Wednesday, July 04, 2012

Poor training, low-quality certification and high error rates dog police canine units

Via Simple Justice.
A couple of recent stories from out of state have brought drug sniffing dogs back into the news, as the US Supreme Court prepares to consider their use at people's homes without probable cause this fall.

Dog-sniff evidence is problematic because dogs can't be cross examined. There's wide variation in their accuracy, not to mention the quality and quantity of their training. And dogs naturally pay careful attention to cues from their handlers, creating a significant risk of false positives. Since no attorney will ever get a dog on the witness stand, the other elements (accuracy, training, and methods to avoid cues from the handler) have been critical to the creation of probable cause in previous court decisions. Unfortunately, the bar on all those measures is lowering rapidly.

Courts already accept dog "alerts" as a legitimate basis for a search, even when the dog regularly "alerts" where no contraband is found. In this story out of Roanoke Virginia, a dog name Bono found drugs "just 22 times out of 85 'alerts'," or 26% of the time, and yet a judge upheld the search. Arguing that the dog's performance was actually better than the statistics tended to show, prosecutors in Bono's case argued the dog should get credit for some "hits" that found no contraband, claiming that: "In some cases where nothing was found after an alert by Bono, police later determined that drugs had been in the vehicle earlier, likely leaving an odor the dog was trained to detect." The judge accepted that argument, but even granting such a (highly dubious) premise, Bono's alerts were clearly mistaken half the time, and in another quarter of his hits he's detecting traces that did not actually indicate drugs were present. That fact undermines the US Supreme Court's past rulings (e.g., US v. Place) which upheld dog sniffs in part because they disclose "only the presence or absence of narcotics, a contraband item." In Bono's case, his "hits" mistook the absence of drugs for their presence nearly 3/4 of the time.

Detecting drugs that used to be in the location is  not the same as detecting contraband presently in a suspect's possession. In Texas, Deputy Keith Pikett claimed his dogs could distinguish smells in "scent lineups" up to 17 years after the sample was gathered, so if you take that preposterous claim at face value (as many Texas prosecutors have) heaven knows how long the lingering odor of marijuana might trigger a "hit"? If I buy a car from a pothead and years later a cop runs a drug dog around it, will the dog still signal that marijuana is there? Should that constitute probable cause to search my car today? That's the logic of the Roanoke court.

The judge in Bono's case upheld the dog sniff as probable cause mainly because,  while the dog "may not be a model of canine accuracy," police were justified in acting on it because of the "dog's training and flawless performance during re-certification sessions." Unfortunately, even flawless performance means different things to different agencies.

A case out of Florida which will be argued before the US Supreme Court this fall will determine whether drug-sniffing dogs may be used to glean probable cause outside someone's home without a warrant. The Florida Supreme Court's ruling being challenged by the state made the following argument to say mere certification was insufficient to demonstrate dogs' competence, because:
conditioning and certification programs vary widely in their methods, elements, and tolerances of failure. Consider, for example, the United States Customs Service regime: 

The Customs Service puts its dog and handler teams through a rigorous twelve-week training course, where only half of the canines complete the training. Customs Service dogs are trained to disregard potential distractions such as food, harmless drugs, and residual scents. Agents present distractions during training, and reward the dogs when those diversions are ignored. The teams must complete a certification exam in which the dog and handler must detect marijuana, hashish, heroin, and cocaine in a variety of environments. This exam and the following annual recertifications must be completed perfectly, with no false alerts and no missed drugs. If a dog and handler team erroneously alerts, the team must undergo remedial training. If the team fails again, the team is disbanded, and the dog is permanently relieved from duty.
In contrast, the testimony below disclosed that Razor and his handler had undergone just one initial thirty-day training course and one week-long annual recertification course. In neither course was Razor conditioned to refrain from alerting to residual odors. Whereas the Customs Service will certify only dogs who achieve and maintain a perfect record, Razor's certification program accepted a seventy percent proficiency. These disparities demonstrate that simply characterizing a dog as "trained" and "certified" imparts scant information about what the dog has been conditioned to do or not to do, or how successfully. (Emphasis in original, citations omitted.)
CBS News' 60 Minutes addressed the subject of low-quality training for sniffer dogs in a 2009 story:
"We don't have a set of standards," says Auburn University professor Larry Myers, one of the country's leading experts on canine detection programs.

He believes there is insufficient regulation and no universal tests for the dogs: "We honestly don't have a set curriculum. We've got to get everybody up to some minimal agreed level."

He also says there are good programs that turn out reliable dogs, but some produce dogs that make lots of mistakes -- and even tell lies.

"They can tell you that something's there, that's not there, simply to get praise, to get food, to get whatever they're working for," says Myers, who adds that canine programs are supposed to train dogs specifically to avoid that problems like that. "It is a tremendous problem. We have trainers that can't train. Dogs are being used that can't --don't know how to do anything."

Do trainers have to be licensed? "In some programs, yes. And in most programs, no," says Myers.

This means that poorly trained dogs and handlers are working in many parts of the country.
Meanwhile, Scott Greenfield at Simple Justice brings word of a drug-dog case from across the country in Nevada that should give everyone pause. Nevada state troopers have sued their superiors, alleging that:
the drug-sniffing dogs used by troopers in the program were intentionally being trained to operate as so-called trick ponies, or dogs that provide officers false alerts for the presence of drugs.

The dogs were being trained to alert their handlers by cues, instead of by picking up a drug's scent by sniffing, the complaint said. When a dog gives a false alert, this resulted in illegal searches and seizures, including money and property, the complaint said.
Dogs naturally pay very carefully attention to their people. They've been doing so for millenia. In Nevada, troopers are accused of intentionally teaching the dogs to alert based on their cues. Which raises the question: How can any court ever determine whether a dog alerted to the smell of drugs, an intentional or unintentional cue from their handler, or for that matter the package steaks recently purchased by the suspect? As pointed out in a recent article from the National Association of Criminal Defense Lawyers' trade journal:
Dogs are not motivated in the same way as humans. Dogs have no interest in ridding the world of illegal drugs. Dog trainers, including police K-9 trainers, use treats, toys and praise to reward dogs when they do what they have been conditioned to do. If a police K-9 alerts, it gets a reward. K-9 handler/trainers know this, and the dogs quickly learn that an alert results in a reward in most instances, even if nothing is found. 
Even if such cues are not intentional, the fact is that dogs seek to please their handler raises questions of inherent bias, particularly when the dog's "hit rate" is as low as Bono's above. As the NACDL article pointed out, "If law enforcement or magistrate judges, who are presumed to be impartial, were to be similarly incentivized, it would constitute violations of the Fourth, Fifth and Fourteenth Amendments." To the dog, after all, it's all just a game in which they want to find reason to alert in order to get a reward. To law enforcement and those targeted, though, it's not a game at all, even if federal jurisprudence on the subject has been mostly a joke.

19 comments:

Anonymous said...

Check out the recent UC Davis study. It shows how inaccurate these dogs really are.

Anonymous said...

Sounds like a bunch of drug dealers and couriers who got CAUGHT who are doing all the complaining! Figures.

Gritsforbreakfast said...

Actually 5:42, I'm complaining and guess what? Not a drug dealer.

And of course, the Florida Supreme Court "complained," too.

Hook Em Horns said...

Pull me over with your mutt 5:42 and get ready for one hell of a ride!

Vincent van Gogh said...

Over the years it has been routine for drug dogs to be run through the parking lots at TDCJ facilities. More than once I was called to the parking lot to unlock my truck because the dogs alerted on it. The questions were always the same. "Is there anyone else who drives your vehicle or do you have total control over it"? "Can we search your vehicle"? Every time this happened what was found in my case was a McDonald's bag with a half eaten hamburger in it my children left in the truck. I asked on more than one occasion, "do the dogs alert on food", and the answer was always, "Oh no that would never happen". In fact, every time the unit went through one of these drug dog searches the dogs alerted on several cars and nothing was ever found. Have you ever seen these drug dogs? They are skin and bone. In my opinion they alert falsely on lots of scents and food is at the top of the list. Of course they make mistakes!

Anonymous said...

They should just throw us all in jail on general purpose and be done with it, since that's their ultimate goal anyway.

Anonymous said...

"Dog-sniff evidence is problematic because dogs can't be cross examined." Really?

Not a drug K9 handler out there worth his salt that wouldn't bring his dog to the judge, let the judge hide the drugs and then sit back and watch.

Why do drug organizations and some amateurs lace their loads with coffee, detergent and the likes or go to the extent of using false compartments in gas tanks?


The Pickett thing....I would have to see that one.

Red Leatherman said...

@5:42 I've been detained and stood out in the dark on a very cold night on the side of the road with my wife over an hour was searched physically and our car practically disassembled due to a false alert and I am complaining too. It wasn't just inconvenient. The excuse? All of North central Texas is a high crime drug trafficking area.
@8:56 How about do the demonstration for the Judge without drugs instead? Looks like the dogs are almost just as likely to alert anyhow.

Gritsforbreakfast said...

8:56, when they do that demonstration for the judge, the protocol should be to hide the drugs in one spot, and then eight McDonald's happy meals elsewhere around the room. Take the dog off the leash and see which one it "alerts" on. The Customs dogs are trained (and rewarded in training) to ignore the happy meal: Is yours?

On the Pikett thing, I linked in the post to an item on Anthony Graves' case where they tried to use his dogs on 17-year old scent evidence. And if you don't think dogs make false accusations, ask this Victoria Sheriff's deputy who was falsely accused before DNA cleared him.

Nobody's claiming dogs are always wrong (or more accurately, their handlers, since the dog can't testify), but they're not infallible, some fail far more than they succeed, and a "hit" does not necessarily in any given case indicate the presence or absence of drugs. In Bono's case, in fact, when his handler said there was a hit it meant there's a 74% chance there are NO drugs: The question is, does that amount to probable cause?

The dogs aren't the problem: The problem arises from flawed humans assuming they can read Fido's mind.

Anonymous said...

@8:56 How about do the demonstration for the Judge without drugs instead? Looks like the dogs are almost just as likely to alert anyhow

No problem Mr. Red. Like I said, any handler worth his salt won't back away from the challenege.

Little lengthy but worth scanning this link.

http://www.nndda.org/

john said...

OF course--they train cops to lie, because you can lawyer up. Judges lie, because you can appeal. That's the thinking. So dogs are invalid, but it's expedient and revenue-raising and face-saving for the dept/cops/courts/police state.

And the NRA et al., stand down.

But not Captcha cha, it's always blinding.

Gritsforbreakfast said...

"any handler worth his salt won't back away from the challenge"

How 'bout the Happy Meal challenge, 8:00? If you let your dog off the leash, will it ignore the eight happy meals and go for the dope?

David O'Niell said...

In the federal system, dog and handler records are sealed for "National Security" reasons. BS- the alert is subjective, and how can you trust an animal who is locked up all day until 'play time'; i.e. their reward for pleasing the handler is interaction w/ said human?

dahowa said...

5:42: I guess what you mean to say is that you don't believe we have the right to be secure in our papers and possessions. Of course only the drug possessors complain because they're the only ones with a forum to complain. I am offended by the very notion that our 4th Amendment rights rest on the nose of a dog.

Red Leatherman said...

The nndda site had an interesting fact on the narcotic detection standards.

They use a minimum of 10 grams of powder and a minimum of 1/4 oz marijuana per stash.

For certification "There will only be one (1)negative response allowed".

Seems with a 50% - 80% error rate, by counting on my fingers, ends up with using Bonzo in something close to two (2) scenarios before he's somewhat beyond cert standards.

Gritsforbreakfast said...

Red Leatherman, I also noticed a couple of other things in their Narcotics Detection Standard (pdf).

First, if they fail the dog team can take the test again as many times as they want, whereas the Customs Service retires a dog if it fails certification twice, with 100% accuracy required.

Also, they do NOT require certification on ignoring non-drug smells, so dogs trained under those standards aren't trained to pass Grits' "happy meal" test.

Anonymous said...

The problem with relying on organizations like the nndda is that they have an inherent bias. They only exist as an organization if these dogs are reliable. If the use of these dogs fell out of favor the organization would no longer exist. Radley Balko made an astute observation regarding the fact that psychics also have a certifying organization. I'm sure this organization also asserts that psychics are reliable.

A more reliable source of info is the UC Davis study: http://www.ucdmc.ucdavis.edu/welcome/features/2010-2011/02/20110223_drug_dogs.html This was a very well done study by people who didn't have a dog in the fight, so to speak. The results were very interesting. "The study, published in the January issue of the journal Animal Cognition, found that detection-dog/handler teams erroneously “alerted,” or identified a scent, when there was no scent present more than 200 times — particularly when the handler believed that there was scent present."

There is other research out there for those that are interested that really shows the reliabilty of these dogs is much to low to say that they establish probable cause. Of course, if one only wants to defend the use of these dogs they won't bother to look at legitimate research done by unbiased research and will only rely on the info provided by the certifying organizations whose very existence depends on showing that the dogs are reliable.

Anonymous said...

I've been studying this issue and I believe that, at least concerning vehicle searches, most of it collapses into the following questions:

Q1: At what rate does a particular K9 unit "alert" when no drugs are then found?

Q2: Does a defendant have the right to the information in Q1 concerning the K9 unit used to search his vehicle?

Q3: If Q2 is 'yes', then are defendants routinely provided with this information prior to being offered a plea deal by the prosecutor?

Q4: What mechanisms are in place to ensure that police are properly recording the accuracy rates of their K9 units?

Q5: Does the public (ie, The Press) have a way to verify that the mechanisms in Q4 are working properly, and can this verification be obtained with a reasonable amount of effort?

* I recently spoke with a defense attorney in my city in Ohio, and he told me that while the accuracy rate of a particular K9 unit can, in theory, be obtained, in practice it can never be obtained prior to the prosecution's offer of a plea deal. Obviously, this is of no evidentiary value to the defendant and is therefore worthless. Furthermore, this attorney told me that while officers are supposed to catalog K9 units hits versus misses, they routinely "forget" to instances where K9 units alert but no drugs are found.

What this all adds up to is that the current use of K9 units to establish probably cause for a vehicle search essentially amounts to a "general warrant". General warrants were what the British colonial forces used to search the homes of early colonial Americans and were the motivation for our constitutional protection against "unreasonable search and seizure". A current analogy to the K9 vehicle search situation would be if you were defending yourself against a DUI charge, and your attorney asked to see police records proving that the machine used to test your blood had been properly calibrated, and the police refused to produce these records until after you either accepted or turned down a plea deal. Obviously this situation wouldn't fly, and the current situation with K9 units shouldn't either.

Anonymous said...

You really are a jackass. Never mind rights being violated.