Educated Guesswork

Let's see if I've got this straight....in Texas, they're planning to require every car to have a special tag that uniquely identifies it, and that anyone can read from some distance away? Boy, I sure hope other states don't follow suit!

Seriously, though--if OCR hasn't reached the point yet where it can automatically read license plates, I assume it will pretty soon. Wouldn't an RFID system simply be a cheaper, more accurate way of doing the same thing? And isn't the whole point of license plates that your car is supposed to be uniquely identifiable when on the road in the first place?

Actually Dan, agencies are already using license plate OCR systems. For example:

http://forbesbest.com/business/smallbusiness/forbes/2005/0228/044.html

So Eric's point here is the same one he made a few months ago about the police using tracking devices. It's not about whether the technology gives agencies a completely new capability. It's about giving them the capability on a whole new scale.

The above OCR system costs $25K a piece and has line-of-sight limitations. This is a whole lot different than it being affordable for agencies to track every car going through every major intersection.

Kevin, $25K doesn't sound like a lot of money to me, and I'd expect the price to come down substantially if volume picks up. How much cheaper are RFID readers likely to be, anyway? Let's say they're a tenth the price. That's pretty significant from a saving-taxpayers'-dollars perspective, but if what you're really worried about is nefarious government "agencies [intending to] track every car going through every major intersection", then an OCR scanner at every tenth intersection doesn't sound much more comforting than an RFID reader at every intersection. And that's assuming that the government you're so afraid of won't simply jack up its spending to achieve the coverage it deems sufficient.

On the other hand, accuracy differences could have a much bigger practical effect. If RFID readers can cut the number of incorrect readings by, say, a factor of ten, then they can similarly reduce the number of innocent people mistakenly inconvenienced due to an incorrect license plate reading. I'd think you'd be all in favor of that--no?

25K is evidently quite a bit to police departments from reading my local newspaper.

In any case, if you accept my premise that scale is important, then I'd be happy to debate whether there is a scale change here or not.

But I'd guess that you're not willing to concede that point. So it hardly seems worth my time to put forth a specific analysis of the scale change here.

Just ran across this:

http://www.iautomate.com/r500sp.html

Long range RFID reader for ~400. So we're talking almost two orders of magnitude difference.

Kevin, I'm having trouble understanding your model of the government. What do you take their motives to be? What resources do you think they have at their disposal? What are the constraints on their behavior?

It's difficult for me to imagine a government that has the will and the resources to place RFID readers at every intersection, for simultaneous precise tracking of every car on the road--but not to to place OCR license plate readers at some fraction 1/n of intersections (for some reasonable value of n) for simultaneous approximate tracking of every car on the road. It's even harder for me to visualize this government caving more readily to pressure not to deploy RFIDs at all, than to pressure to limit their use in fairly obviously desirable ways. Finally, I find it hard to imagine such a government not reacting to the larger error rates of license plate OCR by harassing more innocent people than if they were allowed to use RFIDs.

Perhaps once I've understood how these things can be plausible, I'll be able to understand why different reasonable values of n actually make a difference.

Well Dan, I imagine it's more than my model of government that you're having trouble understanding. From our past discussions, you pretty clearly don't buy into decision and game theories nearly as much as I do in terms of how you personally behave.

Given decision and game theories, my objections to RFID over OCR are based on two additional beliefs: (1) the government often tries to exercise its power through coercion and (2) individuals in government use its resources for their own purposes.

With RFID's, the government's ability to track my whereabouts in a car may be 75% of the time with 90% accuracy. With OCR, it may be 10% of the time with 50% accuracy. (note that I'm just using these numbers as a strawman for analysis)

Now let's say that I want to engage in behaviors that are legal, but potentially embarassing. Or might be of interest to a particular person or group in government. The differences in probabilities dramatically alter the expected utilities and game structure.

My chances of being detected are lower. But also, the ability of my "opponents" government agencies and individuals using their resources is greatly degraded. They can't commit resources based on my detected actions because their certainty is too low. They can't threaten me with exposure because the detection methods aren't credible enough.

This all seems pretty obvious to me. Let's say I'm in a boxing match and the other guy wants to switch from padded gloves to padded gloves with an aluminum shell. Sure, I was willing to take 1/n damage before. But now the game is changed and I really don't want to play that game.

No, Kevin, I'm really quite happy to assume, for the sake of this discussion at least, that the government is a rational actor trying to maximize its benefit. My trouble really does begin with your model of the government's, goals, incentives, resources and limitations--as I already explained. (It's generally considered polite to actually pay attention to what someone is saying in these discussions, rather than simply inferring what their points must be based on your a priori notion of their beliefs.)

In fact, what "seems pretty obvious" to you is implicitly based on exactly the assumptions that I had trouble accepting:

Dan,

I'm surprised by your first point. Surely, you don't think that the government has unlimited cash flow and isn't at all cost sensitive. Yes, it's true that that money is extracted from citizens via taxes, but the citizens aren't happy about it and there's constant pressure to cut taxes.

Let me ask you this: the US Air Force would no doubt love to replace all of their old F-15s with F-22s--and the manufacturers would love to sell them to us. Why haven't they done so?

Eric, my thinking is as follows: constructing a vast RFID infrastructure of the kind Kevin describes is an expensive, difficult undertaking of unclear usefulness, that would no doubt face a lot of opposition. If the government is capable of getting it done, it would most likely be by invoking some powerfully compelling justification--protection against terrorism, perhaps, or massive improvements in crime control. In that case, I expect the cost would become a relatively minor issue in the debate over the clearly massive social implications, positive and negative, of having the system.

To use your example, getting all those F-15s in the first place was spectacularly expensive, and the Pentagon only managed it by invoking a compelling overarching need--the Cold War--that shifted the debate from, "can we get by with cheaper planes?", to, "do we need to purchase a large number of the very best planes we can build, almost irrespective of the cost?". If we're not upgrading to F-22's today, it's presumably because the contractors haven't convinced enough people that they're essential to the War on Terror. (I'd bet that when it comes to, say, the latest JDAMs, special forces gear and urban warfare technology, though, money is very close to no object.)

You always go ad hominem so quickly Dan. It's tiresome. I did read your comments carefully. I read them as, "If you're willing to put up with x, it's not a big deal to put up with 10x." I think that's a fair reading. And given our number of past interactions on this forum, while my surmise about your beliefs is a priori, it's certainly not unfounded given the evidence.

As for your specific points in your last post, I read about about police budgets being constrained all the time. But gee, for just $400 a pop, you could put them in every patrol car. You know, for the safety of the officers. So they can quickly determine if the people they stop are likely to be dangerous.

Of course, these patrol cars are wired to the network, so it's a simple matter of programming, to build a pretty good distributed surveillance net. And if you don't believe the government attempts to bootstrap measures sold to the public for one purpose into much broader powers, may I point your attention to the recent statistics on the use of Patriot Act warrants for non-terrorist related investigations.

As for your other two points, perhaps it was you that didn't read my post carefully. I thought it was clear that my concern is about credibility. I don't really care if the government threatens people more often, per se. I care if those threats are _credible_. Because if they aren't credible, the frequency will take care of itself. And low accuracy and coverage decreases the credibility. Game theory Dan. Which, if you believe in, you certainly don't use much in your analyses.

Kevin, since you don't make any substantial points in your comment, nor address any of mine, I'll happily return the favor--with one exception: I don't understand what you mean by "credible" threats. Threats of what? To whom?

....And (I meant to add) "credible" to whom?