the world.
The people who control the operation will simply
look up what they want about whoever they have
predetermined is suspect, according to their
political agenda.

--GJ

---

From: Amos Satterlee <asatterlee () inta org>
To: "'declan () well com'" <declan () well com>
Subject: RE: Data miner replies to Politech, says TIA can ID terrorists
Date: Tue, 10 Dec 2002 17:18:38 -0500

Declan:

Fascinating point/counterpoint. I think, however, that Brunk's point central
concern is not addressed by Piatetsky-Shapiro, to wit:

1. Intelligent analysis of business behavior or buying patterns occurs
within reasonably defineable contexts. TIA can't limit itself to a
particular context because the terrorist then acts outside the context.

2. Piatetsky-Shapiro is correct that increasing the number of trials reduces
the instance of false positives. Brunks point is that the terrorist acts in
such a way to try and reduce the number of trials of a particular pattern of
behavior.

3. If there is no context other than Everything, Brunk's point is that the
time, effort, manpower, and invasiveness needed to analyze the data that
intends to be unpatterned is so high as to be unacceptable and the
unintended consequences are shocking.

Amos

---

Date: Tue, 10 Dec 2002 17:42:30 -0500
To: declan () well com
From: Stephen Cobb <scobb () cobb com>
Subject: Re: FC: What's so bad about Total Information Awareness? by
  Ben Brunk
In-Reply-To: <5.1.1.6.0.20021209235549.0161eec0 () mail well com>

At 12/9/2002 11:57 PM -0500, you wrote:
> Date: Mon, 09 Dec 2002 22:34:13 -0500
> From: Ben Brunk <brunkb () ils unc edu>
> To: declan () well com
> Subject: Debunking TIA
>
> My hope is that this editorial will awaken those who are even more skilled 
> in computer science, statistics, game theory, etc. and that they find the 
> courage to speak up so we can put the brakes on the wasteful and 
> destructive blind alley called TIA.
>
> Benjamin Brunk

Back when Safire wrote his column on TIA my comment was "The reality is 
that no government agency could possibly link more than two databases in 
under ten years for just $200 million." (11/14/02)

Maybe I should have said it louder, with more credentials :-)

Stephen

---

Date: Tue, 10 Dec 2002 19:45:49 -0500
From: "J.D. Abolins" <jda-ir () njcc com>
Subject: Re: FC: What's so bad about Total Information Awareness? by 
Ben        Brunk
In-reply-to: <5.1.1.6.0.20021209235549.0161eec0 () mail well com>
To: declan () well com

On Mon, 2002-12-09 at 23:57, Declan McCullagh wrote:
> Date: Mon, 09 Dec 2002 22:34:13 -0500
> From: Ben Brunk <brunkb () ils unc edu>
> To: declan () well com
> Subject: Debunking TIA
>
[...]
> All in all, I can't see how TIA will do anything except harm innocent
> people and create new jobs for bureaucrats.  Any numerate person who spends
> five minutes thinking about what is proposed will come to the same
> conclusion.  If our system is going to become this arbitrary, there are
> going to be an awful lot of lives ruined in this country.

Exactly so!

If TIA works in linking patterns to some useful, the errors will be bad
news. If TIA doesn't work to any useful anti-terrorism purposes, it will
be bad.

Whether or not a system like TIA works is not as critical as the belief
by its users that it works.

Recently, I got a good laugh from a Wall Street Journal article, "My
TiVo Thinks I'm Gay, How to Set It Straight." The article described the
grossly mistaken customer preferences generated by TiVo and Amazon.com
routines. (Amazon.com's customer suggestions pegged a gay man as a
"pregnant gay man" after he bought some books on pregenancy to give to a
friend. What I'm not laughing about is the prospect that TIA thinks I'm
a Middles Eastern terrorist if I bought some couscous and browsed some
Middle Eastern Web sites. <grin> Then I'd have to buy some canned hams
(which Iwouldn't eat but give to a soup kitchen) to adjust the profile.
But it is likely that somebody with TIA development would try to catch
TIA hacking attempts and peg eractic profile changes as suspicious.
<grin and groan>

In real life, however, I believe that likely government response to the
occaisional rousting of innocent people, perhaps with soem rough
treatment, is "They should thank us for being so diligent in following
up possible terrorism leads instead of grumbling about the SWAT fellow's
boot on the back of their necks."

I hope to come to back to Mr. Burke's good comments about the
complexities and difficulties of mining data in a latter email. Even
though there are many problems, the perception (a shakey one) is still
being sold as a great business tool.

J.D. Abolins

---

Date: Tue, 10 Dec 2002 22:33:11 -0500
To: declan () well com
From: Stephen Cobb <scobb () cobb com>
Subject: Re: FC: Data miner replies to Politech, says TIA can ID
  terrorists
In-Reply-To: <5.1.1.6.0.20021210150041.021d9048 () mail well com>

I'm sorry but this is a laughable defense. Personally, I see a lot of 
technical problems with TIA long before you get to the statistics, but if 
TiVo is any indication of the technology involved, we are in deep trouble...

"If TiVo Thinks You Are Gay, Here's How to Set It Straight
What You Buy Affects Recommendations
On Amazon.com, Too; Why the Cartoons?
By JEFFREY ZASLOW
Staff Reporter of THE WALL STREET JOURNAL
http://online.wsj.com/article/0,,SB1038261936872356908,00.html
Basil Iwanyk is not a neo-Nazi. Lukas Karlsson isn't a shadowy stalker.
David S. Cohen is not Korean.

But all of them live with a machine that seems intent on giving them
such labels. It's their TiVo, the digital videorecorder that records
some programs it just assumes its owner will like, based on shows the
viewer has chosen to record."

Stephen





-------------------------------------------------------------------------
POLITECH -- Declan McCullagh's politics and technology mailing list
You may redistribute this message freely if you include this notice.
To subscribe to Politech: http://www.politechbot.com/info/subscribe.html
This message is archived at http://www.politechbot.com/
Declan McCullagh's photographs are at http://www.mccullagh.org/
-------------------------------------------------------------------------
Like Politech? Make a donation here: http://www.politechbot.com/donate/
Recent CNET News.com articles: http://news.search.com/search?q=declan
-------------------------------------------------------------------------