Crooked cops, Clever Hans, and drug-sniffing dogs

New York City recently introduced a team of sniffing dogs specially trained to patrol the subway. The cops won't say whether these dogs are trained to detect drugs, explosives, or both.

This story made me wonder: How do police departments prevent unethical K-9 cops from secretly teaching their animals to sniff on command? Obviously, they're trained to sniff only when they smell contraband, but it seems like it would be pretty easy to override that default, deliberately or by accident.

I understand that police dogs often live with their handlers. So, cops usually have plenty of time alone to teach their dogs new tricks. Could an officer train a dog to sniff on command using covert signals, such as eye movements or seemingly natural gestures? It would be very convenient for a bullying cop to teach his dog to sniff on command. In the wrong hands, a subverted sniffer could be a furry license to search anyone.

For that matter, how do the K-9 police guard against the "Clever Hans" phenomenon? In the early 20th century, a horse called Clever Hans became world famous for his apparent ability to do arithmetic and other astonishing feats. Upon further investigation, it turned out that the horse was learning to respond to subtle, involuntary changes in posture. People were asking him questions to which they already knew the answers. Hans would tap out the "solutions" to arithmetic problems with his hoof. It turns out people naturally shifted their heads when he got to the right number.

I wonder whether something similar might go on between police officers and their dogs. Officers develop a close bond with their dogs. Even civilian dog owners report that their animals can be almost eerily attuned into their body language. Do the K-9 trainers use safeguards to make sure the officer's feelings don't subtly affect the dog's behavior? For example, suppose that a subway officer strongly suspects that someone on the subway is up to no good. Could subtle cues from the officer make the dog more likely to sniff a person?

I wonder how you would go about closing these security holes. You could probably detect the Clever Hans phenomenon by designing drills where the officer's beliefs are the independent variable. If the dog is more likely to sniff at packages when the officer thinks there's contraband in it, there's a problem. If it's an honest cop, animal behaviorists could probably work with the officer to control whatever ticks are setting the dog off. I have no idea how to make sure that cops aren't deliberately teaching their dogs to respond to illicit commands, though.

I don't know whether a dog sniffing at a person constitutes probable cause for a search. I hope not. It seems to me that a sniffing dog could easily be a crooked cop's best friend.

Comments

[dan]

my guesses--and they're just guesses--are as follows:

1) a positive reaction from a drug-sniffing dog constitutes probable cause (feel pretty good about this guess)

2) every time a cop conducts a search of a suspect for which there is probable cause, the officer is required to file an incident report (feel less good about this guess, but I'd be surprised if it's wrong)

3) a pattern of false positives would be noticed reasonably quickly (this is my least strong guess, but at least in cities like NYC there's so much computerized police data crunching that it's just got to be the case), and

4) a team with a false positive rate that is significantly above average will be reassigned or retrained.

So I suspect it's largely self-correcting. Although it would be easy enough to go the other way--train your dog not to bark at anyone at the end of a long shift, when you just want to go home.

[dan]

incidentally, if your claim is that drug-sniffing dogs would make it easier for cops to get away with planting drugs on innocent suspects, well...I guess that's true, but I think it's taking it a little far. First off, awfully expensive. I'm also used to seeing dogs stationary in high-traffic areas, where there really wouldn't be an incentive for a cop to place drugs on anyone in particular. It would also get pretty expensive...and after all, much easier to just plant the drugs and make up a story about the perp giving you consent to a search, or dropping a gun, or otherwise giving probable cause.

Equally incidentally, anybody ever tell you your anti-bot distorted letters and numbers system is incredibly hard? I got my first two attempts to post that last comment wrong, and was nervous that if I failed a third I'd be told I was actually engaged in a Turing test, and had demonstrated that I was A.I. All I want to do is speculate wildly about police procedure, and all of the sudden I find myself in a Bladerunner nightmare...

[Lindsay Beyerstein]

My anti-bot system hassles me all the time, even when I'm posting from my home computer!

I wasn't thinking in terms of planting drugs or explosives. I was envisioning a scenario where a subway cop just wanted to harass or embarrass a suspect. Especially if a cop just wanted to freak out some sketchy looking kid in a crowded subway. That's a pretty minor problem all things considered, but it certainly could happen.

Here's a more serious problem. Imagine a case in which a cop just knows he's got something on some guy or gal, but not something his dog is picking up on--a gun, meth wrapped in bubble wrap smeared with camphor, or whatever. So, the cop is convinced of a big payoff and won't be deterred by the possibility of a false positive.

Some unscrupulous law enforcement officials privilege their hunches above the evidence. If I were one of those cops who felt deep down that I had the right to search anyone based on my own hunches, and I happened to have a police dog at my disposal, well, I know what I'd do...

[TW Andrews]

If the dog is more likely to sniff at packages when the officer thinks there's contraband in it, there's a problem. If it's an honest cop, animal behaviorists could probably work with the officer to control whatever ticks are setting the dog off.

I think this is pretty unlikely. Cops, to some degree, are trained to be suspicious, and it would be essentially impossible for them to control all of the non-verbal indicators of their suspicion to the point where a dog wouldn't recognize their posture. I'm not even sure that a another human would be able to notice all of the cues so that they could try to adjust them.

Basically, K9 units should work in teams where only one of the officers is the animal handler. If a dog sniffs someone and there wasn't a clear, verbal cue from the handler, any evidence uncovered in a subsequent search is inadmissible.

[dan]

"Here's a more serious problem. Imagine a case in which a cop just knows he's got something on some guy or gal, but not something his dog is picking up on--a gun, meth wrapped in bubble wrap smeared with camphor, or whatever. So, the cop is convinced of a big payoff and won't be deterred by the possibility of a false positive."

well, that's certainly a risk, and I don't think there's any sure way to control for it. I just don't see why it's a risk unique to K9 handlers; indeed, I'd think it would be less of a risk for them, because a) they'd have to train their dogs for a fairly remote, or at least rare, contingency, b) I am pretty sure there are much easier ways to make up probable cause than to get your dog to bark, and c) as I said, I think they're usually stationary, letting crowds go by them, rather than singling out particular people. I can easily see how a detective or a beat cop could be convinced some particular guy they've had contact with or are investigating has something illicit on him and gin up evidence to make it so, but I don't see much of an incentive for a K9 cop to do the same.

I could be wrong on the stationary nature of drug-sniffing dogs, though. I work on Wall St., so I see literally half a dozen bomb-sniffing dogs every day, but seldom see any drug sweeps (although it would sure be interesting to see what a drug sweep would turn up), so maybe I'm overly inured to the ways of the bomb-sniffers.

(by the way, the police dog scandal I'm most familiar with is the use of police dogs by lonely cops to hit on tourists. Pretty comical in its regularity)

[Aerik Knapp-Loomis]

I'm surprised the Clever Hans effect on police dog units has never occurred to me before. I'm actually kicking myself inside a little. I recently watched a video online of what was alleged to be cops shooting an unarmed victim 81 times (actually he did have a gun, even if the impetus for shooting the young man was him throwing it away) and in all the shooting, some idiot let his dog go and they ended up killing the dog. Instead of preserving the integrity of the scene, the dog's trainer immediately walked up and took the dog away. A lot of officers are probably already incompetent trainers and/or take their dogs into inappropriate situations.

Also, think about how this might affect alleged 'cancer sniffing' dogs.

[Amanda Marcotte]

I doubt that the controls are good enough, but often one officer trains the dog and then another officer keeps the dog. My uncle---who is a weirdo asshole and I try to avoid him---is a K9 dog trainer. He usually only keeps a dog for 6 months to a year.

[Matt]

What most people don't know is that horses can not only do arithmatic but also algebra. Mostly they keep it to themselves, though.

[PhoenixRising]

My current dog is a washout from an assistance program. (She's just too damn happy about all human contact to be a guide dog.) The breeder also provides dogs to various law enforcement folks around the state.

What she told me about training, and my experience with this dog reflects her instructions, is that dogs trainable enough to do any kind of law enforcement work generally bond strongly to the trainer. I've noticed that my dog pretty much does whatever I want her to without my having invested more than a few hours in teaching her--she seems to have very strong intuition about things like who I find suspicous.

So I would be quick to conclude that a dog might sniff on whoever the handler thought smelled funny, given that the dog cares enough about what people want to get through the training. However, a dog that alerted--different than sniffing, usually a bark or sit--falsely would be sent back for a tuneup.

You can teach a dog who you think smells, without knowing that you've done it, and without knowing that the dog knows. But I don't think you can teach a dog that is trained to alert on a scent to give false positives on humans that you find suspicous. The point of training the dog to alert on scents is that its nose is more accurate than human observation can be.

In sum, the kid that the cop thinks looks hinky on the subway? He's going to have problems only if the dog agrees that he smells like a controlled substance.

The question I think is whether being sniffed by a dog is an unreasonable search, and I think that's settled law. Or at least, every time I fly into certain airports I know I'm going to be sniffed by a dog and have no idea what she's looking for.

[PhoenixRising]

Posted before editing again...the point I jumped past is, Sniffing someone is a search. It's not what happens right before a search. If the dog alerts on someone who is being sniffed, further searching is permissible--because the dog alerted, not because the handler gave the sniff command. I don't think a handler could teach the dog to alert on anything other than the scent trained for without breaking the dog's specialized training and making it unpredictable and useless.

So teaching the dog to only 'search' folks you feel are suspicious, whether actively as you describe or accidentally which I think is unavoidable, is probably going to result in fewer busts. (Because what you want the dog to do is sniff the subway car or platform, honing in on the source of the scent it cares about.) Given that all rational actors are most concerned with their employment being included in next year's budget, it seems an unlikely abuse of power.

Compared to only writing jaywalking tickets to the cute girls.

[dan]

actually, a sniff isn't a search. You've got no privacy rights regarding what is out in public, including your scent.

(there was a really interesting supreme court case a year or two back about whether the principle applied to cops using infared detectors to look for evidence of marijuana production in homes--the argument was that you've got no expectation of privacy regarding light emanating from your home, and if the police look at a different frequency that happens to indicate likely criminal activity, there's no privacy violation [i suspect I'm getting the technology wrong, but I think I have the law right]. I can't remember what the outcome of the case was, actually, but I'm pretty sure the right to smell people without a warrant was the most important precedent the government relied upon)

I believe the sequence is dog sniffs and gives a signal, and handler then has probable cause to request a search.

[Utica]

insofar as they are dogs, i like police dogs. they may hang out with some bullies, but the police dog knows he's running the show. you can see it in his strut.

[pili]

a positive dog sniff is enough for probable cause for a search warrant. often the threat of getting a dog team is used to get people to consent to search without a warrant. NEVER consent to a search even if you think your car/home is clean. dog sniffs depend on the honesty and integrity of the handler. some are, some aren't.

[pili]

a positive dog sniff is enough for probable cause for a search warrant. often the threat of getting a dog team is used to get people to consent to search without a warrant. NEVER consent to a search even if you think your car/home is clean. dog sniffs depend on the honesty and integrity of the handler. some are, some aren't.

[saoba]

I once, long ago, had a drug dog alert first on my car trunk and then on my infant son's playpen.

I knew I didn't do drugs and I was pretty sure my 11 month old wasn't nipping out to buy pot while I was asleep.

The dog, it turned out, was trained to get a treat after a 'hit'. His treat was an arrowroot teething biscuit.

You can guess where this is going, I'm sure.

He alerted on the car, evidently, on the 'there's a treat, there must be drugs!' theory.

The handler nearly died of shame on the spot.

[togolosh]

This link talks about a supreme court case which found that dog sniffs at traffic stops aren't unreasonable search and seizure. I'm trying to find the source of something I read at the time which suggested that under the ruling a cop could stop you for a legitimate offense (broken taillight, etc) and once you'd been stopped you could be held while a dog and handler were brought to the scene for a sniff around. When I have more time I'll check Radley Balko's site, since he's usually pretty on top of this sort of thing.

[GerwingR]

This is a streach right ? you get us to bark when we hear the bell.... geeer... yelp

[John Padilla]

My son and two other boys... that he was not suppose to be with ... got stop by the cops and all three of the boys said that the dog was walked around the car twicce and didnt make a hit .. then the handler walked up to the trunk.. tapped on the trunk really hard and called the dog by name .. and called it a hit ... there was stuff found in the car that was narcatics .. I guess the cops can do as they please without any worry about the way they do things ...

[defcon29]

This is to Phoenix Rising, who posted earlier: (quoted him on bottom of my post)

Dang, I did not look at the link you provided, but either you or your source is talking out of his behind about a case that they can hardly comprehend in the slightest! If you are going to talk about Supreme Court cases, especially one of the more important ones regarding the 4th amendment, please read a simple link like this so you don't spread false information, and don't make yourself look like an imbecile: http://www.norml.org/index.cfm?Group_ID=4075

First off, The Supreme Court ruled that the use of infrared thermal imaging was unconstitutional. They describe the act of imaging that occured as opposite of what a drug sniffing dog does; whereas a dog will alert in immediate vicinity of contraband only (in theory); and the imaging "alerted" from every single source of radiation around it. That would've been a tragedy if the ruling was overturned.
In regards to your technological massacre, read this to find out what thermal imaging is: http://en.wikipedia.org/wiki/Thermography#Thermographic_imaging

Thermal imaging is radiation converted to images, the radiation changes in relation to heat. Has nothing to do with light eminating from someones house! And they aren't listening to a CB, so there wont be any "Frequencies" that they are going to tune in to. And you were wrong about the "sniffing" part, nowhere was sniffing even mentioned in the decision.

Here is the original shenanigan by Phoenix rising, nov. 29, 2006:

"actually, a sniff isn't a search. You've got no privacy rights regarding what is out in public, including your scent.

(there was a really interesting supreme court case a year or two back about whether the principle applied to cops using infared detectors to look for evidence of marijuana production in homes--the argument was that you've got no expectation of privacy regarding light emanating from your home, and if the police look at a different frequency that happens to indicate likely criminal activity, there's no privacy violation [i suspect I'm getting the technology wrong, but I think I have the law right]. I can't remember what the outcome of the case was, actually, but I'm pretty sure the right to smell people without a warrant was the most important precedent the government relied upon)"

[Michael Pace]

I realize this is kind of an old article, but I was just wondering about this the day before yesterday when I was stopped by a K9 unit for my left driver's side mirror being gone. I was immediately asked to step out of the car and pressed if I had anything illegal in the vehicle. I did not, and I told him so. He asked if he could search the car and I told him no, I was on my lunch break and didn't have time for this. He then asked again if I had anything illegal and if he could search. I again told him no.

So out comes his dog. He walks him around my car twice as the dog sniffs the tires, the door handles, the license plate, etc. The cop then says that the dog had "alerted". So now he has probable cause to rifle through all my belongings for 15 minutes. To his massive disappointment, he didn't find anything.

So I don't know why the dog alerted, but I do know I didn't have any drugs in the car.