Muddled Thinking About Universal GPS Tracking

I knew as soon as I read George Washington University Professor Amitai Etzioni’s ridiculous bootlicking CNN op-ed (via a tweet from Radley Balko) that I wanted to write something about it. As often happens, Scott Greenfield beat me to it with a biting response to the legal issues. Still, let me give you a taste:

The Supreme Court is about to hold hearings on whether the police need a warrant to attach a GPS tracker to a suspect’s car and trace its movements while it is in a public space…

The intense debate the case has already elicited among legal scholars, civil rights and libertarian activists, and those particularly concerned with public safety and national security is largely focused on the question: what would the Founding Fathers have said about the case? As I see it, at least equal weight should be accorded to the question: How well are our public authorities doing in their dealings with criminals?

Yes, because police expediency is such an excellant way to decide what our rights should be. In fact, it would really speed things along for our overworked police forces if young black males would be so courteous as to report to the nearest prison when they turn 16. Thank you for your cooperation.

Basically, Etzioni seems willing to give police whatever power they want, as long as they say they need it to protect us. As is usually the case when someone makes this argument, Etzioni seems unwilling to consider that giving police this power might come with a significant cost, that the cost might include significant abuse by the police, and that the giving the police this power might not actually make us safer.

It’s a common enough way of thinking. But what got my attention is Etzioni’s depressing misuse of science, statistics, and logic.

According to national statistics for 2010, less than half (47%) of violent crimes committed in this country are “cleared” (that is, suspects are arrested, charged, and turned over for prosecution) and only one out of five (18%) criminals who commit nonviolent crimes (such as burglary) are caught and tried.

Etzioni offers no context for these numbers. Is a 47% solve rate bad? Or is it an all-time high? Crime rates have been falling since the 1990s and if clearance rates are not unusually low, it’s not clear that we need to grant police such a broad privacy-destroying power.

For obvious reasons there are no such statistics available for terrorists, and the fact that there was no successful attack in the U.S. over the past 10 years tends to make us complacent.

However, if one takes into account that there are many millions of people in the world who hate us and wish us harm (and at least a few right here in the U.S.), we should maintain our vigilance. As one terrorist group once put it, “You have to be lucky all the time. We only have to be lucky once.”

Etzioni likes the statistics when they are in his favor (the low clearance numbers, which imply a large number of criminals roaming around) but when the statistics cut against his argument, as with the low terrorism rate, his response is to downplay the statistics and warn us not to get complacent.

I get the feeling this is due more to intellectual laziness than dishonesty, since Etzioni is actually wrong on the facts when he asserts that there has been “no successful attack in the U.S. over the past 10 years.” Thirteen people died in the shooting attack at Fort Hood in 2009, and twelve more people have been killed in smaller incidents that are considered terrorism. (Five more died from the anthrax mailings after 9/11, but that’s technically just outside the 10-year boundary.) It averages out to about three domestic terrorism deaths per year.

As to what is reasonable, it obviously changes with the circumstances. Given that criminals can use freely all the new technologies — including of course GPS trackers, smartphones and spyware — it seems eminently reasonable that the police should also be able to use some of these, especially in public spaces, in which people have no expectation of privacy (or at least should not have one).

Huh? Criminals also smoke crack. Does that mean we should let the cops smoke crack?

The proper question is not whether police have the same technology the criminals have, it’s whether Police have the right technology to do the job. For example, cops have been arguing for years that they need better weapons because the criminals have better weapons. That makes some sense, because one of the best ways to counter an opponent’s weaponry is with weaponry of your own. But it’s not clear that criminals are doing anything with GPS systems that the police could counter with GPS systems. What’s the logic for demanding GPS parity? This seems like lazy thinking.

Moreover, often some such surveillance is needed before a tip or lead can be developed to the point that it meets the standard of probable cause.

The counter-argument that if the police are allowed to proceed, we shall be all tracked seven days a week, round the clock does not withstand minimal criticism. At most, the GPS data tells us that someone drove a car to certain places. Who lives or works there, what happened, etc. etc., all remain to be investigated.

Again, Etzioni is speaking out of both sides of his mouth. If we don’t need to worry about cops having our GPS data because it’s so useless, then why do cops want it so badly? The obvious answer is that knowing where a person travels in a day or a month tells us a lot about that person. That’s why the cops want it.

And then Etzioni says something really stupid:

If the police put GPS devices in all the cars on the road, or even only in one out of every thousand, cops would be buried under an endless flood of data points — among which suspects would be lost.

Um…no. Not at all. Not even a little bit. This one really bothers me, because I’ve encountered this misconception before and it’s seriously wrong. Etzioni is trying to sound smart by talking about things he doesn’t understand. Modern internet-scale computing routinely solves these kinds of problems.

I got curious and did some of the math. Assuming we can store GPS latitude, longitude, altitude, timestamp, and transmitter ID in five 64-bit fields, and that we want to get a GPS fix every 10 seconds for all 309 million Americans, that works out to about 99,000 gigabytes per day, or 3 million gigabytes per month, equivalent to the data bandwidth used by 11 million iPhones. Using prices from Amazon Web Services, I estimated that it will cost $150,000 per month to transfer the data, and if we to retain the data for a year, that will cost 36.3 million gigabytes, the storage rental will eat another $2 million per month.

Computing cost is harder to estimate because it depends on what sort of data processing we’re planning, but assuming that a large Amazon server instance can process 1000 samples per second, we’ll need 30,900 running server instances, which will cost us $8.9 million per month. Now that’s a heck of a lot of servers, but it’s not unheard of. Large sites like Yahoo and Facebook are probably using that many servers, and Google is widely estimated to operate 900,000 servers.

Retrieving the full year of data for any single person would probably take a minute or two (mostly transfer time) and with that many servers, we could retrieve tens of thousands of records simultaneously. If you want to be able to answer geographic questions  — “Who was within 500 yards of the crime scene on the day in question?” — you’d probably need twice as much storage for the geospatial index. On the other hand, you could also save space by eliminating duplicate GPS data from when people aren’t moving. Call it a wash.

There are lots of other costs, including programming and management, but I think we can safely say that the cost of handing the data would be less than $150 million dollars per year. And if we cut our sample rate from every 10 seconds to every minute, the cost scales down to only $25 million per year. The government could contract this out, and depending how you classified the work, it might be considered a small business.

The thing is, my datacenter calculations are just a diversion, because if Etzioni had given it a little thought, he’d realize we already have a working example. When you call a mobile phone, the cellular network has to know which antenna to use to connect to it, and since the network can’t know in advance when the next call will come, it has to be constantly tracking every phone. So at this very moment our national cellular phone networks are using 250,000 cellular antenna sites to track over 300 million mobile phones in real time. It’s not as accurate as GPS, but it’s just as much data, and it’s fast, cheap, and reliable.

At the same time, the police should be required to file reports after the fact about their use of GPS trackers. If it turns out that they are employed too often or to track people who are, say, political activists, the police should be reprimanded and if they persist, elected officials (say, a city council) should set limits on the use of this and other crime-fighting technologies and punish those who abuse them.

Or, we could set the limits and punish the abusers right now. Keep in mind, the argument isn’t about whether the police should be able to track bad guys with GPS, it’s only about whether they should be able to do so without a signed warrant from a judge.

If the data ran the other way — if most criminals were neutralized and we did not have to be concerned about terrorists — reasonable people who seek to deny the police the use of GPS trackers without probable cause might have a much stronger case.

“If the data ran the other way”? What data is he talking about? There is no data. The only data he presents is the low clearance rate. He offers no evidence that any part of this low clearance rate is due to the inability of police to use GPS trackers without a warrant. He simply takes the police at their word that this is a tool they need.

Personally, I’m against this kind of big-brother system even if it would produce a substantial increase in clearance rates, because I think the government would use it to do more harm than good. But even if you don’t agree with me, don’t you think we should make the police prove they need this power before we give it to them? Make them gather statistics about how many bad guys got away without GPS tracking that would have been caught with it. Make them show us the math. My guess is, they can’t.