The problem with drug-sniffing dogs is not that dogs aren’t capable of sniffing out drugs; it’s that we’ve bred into domestic dogs a trait that trumps that ability — a desire to read us and to please us. If a drug dog isn’t specifically trained to compensate for this, it will merely read its handler’s body language and confirm its handler’s suspicions about who is and isn’t hiding drugs. This has been confirmed by tests of K9 units that have shown that controlled tests designed to fool handlers are much more likely to trigger false alerts than controlled tests designed to fool the dogs. The fact that mine-sniffing dogs tend to be more accurate than their drug-sniffing cousins further illustrates the point — handlers not only have fewer preconceptions about where mines are located, but they also have an incredibly strong incentive for the dogs to be accurate about finding them. But even here, the dog-handler bond can become problematic, which is why some detection experts are turning to rats.
The problem here is that invasive searches based on no more than a government official’s hunch is precisely what the Fourth Amendment is supposed to guard against. Unfortunately, the way the U.S. Supreme Court has ruled on this issue not only doesn’t account for the problem, but also has given police agencies a strong incentive to ensure that drug dogs aren’t trained to act independently of their handler’s suspicions. A dog prone to false alerts means more searches, which means more opportunities to find and seize cash and other lucre under asset forfeiture policies. In fact, a drug dog’s alert in and of itself is often cited as evidence of drug activity, even if no drugs are found, thus enabling police to seize cash, cars and other property from motorists. For example, I’ve interviewed dog trainers who have told me that drug dogs can be trained to alert only when there are measurable quantities of a drug — to ignore so-called “trace” or “remnant” alerts that aren’t cause for arrest. But these trainers say that police agencies don’t want dogs trained to ignore remnant odors, because any alert is an authorization for a more thorough search.
In 2013, the Supreme Court made things worse in Florida v. Harris. In that case, the court unanimously ruled that mere certification of a drug dog was enough to establish a presumption that a drug dog is reliable, regardless of the reputation of the certifying organization, regardless of whether that organization understands and appreciates the importance of training dogs to ignore their handlers’ suspicions, and regardless of the dog’s performance in the real world.
In U.S. v. Bentley, we see just how damaging the Harris decision really was. Lex, the drug dog that searched Bentley’s car, had a 93 percent alert rate. That is, when Lex was called to search a car, he alerted 93 percent of the time. He was basically a probable cause generator. His success rate was much lower, at 59 percent. That is, the police actually found drugs just six of the 10 times Lex told them they would. That means that four of every 10 people Lex alerted to were subjected to a thorough roadside search that produced nothing illegal.
It gets worse. Even a dog that was well trained initially can be conditioned to pick up bad habits once it’s in the field. This is exactly what was happening with Lex. It turns out that Lex’s handler gives the dog a reward every time he alerts, regardless of whether that alert is accurate. Lex isn’t getting rewarded for filtering innocent motorists from guilty ones. He’s being trained to authorize a search, each and every time he’s called to duty.
The Seventh Circuit found all of this troubling, even conceding that the reward policy was “a terrible way to promote accurate detection.” And yet according to the three-judge panel that heard the case, none of this was enough to amount to a violation of the Fourth Amendment. Following the Supreme Court’s prescription in Harris, the opinion notes that that the dog had passed tests in “controlled settings” and cited testimony about the dog’s reliability. But that testimony came from the dog’s handler. And there’s no further explanation of what those tests in controlled settings meant. Were they conducted by the police department? By the handler? Did they include tests designed to fool the handler as well as tests designed to fool the dog?
The head of the agency that trained the dog also provided testimony about the quality of his group’s training, but of course he’d say that. He also expressed concern about Lex’s high alert rate and relatively low accuracy rate, and about the fact that Lex actually had failed “two simulated vehicle searches.”
Yet the opinion points out that in the past, the Seventh Circuit found no problem with a drug dog whose accuracy rate was 62 percent. Lex’s was only slightly lower. Moreover, the court notes that the U.S. Court of Appeals for the Fourth Circuit gave its okay to a dog with a success rate of 43 percent, or less accurate than a coin flip. This even lower number jibes with a 2011 Chicago Tribune investigation of suburban Chicago police dogs that found a success rate of just 44 percent. That review also found that with Latino drivers, the accuracy rate plunged to just 27 percent, more evidence that the dogs are merely reflecting the biases and presuppositions of their handlers. Other studies have shown false positive rates of up to 80 percent. With success rates that low, it’s hard not to conclude that drug dogs aren’t tools to determine probably cause, but basically a “search warrant on a leash.”
In Bentley, it turns out that the dog was correct. The suspect was found to be transporting a large supply of cocaine. But guilty people tend to bring the appeals that set precedents because guilty people have a lot more to lose. A false drug dog alert that turns up nothing won’t result in an arrest. To get a court to rule that search illegal, the victim would need to file a civil rights lawsuit. That’s a lot of hassle and expense for someone to endure to get compensation for a wrongful 45-minute search on the side of the road. And that’s assuming they can find an attorney to take the case. Civil rights suits take a long time to resolve and are very difficult to win.
At worst, a false alert may lead to the wrongful seizure of a motorist’s cash, car or other property. Here, there’s more incentive to go to court — the motorist wants his money back. But here too, other incentives cut against legal action. First, the cost of hiring an attorney and going to court can often exceed the value of the property that was seized. It can also take months, sometimes years. And even if a motorist does succeed in getting his property returned, it’s another huge step and another round of litigation, expense and risk of failure to get the courts to declare that the stop itself was a constitutional violation. Most people are just happy to have their property back.
The point here is that it might be tempting to shirk at the injustice of this ruling: Maybe the dog was a ruse to let cops search this guy, but that search led to a huge stash of cocaine. What’s the problem?
The problem is that this ruling gives wide latitude to police agencies in the Seventh Circuit to use drug dogs as a end-run around the Fourth Amendment. And that affects everyone, not just drug dealers. That means more latitude for forfeiture, and more potential for the sort of corruption and legalized highway robbery we’ve seen reported countless times over the past few years. Even if no property is seized, a roadside search can be a humiliating and time-consuming experience. The cops will typically go through everything you own, including bags, suitcases, pockets, purses and so on. Sometimes they’ll tear out the upholstery of your vehicle. They’ll scatter your belongings along the side of the highway.
The opinion in Bentley cautions against a “race to the bottom” when it comes to drug dog accuracy, but it’s hard not to think that we’re already there. Remember, Lex was called out only when the police suspected someone was in possession of drugs. If we assume that cops are at all skilled at detecting drug runners, we’d already expect a fairly high percentage of stops in which the drug dog is called out to produce illegal drugs. If the drug dog is still wrongly implicating four of every 10 people, it’s actually quite a bit worse than a coin toss. The dog really isn’t filtering out innocent people at all (an assertion already backed by Lex’s 93 percent overall alert rate).
Now consider that Fourth Circuit drug dog that was wrongly implicating nearly six in 10 people, thus subjecting more than half the motorists it sniffs to a wrongful search. First, this means that the police themselves in that jurisdiction were already pretty bad at identifying potential drug offenders. But we can at least assume that the police there were good enough that the pool of people who get a dog sniff will include a significantly higher percentage of drug offenders than the overall pool of drivers. (I’d imagine that even the average citizen could pull this off.) Assuming that much, it’s hard to imagine a drug dog doing much worse than a false alert rate of 60 percent. To get to a false alert rate of 70 or more percent, as the Chicago Tribune found with Latino drivers in the suburbs, you’d almost have to intentionally train the dog to make mistakes.
The Supreme Court originally gave its imprimatur to drug dog sniffs because when used properly, the dog’s finely tuned sense of smell can detect drugs and their absence with incredible precision. But that caveat — when used properly — is critical. If a drug dog isn’t eliminating any innocent people, if it’s validating the suspicions of the police nine out of ten times resulting in searches for which up to half or more of the suspects are innocent, then descriptors such as finely tuned and incredible precision no longer apply.
And neither does the Fourth Amendment.