We rely on the police to keep us safe, to protect us from violence, and to safeguard our property. We also accept that a certain level of crime is inevitable — it may be technically possible to eradicate crime, but the cost of a dystopian police state is too high.
As technology advances, though, the line that we draw between privacy and security is shifting.
George Orwell’s vision of mass surveillance in Nineteen Eighty-Four seemed like science fiction when it was published in 1949, and even when 1984 rolled around. Today, many of its far-fetched concepts seem eerily close to reality. Technology has a valuable role to play in enabling the police, but it also raises serious legal, ethical, and moral questions.
The telephone, fingerprinting, polygraphs, and two-way radios have all advanced the cause of crime detection and prevention. The universal emergency number, 911, was established in 1968. The next three decades saw the rise of community policing, computerization, and DNA technology. Since the turn of the century, cameras have rolled out everywhere, and now there’s hope that big data analytics will bring new crime prevention strategies through predictive policing.
Will these new technologies make us safer than ever, or just give Big Brother a new set of eyes?
Surveillance has increased dramatically in recent years. There were 245 million video surveillance cameras in use in 2014, according to IHS research. Cameras are ubiquitous now, from closed-circuit television (CCTV) and dash cams, to smartphones and body-worn cameras. They’re even being mounted on unmanned aerial vehicles (UAVs) or drones. We can also be tracked through our phones and even scanned for weapons from a distance.
The impact of all this surveillance isn’t clear. CCTV has been increasingly adopted in the United States and even more so in the United Kingdom. The British Security Industry Association estimates that there are between 4 and 5.9 million CCTV surveillance cameras in the U.K. covering a population of around 65 million but the research on its effectiveness in preventing crime is disappointing.
“The results show that in terms of violence and aggression, there’s actually no effect on crime or criminal behavior,” Dr. Barak Ariel, lecturer and analyst in Experimental Criminology at Cambridge University told Digital Trends. “If you’re an experienced offender, you pretty much know already that if you put your hood up, CCTV is almost useless in investigations.”
There are exceptions. For example, it was CCTV footage that helped police capture the Boston Marathon perpetrators. Good-quality cameras can capture faces in well-lit settings when the people don’t move around much. Cameras in parking lots with limited access can help to reduce auto theft. Transport police have garnered good results with speed cameras, and cameras on escalators or public transport.
In 2012, the Federal Aviation Administration (FAA) granted approval to various law enforcement agencies to train operators in the use of drones. There are many potential future applications, from serving warrants to equipping drones with stun guns, but currently unmanned drones primarily offer aerial surveillance in situations where a manned helicopter would be too expensive or dangerous.
Several police departments [n1] are using drones now including Little Rock, Arkansas, Miami-Dade, Florida, and Arlington, Texas. While some are limited to car chases and siege situations, others are being used for general surveillance. They can fill in gaps in CCTV coverage and offer police greater capability to track people.
“Do we want to live in a country where everybody is in the system?”
Serious privacy concerns have slowed adoption. After a public backlash, Seattle dropped its drone program before it got off the ground. Some states, including Florida, Texas, Idaho, Iowa, and Utah, have passed legislation requiring law enforcement to obtain a warrant before using drones, but there’s still no national legislation.
There’s another problem if police want to track suspects across cities – reviewing video footage is a huge burden and few police departments have the resources.
Various companies, like SeeQuestor, are trying to address this issue by offering software that enables law enforcement to quickly review people and faces in video, but it still requires a review by a human being. Facial recognition software is not yet up to the task.
“I’ve seen four demos on facial recognition and I wasn’t impressed,” says Ariel. “The technology isn’t very good at picking out people or recognizing faces, especially when they’re moving and the resolution isn’t very high. It also only recognizes people who are in the system, so it won’t help with first-time offenders.”
The FBI already has a database with more than 30 million mug shots, and it can also access driver license photos from many states and passport photos from the State Department. But there’s a big difference between matching two mugshots and matching a mugshot with grainy CCTV footage.
Law enforcement may eventually benefit from the work that tech giants like Facebook, Google, and Microsoft are doing in this area. None of the problems with facial recognition are insurmountable — it’s simply not reliable enough yet.
For this kind of technology to work well, you’d need a database of every person’s face in the country and an enormous amount of processing power to make a search fast enough. There will also inevitably be a lot of false positives, and there are serious privacy issues around consent.
“If you do think about having this technology, there’s a cost to it,” says Ariel. “Do we want to live in a country where everybody is in the system?”
It’s not just the public that is under greater scrutiny. In the wake of high-profile incidents in Baltimore, South Carolina, Ferguson, Missouri, and elsewhere across the U.S., serious concerns have been raised about police misconduct and even brutality.
This has driven a movement of concerned citizens to take to the streets to try and document and expose unprofessional police behavior. There are Cop Watch organizations in many cities including New York, Los Angeles, Berkeley, California, and Portland, Oregon. They offer advice on how to safely record the police without getting arrested and share footage and photographs on social media networks.
This crisis in police-community relations has driven the rapid adoption of body-worn cameras for police officers.
Beyond this kind of organized activity, everyone has a smartphone with a camera in their pocket now, and it’s easy to record an incident and upload directly to social media to share it.
Mobile apps can also be used by the public to track cops and even crime, but not without a lot of controversy. An app called Vigilante, designed to alert nearby users about crimes recently reported to 911 in the area, was recently kicked out of the App Store by Apple.
The Nextdoor app, which some users have adopted as a kind of neighborhood watch, made headlines because users kept sharing reports about supposedly sketchy characters in the vicinity. Unfortunately, it was often the color of their skin that put them under suspicion, prompting the makers to redesign the reporting interface to combat racial profiling.
Law enforcement has repeatedly asserted that police tracking in the Waze app should be turned off because it’s endangering officers, but so far Google has not complied. It has never been easier for people to share information on crime and cops.
Some police officers are unhappy about the increased scrutiny. St Louis Police Chief Sam Dotson coined the term “The Ferguson effect,” suggesting that a decrease in public confidence in the police following the shooting in Ferguson in 2014, where an unarmed 18-year-old black man was fatally shot by a white police officer, [E1] has led to a jump in the murder rate in major U.S. cities.
The idea is that police officers are being more cautious and are reluctant to make the same arrests they would have in the past, which is emboldening criminals. It’s a concept that many skeptics reject, but it’s still the subject of fierce debate.
What isn’t up for argument is the fact that this crisis in police-community relations has driven the rapid adoption of body-worn cameras for police officers.
“Everybody is buying them, everybody wants to implement them,” says Ariel. “Broadly speaking, body-worn cameras look like an effective technology. There’s strong evidence from many of the experiments we ran that officers are much less likely to get a complaint against them for misconduct or use of force.”
A yearlong study of almost 2,000 officers across U.K. and U.S. police forces showed a 93 percent drop in complaints made against police by the public. Could this be the technological solution that restores confidence in police legitimacy? Many senior criminal justice officials seem to think so.
As of August 2016, 43 out of the 68 major city police departments in the U.S. have adopted body-worn camera programs. However, there’s a lot still to figure out. Even the positive effect that they’re having isn’t fully understood.
African-American men were much more likely to be stopped, handcuffed, and searched than white men.
“The question is still open on whom they have the effect,” explains Ariel. “Is it the officer using the camera or the suspect who sees the camera?”
There’s also a huge variation in how the cameras are being used, as a policy scorecard by The Leadership Conference recently illustrated. Guidelines on how much officer discretion is involved, and what happens to the footage are still being developed.
“I’m a fan of body-worn cameras, and if I was still a police officer today, I would want one, but body cameras are a law enforcement tool, not a panacea,” Dr. Tod Burke, professor of criminal justice at Radford University and a former Maryland police officer, told Digital Trends. “A lot of the focus was post-Ferguson. People thought that if the police had body cameras, this would have solved the problem, and so they were thrown onto police officers without proper policy being implemented.”
Perhaps it’s unsurprising that’s there has been some resistance from police officers on the street. The idea of being recorded doing your job isn’t very appealing.
“One of the fears police officers have is about who is going to have access to this video,” explains Burke. “Is internal affairs going to have access? Is it going to be used as part of an evaluation? Is it going to be used as part of a training video?”
There is potential for body-worn camera footage to help police modify their behavior and combat bias through analysis and training. Stanford research conducted with the Oakland Police Department on traffic stops involved computationally analyzing linguistic data from body-worn cameras.
The two-year study found a persistent pattern of racial disparity. African-American men were much more likely to be stopped, handcuffed, and searched than white men. The researchers also scrutinized the specific language and tone officers used during stops. They didn’t uncover overt racism, but there was a subtle bias problem. Researchers hope that by collecting and analyzing data like this, training can be improved and officers may be able to self-audit racially charged footage. The idea is that reviewing video of tense situations, such as a member of the public calling a police officer racist, is an opportunity to learn and develop best practices.
“There are also privacy issues with body-worn cameras,” says Burke. “Often police officers are responding to very sensitive incidents. Think about bystanders in the background or children in the home, should they be recorded? It could also discourage people from giving information to the police.”
There are technological challenges, too. How will footage be stored? Who will have access to it? How is it analyzed and redacted? How is it linked to calls and crime reports?
The leading provider of body cameras, Taser International, believes it can answer these questions. It accounts for more than 75 percent of the market right now, and it offers the Axon range of body cameras that plug into a backend system called Evidence.com.
The company’s interest in cameras grew out of a desire to make Taser weapon use more transparent. Tasers are employed by more than 18,000 police departments across the U.S. today. In the face of complaints about their misuse, Taser worked on ways to make them more transparent. The latest electrical weapons from Taser have internal logs that track weapon use, so it’s possible to review when it was used, how many times it was used, and see exactly how much electrical current was delivered.
Tasers are employed by more than 18,000 police departments across the U.S.
In 2006, the company added the Taser Cam, which is triggered to record the incident any time a Taser was used. On average, officers only use their Taser twice a year, so the company began considering a camera that might be used all the time. That led to a lipstick-sized camera design that Taser developed in partnership with Oakley, thinking that sunglasses would be the ideal mount for a police officer’s point of view.
The new camera was released in 2009, but there were problems with the initial design. It had a dedicated recorder attached with a touchscreen for playback, built-in GPS, and a big power pack.
“Size, wires, and comfort were the three biggest complaints,” Steve Tuttle, Taser International’s vice president of Strategic Communications, explained to Digital Trends. “Officers hated it, but no one hated the concept, so we went back to the drawing board.”
The redesigned Axon line of cameras is what they come up with. There are a variety of different mounting options, so they can be fixed to a uniform pocket or attached to glasses. Instead of having recording units or touchscreens attached, they hook up with the officer’s smartphone.
These cameras are on continuously during a shift, but as a concession to police concerns over being constantly monitored, they only save the last 30 seconds of footage. This also cuts down on the amount of video that must be stored and analyzed.
There’s a big round event button that officers double tap to record an event. It saves the buffered 30 seconds, without audio, but records from then on, with audio, until the officer holds down the button for five seconds to turn it off again.
Department policy dictates when officers must trigger an event video. It may be when they get a radio call, when they see a crime in progress, or when they have any interaction with the public.
The accompanying smartphone app can add meta data and GPS information, and it also allows officers to review the video and add notes. They can’t delete videos and all footage is encrypted. At the end of the shift, they dock the camera back at the station, where it recharges and securely uploads everything.
Each department can determine who has access to that footage. That means video of a homicide, for example, may be limited to the chief and assigned homicide detectives, and it will be retained permanently, so it can’t be deleted.
If departments are recording every interaction with the public, they might mark some encounters as innocuous, so that the video can be tagged for deletion after 60 days, or whenever it passes its statute of limitations if no complaint has been made.
Department policy dictates when officers must trigger an event video.
Taser is also trying to tie all the evidence together digitally on the back end with Evidence.com. Different cases can encompass body camera and CCTV footage, crime scene photos, and reports, and they can be shared digitally with the district attorney. Of course, this depends on departments and prosecutors licensing the software, which costs between $15 per month per user and $79 per month per user, depending on what features you need and whether you want Axon cameras.
That may sound expensive, but it’s proving to be very popular. Taser reports that sales of body cameras and the related software are now surpassing stun gun sales.
The latest feature for Taser’s Axon camera line is Wi-Fi connectivity, so footage and data from body-worn cameras may soon be rolling directly into databases.
“We’re preparing for the future, so you can take the data and apply algorithms and machine learning, to use it effectively to combat crime and uncover trends,” says Tuttle.
Body camera footage could have a major role to play in facial recognition and people tracking. The technological barriers to real-time streaming are coming down. There is scope to tie together all this data and camera footage into a real-time system to help the cop on the street.
The Domain Awareness System, developed by Microsoft and the New York City Police Department (NYPD), looks like a step in this direction. According to former Mayor Michael Bloomberg, it allows police to “access relevant information gathered from existing cameras, 911 calls, previous crime reports, and other existing tools and technology.”
All of this data and footage can also potentially feed into models that inform deployment and seek to identify when and where crimes will happen and even who might be involved in them.
“The idea of predictive policing is to bring systematic intelligence to bear on policing that moves beyond merely responding to calls,” Professor Peter Manning, chair in the School of Criminology and Criminal Justice at Northeastern University, told Digital Trends.
The theory and research behind this dates back to the 1970s. In simple terms, it was about recording where crime happened and using that information to predict where it might happen next.
“When I was a police officer, we had something called pin maps,” says Burke. “We had a map at the station house and we’d put a little blue pin, red pin, yellow pin depending on the crime wherever it occurred, then after a while, we see where these little clusters are and say, ‘Okay, that’s where we need to focus our attention.’”
As computing power increased and records improved, many police departments started to produce crime maps (which look like heat maps) highlighting crime hot spots, and sometimes even heat lists of people likely to commit or be the victim of crimes.
Predictive policing has evolved out of crime mapping, which Manning studied and wrote about in his 2008 book, The Technology of Policing: Crime Mapping, Information Technology, and the Rationality of Crime Control. He found that positive claims about the impact of crime mapping and CompStat (short for computer and statistics) programs were largely overstated.
“Anyone less than 100 percent committed to perpetrating crime can be deterred.”
“There’s absolutely no evidence that anyone has shown that [S1] the technology of mapping and analysis has any bearing on police practice,” Manning explains. “In fact, all the research shows that it has none.”
The problem isn’t necessarily with the ideas or the analysis, it’s the implementation.
“Unless the police alter their pattern of deployment, it doesn’t matter what information they have,” says Manning. “My argument is that the technologies that have been adopted by the police historically have always been fitted into the present structure or practice, they have not altered the practice much and they have not altered the structure of how policing is done, with some exceptions.”
Some studies show that hot spot policing has a positive impact on crime reduction, when compared to routine patrols with no link to the data and distribution of crime. But it may seem like common sense to say that putting more police in small geographical areas with high rates of crime will reduce crime.
This kind of analysis and mapping has grown more sophisticated in recent years, giving rise to the term predictive policing.
“Faster and more regular data collection by police over the last decade and increased computing power is enabling us to look at, not just where crime has happened in the past, but where it’s likely to be in the future,” Jeffrey Brantingham, a UCLA professor of anthropology, told Digital Trends.
“Our paper published at the end of 2015 on randomized controlled experiments we ran in LA suggested positive impacts. Not only were there increases in the fractions of crime that you can predict, but also, when you put that into police officer’s hands, it has a sort of doubling of the effect of crime prevention.”
Brantingham is also a co-founder of PredPol, which supplies predictive policing software to several police departments including Los Angeles and Atlanta.
PredPol strictly looks at when and where crime might happen, and it only draws upon past records about when and where crime has occurred. Predictions cover 500 x 500 foot boxes, roughly the size of a city block, and they’re done on a shift-by-shift basis.
“We could predict at finer scales and in real time, but we’re looking for the scale that’s most appropriate for the way in which police do their job,” says Brantingham. “The truth is that no algorithm is ever going to get out of the car and police the problem.”
In trying to build a mathematical model that can anticipate and forecast crime, algorithms may weight very short-term patterns of crime more heavily, but long-term historical data and structural features of the environment must also be considered. If a burglary occurs at a house, it may be because an adjacent parking lot made for easy access, or perhaps there was a successful burglary next door a day or two earlier and this home has the same layout, making it a softer target.
But if you foil a crime in one location, does the criminal just go around the corner?
“Studies suggest the opposite is true,” Brantingham explains. “You put police officers in a particular place and not only does crime go to zero in that location, crime is actually reduced over a much wider area.”
This is known as the diffusion of benefits. The theory is that you’re pushing offenders out of their comfort zone. They understand the targets and how to be successful in this area, so things won’t be quite as easy if they have to go around the corner. At least some of the time, they’ll hit the tipping point where they weigh things up and decide not to commit a crime at all.
“You don’t want to turn this into Minority Report, what privacy rights are we willing to give up for security?”
“Hollywood has led us to think of criminals as walking bombs that want to commit a crime all the time, but most offenders are not actually that committed to what they’re doing,” explains Brantingham. “Anyone less than 100 percent committed to perpetrating crime can be deterred.”
Not everyone agrees that predictive policing is effective. A study conducted by the Rand Corporation of a seven-month field trial of predictive policing in Shreveport, Louisiana, found there was no statistically significant reduction in property crime.
“There were no effects,” Jessica Saunders, senior criminologist at Rand, told Digital Trends. “What we saw there was that a lot of people already use hot spot mapping, and there’s only a marginal increase in accuracy using a predictive model.”
It seems that there isn’t a big difference between freshly branded predictive policing and what police departments are already doing. There may also be a disconnect between the top brass and the officers on the beat.
“We have a cohort of really professional, forward-thinking, modern police chiefs,” says Saunders. “But we also need buy-in from lower-level people in the department who are actually supposed to be actioning these predictions.”
In other words, once the police have the data, what do they do? This is tougher to answer when you try to predict who is going to be involved in crime, rather than where and when it might happen.
“In Chicago, they predicted people who were at higher risk of becoming homicide victims, but they really didn’t know what to do with that information,” explains Saunders. “We’re getting better at predicting, but until we know what we’re going to do with those predictions, we’re not actually going to be fulfilling the mission, which is preventing the crimes from happening.”
The Chicago “heat list” used an algorithm to draw up a list of more than 400 people thought to be at the highest risk of gun violence in the city. When Rand investigated the impact, the study found that there was a lack of clarity about how to use the predictions, and worse, that some officers may have been using the list as leads for closing shooting cases. Ultimately, there was no reduction in crime.
Part of the problem is that we don’t have the same depth of data on who is committing crime as we do on where and when it happens. People move around, their lives can be chaotic, and many crimes go unsolved.
The best predictor of future behavior is past behavior, but it’s not enough on its own. Concerns about racial profiling must be addressed, but squeezing the bias out of models is easier said than done. Plugging in more data may improve predictive accuracy, but how far do you go?
The potential for technology to aid the police has never been greater, but the fundamental tug of war between safety and civil liberty still exists. Maintaining that balance in the face of the current crisis in police-community relations in the U.S. seems to be driving technology forward in some cases, and holding it back in others.
“You don’t want to turn this into Minority Report,” says Burke. “What privacy rights are we willing to give up for security? You could do pat-downs on everyone walking down the street, and you’re likely to find weapons and stop crime from occurring, but at what cost?”
As technology continues to give law enforcement new tools for policing, it’s society – not engineers – that will have to figure that part out.