Our Modern Surveillance Society
While the prevalence of surveillance cameras in metropolitan areas continues to increase, the technology does not always deliver on its promise to prevent crime. As law enforcement agencies rethink the use of video surveillance, societies must assess the trade off between increased security and decreased civil liberty.
They’re Watching
In cities across the world, governments are investing heavily in Closed Circuit Television (CCTV), a form of video surveillance that transmits a 
digital feed to a limited set of monitors. Most recently with the emergence of the cloud, companies offer video surveillance as a service (VSaaS), whereby a user can automatically set and adjust capabilities, access footage via multiple devices, and rapidly scale video capabilities. [1] As the price of video surveillance decreases, the prevalence of surveillance cameras has increased dramatically. While London with estimates of over 1.9 million installed cameras ranks as the most surveilled city, other metropolises across the globe are investing heavily in video surveillance. Worldwide revenue from sale of video surveillance technology reached $15 billion in 2014 and is expected to increase up to over $23.6 billion by 2018.[2]
Video Surveillance and the Prevalence of Crime
Most cities laud the surveillance technology as a way to prevent crime and increase public safety. While the presence of a camera in a parking lot or 
on public transportation may deter a potential criminal from committing a crime, the cameras have no effect on experienced offenders. Despite the ability to store and compare vast quantities of video footage, after a crime has been committed, facial recognition software is not yet capable of dissecting video footage and identifying possible criminals.[3] In order for facial recognition software to ever work, a computer must be equipped with a database of faces against which to compare video footage. Many worry that such a catalog would infringe on privacy rights of law abiding citizens. While law enforcement organizations can use mug shots of past perpetrators, facial recognition is not yet advanced enough to match a clear image to the grainy video footage of a CCTV camera that is taken from an angle. The FBI’s Next Generation Identification (NGI) system, only guarantees an 85% chance of returning a suspect’s name on a ranked list of 50 names after viewing a picture of a suspect’s face, and that will only happen if the suspect’s name is already in the database. Even amongst software developers, many believe that facial recognition software will never deliver “near human” recognition capabilities.[4]
Working around Technological Limitations
Recognizing these limitations, some law enforcement agencies have developed processes that pair their digital libraries with the power of the human brain. The London Metropolitan Police Service created a specialized force of “super-recognizers” to work with computer outputs in order to identify criminals. Officers at Scotland Yard developed this unit after hearing about research from a Harvard postdoctoral student studying prosopagnosia, a condition where a person is unable to recognize human faces. The research showed that humans are capable of a spectrum of recognition ability. While a person with prosopagnosia cannot even recognize himself, super-recognizers can remember the face of a person with whom they have minimally interacted. To become a member of the Met super-recognizer unit, an individual must pass a series of recognition tests.[5]
To identify a criminal, first, software determines the images that contain faces within a stream of video footage. Normal police officers then enter these images of the perpetrator into a database and tag the images with categorical metadata such as age, race/ethnicity, type of crime, and other physical characteristics like tattoos, missing teeth, etc. The database also includes images and mug shots of known criminals tagged with metadata. Using the computer’s algorithm, super-recognizers narrow the number of images in the database to compare against the video footage. They then can determine which images in the database match the perpetrator. In 2015, this unit was able to solve 2,500 crimes using imagery. This method of identification has also proven far less expensive and more accurate than traditional identification techniques like DNA and finger print matching. [6]
Ethical Implications
While use cases, like London’s team of super-recognizers, prove that video surveillance can effectively improve security, there are still many ethical quandaries that need further exploration. Governments must balance civil liberty and safety. As evidenced by the revelations of Edward Snowden, most law-abiding citizens do not feel comfortable submitting to government surveillance. As technology becomes more advanced, societies must constantly assess the freedoms they forsake by living in a digital world.
Word Count: 799
[1]“A Detailed Comparison of VSaaS vs. Internet-Connected Traditional Security Camera Systems”, Dean Drako. Eagle Eye Networks.
[2] “Here’s Looking at You: An Evaluation of Public CCTV Cameras and Their Effects on Crime and Disorder”, McLean, Sarah J.; Worden, Robert E.; Kim, MoonSun. Criminal Justice Review.
[3] “From Drones to Body Cams, Tech is Changing the Fight Against Crime”, Simon Hill, Digital Trends.
[4] “Why Facebook is Beating the FBI at Facial Recognition”, Russell Brandom, The Verge.
[5] “The Super Recognisers of Scotland Yard”, Xan Rice, New Statesman.
[6] “The Detectives who Never Forget a Face”, Patrick Raden Keefe, The New Yorker.
Thanks Ellyn for an interesting discussion on the benefits and limits of using video surveillance technology for recognizing criminals. Although this is an ethically contentious topic, I have no issues with it if it means that criminals will be caught, and perhaps even dissuaded from committing a crime in the first place. However, I question how useful this technology is in the case of a masked criminal? There would have to be so many cameras installed so as to provide the ability to trace the criminal back in video space-time to a point when he/she was not masked. I wonder how committed the government is to providing the funds for installing the infrastructure necessary to make this technology effective in different types of cases.
Also, I thought it was interesting that humans are still necessary to help identify criminals after the software had been used to match images. It seems that signal processing algorithms such as those used in facial recognition would develop a higher accuracy with a greater pool of images with which to compare. I would imagine this would pose a challenge for detecting first-time criminals in which the only image the police department might be able to match the video to would be a government issued ID. I wonder whether there is any value to be gained from complementing video-image matching with video-video matching? Would it make sense to develop a capability that would allow the software to match a face at a given point in time in one video, to a face at a different point in time in other video surveillance? I’m not sure whether this would help eliminate potential options and narrow in on the correct criminal, or make the issue worse and only increase the number of potentials that must be eliminated, but it’s an interesting option to consider.
I found the use of “super-recognizers” by Scotland Yard to be fascinating, especially because I know two individuals with this ability. (One friend in particular will remember people she met once over a decade ago, and when this happens, she usually pretends she hasn’t previously met the other person so she doesn’t risk creeping him/her out.)
This article makes me wonder to what extent law enforcement agencies are able to serve warrants to companies like Facebook to utilize their facial recognition capacities in investigations.
Separately, is the gain to consumers worth the privacy we’re trading to for-profit companies?
Ellyn,
I think your blogpost is covering a topic that will become more and more important in the years to come, as we constantly face the risks of infringement on our rights vs. the expected reward of safety and security.
However, as showcased by the likes of Edward Snowden, “we the people” have never really been given a choice by our governments.
The so called war on terror has taken us astray and shifted our attention towards “protection” at all costs. Ironically, this same path is the one leading us to less freedom; freedom which modern societies have spent hundreds of years to achieve.
The sinister state attempts to keep close watch on citizens will unequivocally put a strain on that. Whether it will actually keep us “safe”, however, we can only hope.
Thanks for the post Ellyn. This is a topic that I struggle with as I clearly recognize the need for additional safety measures in the fight against crime but I also recognize the privacy concerns of normal civilians. I, for one, am a big proponent in video surveillance in public areas as they can be used to both prove someone’s guilt or their innocence. As body cameras become more prevalent in the law enforcement community, the footage can act as an additional accountability measure to ensure that proper procedures are followed and will hopefully prevent much of the ambiguity surrounding their encounters with citizens to our benefit and theirs.
The use of “super-recognizers” is also particularly interesting to me because I wonder how much weight is given to their testimony in a criminal trial. For the most part, someone is relying on the judgment of these “super-recognizers” instead of being able to have an objective resource. To me, this could be construed as merely here-say even though these individuals are specifically tested to obtain this position.