When Algorithmic Surveillance Becomes Structural Surveillance

A recent Wall Street Journal article revealed an ongoing DEA program designed to use automated license plate readers to build a real-time database of vehicle movement throughout the United States. This database would also store this data, giving government agencies the ability to retroactively track the movements of millions of Americans. As we've seen with similar programs, the scope of the effort will inevitably begin to creep.

The effort began in border states like Arizona, California, Nevada, New Mexico and Texas, but the goal has always been expansion, according to current and former federal officials and documents. Officials wouldn’t say how many other states are now feeding data into the system, citing concerns that disclosing such information could help criminals avoid detection.

The problem with these electronic dragnets is the enormous efficiency gained by putting this information into a digital format, enabling algorithms to rapidly search and sort the data based on whatever criteria the law enforcement agency happens to be interested in, including, for example, asset seizure.

One email written in 2010 said the primary purpose of the program was asset forfeiture—a controversial practice in which law-enforcement agencies seize cars, cash and other valuables from suspected criminals. The practice is increasingly coming under attack because of instances when law-enforcement officers take such assets without evidence of a crime.

But, one might ask, isn't increased efficiency a good thing? In many cases, this would be true, but consider Justice Sotomayor's warning in her United States v. Jones concurring opinion:

And because [advanced tracking technology] is cheap in comparison to conventional surveillance techniques and, by design, proceeds surreptitiously, it evades the ordinary checks that constrain abusive law enforcement practices: limited police resources and community hostility.

(citations omitted). Frank Pasquale's recent work (and excellent book) addresses some of these same concerns--what happens when algorithms, often opaque to ordinary citizens, become the de facto method of law enforcement surveillance?

Law enforcement agencies point out that, through this program, they have been able to significantly enhance their ability to solve crimes.

In 2010, the DEA said in internal documents that the database aided in the seizure of 98 kilograms of cocaine, 8,336 kilograms of marijuana and the collection of $866,380. It also has been connected to the Amber Alert system, to help authorities find abducted children, according to people familiar with the program.

Crime prevention is certainly a potential benefit of such a program. If we take this as an overriding factor, one option would be to simply allow these techniques and technologies to become just another part of our growing structural surveillance. But are we willing to bet against future abuses of such a system, and its effects on our society, on this factor alone?