This article attempts to show what a local government could do, but does anyone have access to the sort of location dataset that could give one an intuition about how likely it would be that individuals would be singled out using the data shown?
These cops appear to arrest 23 people per month for driving under the influence, and hand out about 1750 other traffic citations (not linking directly to the PDF to save their server but you can find it under crime stats).
The article implies that there would be more traffic stops due to increased DUI suspicion, and it certainly seems that it could happen, but given these population, police, and police activity numbers, and given that the article itself gives a false-positive example, how likely would that be? Is it reasonable to think that these cops currently have a lighter load and have time to be dispatched to investigate a potential DUI?
I am not suggesting that it is of no concern for the government to have unfettered access to data, and I can imagine a vast number of possible scenarios in which the data could be misused, but possibly we can better quantify that concern.
The capacity to do selective enforcement is a huge issue in itself.
Let's say you bring corruption charges to light, or advocate for something people in power don't like, or even just cut off someone's angry nephew in traffic.
In a country with a rule of law, that's okay --- you're safe. You can't just be arbitrarily punished.
But with this? No problem. They just tell the NSA to flag you, and as soon as you do _anything_, they'll get you. Better still if the data's retroactive --- if they can track everyone, and figure out that three years ago you caught a fish and transported it across state lines or whatever.
If you have a legal system that classes a majority of people as technically a criminal, and the capacity to selectively enforce that, then congratulations, you've got despotism.
You're assuming that individuals would have to be singled out for analysis in the first place. They wouldn't. With enough computing power, it'd be trivial to run this sort of analysis in near real-time for everyone. A few years down the line, this is even less of a concern: increased power, improved analysis tools, and greater efficiencies will all come together to empower just such a system at costs manageable by many local governments (particularly if they start working together to manage costs).
This means you've got a practical means of confidently predicting potential crime. Even better, the same mechanisms that help you predict potential criminal acts also help you close active cases more efficiently. That inevitably results in significant boosts to police productivity.
So armed, you now get to deploy your existing resources more effectively. Instead of parking a patrol car at a "pretty good spot" to nail violators for a few hours, you'll reposition yourself continuously to the location offering the greatest probability of an arrest/citation at that specific moment in time. Instead of passively waiting for violators to happen across your location, you're actively positioning yourself to nail them directly. The specific individual doesn't matter.
But you won't be dispatching patrollers in the current manner. Since you'll also be tracking your own patrol cars, you can correlate the two and dispatch each car to the highest probability incident nearest to them where they're likely to catch a potential perpetrator. By having everyone under effective surveillance, for every minute of every day, each of your resources will always be deployed in the most efficient manner possible.
Where it really gets wild is when you start to consider traffic enforcement automation along with the push for local government drone coverage. Along with the existing red light and speed cameras, you'll have mobile platforms capable of filling in for any holes in your patrol car coverage while also benefitting from the same efficiency gains described earlier. And they'll be gathering additional data (movements, facial recognition, automatic number plate recognition, etc.) to send back to base while they're doing it.
It won't be like The Machine from Person of Interest, but all things considered, it'll be closer to it than anything else to date.
The article uses maps of Peoria, IL, so let's assume we're dealing with Peoria. There appear to be roughly 115K people between 18-64 in its metro area (http://www.wolframalpha.com/input/?i=population+of+peoria+il...), and 11 traffic cops. (http://www.peoriagov.org/peoria-police-department/police-div...)
These cops appear to arrest 23 people per month for driving under the influence, and hand out about 1750 other traffic citations (not linking directly to the PDF to save their server but you can find it under crime stats).
The article implies that there would be more traffic stops due to increased DUI suspicion, and it certainly seems that it could happen, but given these population, police, and police activity numbers, and given that the article itself gives a false-positive example, how likely would that be? Is it reasonable to think that these cops currently have a lighter load and have time to be dispatched to investigate a potential DUI?
I am not suggesting that it is of no concern for the government to have unfettered access to data, and I can imagine a vast number of possible scenarios in which the data could be misused, but possibly we can better quantify that concern.