[phurley]

I have a real issue with the current NSA practices and with much of the government surveillance attitude, that said this generally sounds like a case of doing it right.

I like that there was a court order for specific cell towers near the robberies, it would be better if order specifically requested the intersection of the towers or in someway restricted further use of the information to prevent long term and or continued use of collected data for unrelated cases, but I do not find this to be egregious.

This was a case of law enforcement using reasonable tools to do their job and a judge was in the loop -- I think the there should be a requirement to show probable cause, but a I think that a judge should be able to see that in this case there was probable cause to believe that anyone who had been at X number of the banks at the time of the robberies was a probable suspect and the collection of the data would be warranted.

My issue is when law enforcement (or other 3 letter agencies) collect (and retain) all of this data (and email, and license plates, and security cameras...) for non-specific uses. This is a significant infringement on our personal privacy and a threat to our democracy.

[nkrumm]

Exactly. The phrase "probable cause" can (and should) be directly related to the probability of a false positive; in the context of virtually unlimited 'data dumps', this paradigm must be adapted to reflect the probability of a false positive in a multiple-testing situation.

The FBI in this case was successful because they had 4 (or 3) towers; with only two it is likely they would not have been able to hone in on the criminals as quickly as they did. It's perhaps even possible to calculate what the chance of success is, and establish guidelines for release of data based on this (e.g., minimum 4 towers). Also, probability would have to be taught in law schools.

[rayiner]

> Also, probability would have to be taught in law schools.

The first couple of weeks of my Evidence class was a crash course in probability and Bayes' rule. Our final exam had an entire section on it. I'd imagine you can still get through law school without learning probability, but it's getting increasingly harder as Evidence is reformulated on probabilistic grounds, just as Torts was reformulated on economics grounds in the 1960-1970's. Today you probably can't get through a Torts class without learning about Pareto efficiency or Coase theorem, etc. Also, empiricism is a dominant academic trend right now in legal academia, and it heavily relies on statistical approaches.

[hga]

As long as you're talking about an investigation to develop leads, if the investigators have a clue false positives aren't necessarily going to be bad. In this investigation, they didn't just go arrest the people associated with the two numbers they found, they got the metadata for the two numbers and saw enough additional correlations they could be pretty sure they'd found their men, especially with the added highly suspicious behavior. Presumably a few false positives wouldn't have panned out after their metadata was examined, e.g. in this one's second stage they went from 3 out of 4 to "most of the 16 the bank robberies under investigation".