One overcast day in February 2016, a massive police motorcade coursed through Pittsburgh. Hundreds of officers mourned one of their own—Officer Aren, a five-year-old German Shepherd, a four-year veteran of the city’s Port Authority force.
The dog had died from stab wounds. Days earlier, 37-year-old Bruce Kelly, Jr., a Black Baltimore resident, was reportedly drinking in a park gazebo with his father when the officers arrived. An argument ensued and Kelly fled. He was cornered by several officers—reportedly encircled—a few blocks away, where he was holding a knife. The officers targeted Kelly with multiple tasers; his winter coat apparently diverted their charge. K9 officer Aren was then ordered to subdue Kelly—or literally ‘take him down’ to the ground—and Kelly stabbed the dog in the mouth as the dog attacked him. Officers then shot Kelly repeatedly, executing him in the street.
Kelly’s funeral, by all accounts, was far more modest than Aren’s.
Decades before Trump tweeted a digitally-faked image of himself awarding a medal to Conan—a dog injured in the raid on Islamic State’s leader Abu Bakr al-Baghdadi in Syria—dogs were being bred around the globe to become defense and law enforcement’s authorized tools for dangerous and ethically fraught tasks alike.
While technology is currently being developed that could make these working canines obsolete, the dogs’s treatment, legal status, and their very necessity are being rightfully questioned. Work canines are very useful in cases of search and rescue, bomb detection, and locating electronics which may contain child pornography. But dogs will do as they are told, and police dogs specifically can be used to intimidate, incriminate, attack, and serve as proxy for actions that police do not want to be held directly responsible for. These uses operate under an assumption that we need police dogs to do this—an assumption that isn’t historically supported.
They called him Rin Tin Tin, one of millions of dogs used in the so-called Great War, most of which carried messages or guarded patrols, and over a million of which died. Rescued as a puppy from a battlefield kennel in 1918, “Rinty” soon became a mascot for the US Army’s 135th Aero Squadron, and then began a silent film career that changed US dog lore for generations.
He and fellow 1920s ‘hero dog’ Strongheart also amounted to great PR for working dogs, especially German Shepherds, which American military and police forces had increasingly begun to adopt as tools and weapons.
After WWII, when dogs were used as messengers, scouts, and tested as bomb-delivery systems (among other things), the surge of scientific innovation again drew attention to dogs’ seeming abilities. The inaugural commissioner of the Federal Bureau of Narcotics (FBN), Harry J. Anslinger, who’s considered to be the country’s first drug czar and the godfather of drug prohibition, may have been inspired by British wartime experiments that trained dogs not just to track scents, but rather to seek out specific ones, historian Justin Hubbard, PhD told Gizmodo.
Wherever Anslinger got the idea, the first US drug dogs came about “almost immediately after WWII,” Hubbard said in a phone interview. Anslinger assigned an agent to investigate training programs around the country, which were sparse, and told him to create the agency’s own.
Until the FBN took an interest, police dogs were mostly used as trackers or “general muscle” by individual departments, Hubbard said; then during raids of high-density, low-income Chinese and Black communities in NYC. By the early 1950s, drug dogs appeared in other municipal police forces, starting with cities like Baltimore and St. Louis.
Meanwhile, the Department of Defense was exploring numerous avenues for putting trained dogs to work, including one attempt to see if dogs could detect landmines or communicate with handlers using ESP (an idea the military decided not to follow up on). After a lull in the mid-to-late ‘50s—during which time Anslinger told a Texas police force that he was discontinuing the FBN’s program, calling it a “failure”—municipal and federal dog use really started climbing by the late ‘60s, Hubbard said, due to overlapping factors.
For one thing, the dog-research business was booming. The army developed its own “detector dog” program, and followed up with special training and placement for smelling explosives, narcotics, or both. University programs also got in on the training and behavioral research. “In the 1970s, one thing dogs were great at was getting research dollars,” Hubbard noted.
At that time, those same dogs increasingly appeared in different arms of law enforcement, too. By the ‘70s, the military had started selling its trained drug dogs to police forces around the country, as well as to customs enforcement officers.
This not only made drug dogs a regular part of the nation’s legal landscape, Hubbard said; it also forged “a direct connection between the army and how we patrol our borders.” In short, military dogs and methods for their use became part of civilian law enforcement across the country, but also the systems we use to control who gets to leave or enter the US.
According to Bob Dougherty, law enforcement canine coordinator for the Penn Vet Working Dog Center and a retired K9 police handler, the international defense-dog market has only continued booming after that. Once the US wars in Iraq and Afghanistan began, and the ‘War on Terror’ took hold, things escalated again, Dougherty said. “When ‘terrorism’ became much more of an everyday word, not just here but everywhere, demand [for dogs] increased.”
Today, third-party vendors help fill numerous government contracts each year by sourcing dogs from around the world, generally as cheaply as possible. There are also rental companies to serve public and private clients: for example, if police forces can’t or won’t provide canine security presence at high-profile events, organizers can hire trained dogs short term from private companies, perhaps supplied by a separate broker, for their bomb-sweeping or drug-deterring needs (though some studies have suggested the latter tactic doesn’t really work).
“If people will feel safer when they see two cops, a dog, and an automatic weapon—and some people actually do feel safer when they see that—then [firms] can offer that.”
In the past few decades, authorities and private companies have also boosted their efforts to meet growing security demands with technological tools instead: whether as surveillance, detectors for illicit materials, or methods of using force.
But it’s still unknown just how effective robots can be—or dogs, for that matter—at doing those kinds of jobs.
Law enforcement agencies use dogs in a few specific areas, mostly to do with finding the invisible and controlling or retrieving human bodies. These tasks rely on dogs’ senses of smell, their natural physical features and abilities, and their capacity to take direction. Increasingly, it’s a skill set that engineers in government, private industry, and academia have being trying to reproduce.
By far, dogs’ most in-demand and expensive trait is their superior sense of smell. Roboticists have tried to achieve simulating dogs’ smell detection using different ‘electronic noses,’ with varying levels of success. Their biggest aims are to find cheaper, more convenient, and more accurate detecting tools than trained dogs, which can cost $25,000 or more each, and big, unwieldy machines like mass spectrometers, which can cost hundreds of thousands of dollars.
Like dogs’ noses, these devices generally have to be trained or preprogrammed in order to pick out trace scents—i.e. airborne or other residual particles—of particular substances. Some researchers have focused on finding ways of sniffing out money, for example. In 2014, KWJ Engineering announced its plans for a Bulk Currency Detection System, which could help detect large amounts of US currency being brought into the country without taxes paid. (As of 2019, however, it seems this has yet to hit the market.) Other researchers have focused on sniffing out humans, drugs, or even explosive materials. In 2014, Mexico’s Tecnológico de Monterrey announced that researcher Blanca Lorena Villarreal was developing an electronic nose (complete with two nostrils) that could detect the smell of alcohol, and could also be programmed to recognize and follow the scent of blood, sweat, or human urine, which would be utilized in search and rescue situations. Last year, researchers at Duke University also announced that their digital odor-sensor, which uses live smell receptors from mice, could potentially serve as drug-detector in the future. But they also warned that, in general, electronic noses are a long way from being used in the field.
That hasn’t stopped law enforcement from trying out new scent-detectors, or from testing other kinds of police robots in the field—with seemingly scant consideration and murky ethical implications. This includes potentially violent situations. On police forces and military patrols, where dogs are part of officers’ escalating toolkit for pursuing their goals with minimum risk to themselves, robots have been deployed for dangerous scenarios in numerous cities and countries, and been credited at least once with effecting lethal force.
In 2016, for example, there were several instances of such use: FBI agents and New Jersey police sent a bomb-squad robot into a train station to diffuse home-made explosives, which detonated when the robot was trying to cut a wire. That same month, Politico reported, “sheriff’s deputies from Los Angeles used a robot to disarm a violent suspect who had barricaded himself [in a barn] in a Southern California desert.”
The week before that, “another robot was used to force doors open as police searched for a gunman hiding out in an Amtrak train.” That summer, officers in Dallas also used a robot with a bomb attached to disarm and kill a gunman who’d gone on a fatal shooting spree against police.
In some cases, police have even been known to use both robots and canines in their operations. In 2019, following one dramatic high-speed pursuit in California, “It took police sending in a robot, a drone, stun grenades and even a K-9 unit which appears to have bitten the suspect before he was taken into custody,” as ABC News reported.
The reasons for using robots in these scenarios are multiple: they may be more durable than dogs in some ways, and can be equipped to carry and control various weapons. They can potentially be controlled more closely (if their operating systems and AI permit), and may even collect evidence or send back live video feeds while they’re in there.
Unlike dogs, robots don’t experience pain, nor can they become afraid and increase the risk to present humans and themselves as a result.
Because sometimes, of course, dogs and the people they’re trained to control get hurt. Whether because dogs are scared or poorly instructed, police dogs can often cause more harm to human targets than their handlers presumably intended, or needed. This includes lacerations and trauma to human bodies, the death of other animals, and the death of humans and dogs themselves.
As a result, some cities and countries are considering stronger legal protections for their canine officers—in most cases, protections not from police forces that put them in dangerous situations, but from the civilians they’re told to attack. Given the absence of consistent information available to the public surrounding each incident, it’s not clear that this is the best way to protect police dogs.
The question remains whether dogs can be used to reliably detect explosives, drugs, or other things they are legally viewed as able to find. This question has been around for as long as we’ve used drug dogs in this country, and has shown up in court documents since scientists first tried to define dogs’ abilities in technically legal terms. According to Justin Hubbard, by the early 1970s, courts had started to treat dogs as “wholly legitimate devices” for justifying an initial police search, among other things. But it was often up to police and animal experts, and ultimately to individual courts, to say whether or not this had been the case. In the decades since, some researchers and critics have repeatedly expressed doubts over dogs’ abilities to detect substances, to alert their handlers consistently, and, importantly, whether dog-handler relationships are compromising the results.
In 2015, an Illinois court upheld the conviction of a driver whose drug arrest followed a positive ‘alert’ from a dog named Lex. Police pulled the man over for reportedly changing lanes without signaling, and then decided, for one reason or another, to call in for a drug dog. Lex alerted officers, who searched the car and found 15 pounds of cocaine.
On appeal, the defendant argued that Lex didn’t have a good performance record prior to this instance; as the judges phrased it, “He [had] a point.” Lex was found to be alerting his handlers 93% of the time he was brought to do an “open-air sniff” of a vehicle (which he received praise and rewards for doing), while his accuracy rate in the field, 59.5%, was “not much better than a coin flip,” the judges wrote. Two years later, the dog was also briefly removed from duty after failing two vehicle-sniffing simulations.
Nevertheless, the judges said, the US Supreme Court had already ruled based on a Florida case that evidence of a dog’s training, not its record, is sufficient to deem it reliable. They further said that 59.5% is an acceptable rate of accuracy in the context of dog-related legal precedent. They also pointed out that Lex was only being used “when the police already suspect that drugs may be present,” which an “embarrassed” representative from the school that trained him, like government prosecutors, said had been partly to blame for Lex’s high alert rate.
At the same time, the Illinois judges noted, “Presumably the dog knows he will get a [toy] every time he alerts. If Lex is motivated by the reward (behavior one would expect from any dog), he should alert every time. [This] seems like a terrible way to promote accurate detection on the part of a service animal, lending credence to [the defendant’s] argument that Lex’s alert is more of a pretext for a search than an objective basis for probable cause.”
Cornell University law professor Sherry F. Colb reflected on the issue for Justia in 2012. Colb wrote, “Like machines, dogs have error rates. But unlike machines, dogs may also have their own desires—desires that come into conflict with the objective of identifying whether or not there are drugs in a given car or suitcase.”
“Dogs, for example, are able to pick up on the feelings of the people around them, and they are also (often) eager to please the people whom they love,” she explained. “Like the spouse asked to assess her partner’s appearance, the dog can often tell which of the two responses [its] handler will find pleasing, and which he will find disappointing. Neither human nor dog wants to disappoint the object of his or her affections.”
In other words, police dogs may be inclined to react however they think officers want them to, which could effectively facilitate undue searches or arrests. And unlike the human officers that deploy them, dogs can never be cross-examined about their decisions or motivations.
Even facial recognition algorithms, which authorities have started using despite apparent racial and gender biases, can tell us something about how the results (skewed or not) came about, unlike dogs. That’s another aspect of police-dog life that hasn’t changed in decades, it seems: they are still generally recognized as tools that operate without bias, even while being ‘man’s best friend’ to one individual, and specifically trained for certain outcomes.
In the late 1960s and early ‘70s, as the number of police dogs rose against a backdrop of social conflict and unrest, they represented objectivity in the midst of varying human perspectives, according to Hubbard. For US law-enforcement officers at the time, whether they were manning the border or facing Vietnam War and civil rights protesters, the dogs seemingly offered a way to protect themselves from perceived bias or even eliminate it, he explained.
The same kind of thinking survives today. “If a dog alerts their handler [about] something, police can later say that dogs can’t be biased against this or that group of people—[supposedly proving] that they didn’t arrest someone because of their race, or their haircut,” Hubbard said.
In reality, though, the opposite may be true: dogs might pick up on their handlers’ feelings of hostility or fear, and act accordingly—thus amplifying or legitimizing human bias, and potentially turning it into a legal foothold for police.
Ultimately, the human element can’t be removed from dog-handler partnerships, in the legal or literal sense, Hubbard said. Conscious or otherwise, a lot of collaboration and communication happens between dog and handler, much of which we can’t possibly document or understand.
In some cases, though, the complexity of issues around dogs policing humans can unravel enough, and in enough detail, to be examined with our own eyes—in nearly every case, showing that human decisions and police ambitions are at the core.
In the case involving Illinois drug dog Lex, for example, the story began well before Lex was called to the scene, and even before the car’s driver had failed to signal.
It actually began when the arresting officer ran the plates on “a Chrysler Pacifica he observed at a Circle K convenience store, [learning] that the car was registered to [a driver whose] license had expired 18 years earlier.” Based on that, the officer followed the car onto the highway and “stayed with it until the driver committed a lane violation,” allowing police to begin the search they wanted, and later—with Lex’s help—to make an arrest.
In 2016, Pittsburgh Post-Gazette columnist Tony Norman got a closer look at the police execution of Bruce Kelly and its aftermath, which he described in two newspaper columns about the individuals involved.
The shooting of Bruce Kelly was declared justifiable that summer, and Kelly’s family have since filed suit against police. At the time, Aren was reportedly the eighth K9 officer to be killed so far that year. The night he was killed, “dozens” of Pittsburgh police carried out his body in a flag-draped coffin from the veterinary office where he had died.
About a week after the shooting, Norman and his colleagues watched from the Post-Gazette’s offices as a long procession of vehicles and hundreds of human and canine police officers amassed to mourn Aren’s passing. “It was absurd,” Norman told Gizmodo. “The procession stretched from Downtown to the North Shore, where the Post-Gazette’s offices are, and we saw car after car of police getting out, of dogs getting out.”
“It was like something from a Leni Riefenstahl film: all of this public mourning for a canine life that was purposefully sacrificed by its owner, made to attack a basically unarmed man who was surrounded by cops—it was for the sake of police egos at that point.”
Since this incident, Norman told Gizmodo, he feels like he’s seen fewer K9 units in Pittsburgh than he used to, at least Downtown, where the Post-Gazette‘s offices were previously located. “You used to see canine patrols, canine cars everywhere, but I can’t remember the last time I saw one here, except during football matches, rap concerts—then they’re everywhere.”
Following a heavily protested G7 summit in Pittsburgh in 2009, when police rolled out everything “from sound cannons to gas bombs” to subdue civilians, he explained, “There was so much blowback from that that police apparently decided to go low-tech again, [and] more police dogs came in.”
“But after the Aren thing, there may have been consensus that this is really going too far now: that there’s a certain laziness in law enforcement, that dogs are being used as weapons of oppression, but they are also being oppressed, and being trained to do things they really shouldn’t do.”
At the same time, he’s still wary of aggressive police tactics as they impact both humans and dogs. “When there’s an incident of a dog being ordered to attack someone, that person is experiencing primordial fear: you’re going to try to defend yourself, and if you try to kill the dog, police will kill you.”
“There’s a temptation on the part of police to do what is most efficient for them, and that might usually entail exposing the public to half-baked police tactics that endanger them,” Norman went on. “That isn’t necessary, and there should be a real emphasis put on deescalation techniques. There’s no reason that every conflict between a civilian and a cop has to end in some sort of violence.”
“We don’t have to be that kind of society,” he added. “Dogs on the street, all the paramilitary shit—that’s all basically there to antagonize people.”