A recent article about lie detectors brought up a related topic; drug/bomb sniffing dogs or the more general use of police service dogs.  Let me begin by acknowledging a bias; I love dogs.  There seems little doubt regarding the utility of these animals in search and rescue, body recovery, and even sniffing contraband and explosives.  I want to be clear that this is not an attempt to diminish the importance and the role of these dogs in many of the activities they engage in.  Regardless of their error rates or even false positives, the assistance provided by these animals can be invaluable.

So what's the issue?

As always, the problem occurs when there is a presumption of infallibility to these animals and then it is used to infringe on people's rights and due process.  Specifically one of the primary issues currently being reviewed by the Supreme Court is whether law enforcement has exceeded their purview in the use of these dogs to conduct illegal searches (1).  For a more complete discussion of the legal ramifications, you can review the links provided at the end of this article.  From the scientific perspective, we should be concerned that there is a tacit assumption that the dog is a tool or a technological device that performs on demand and is error-free.  This is flatly wrong.

First we have to consider that there is a huge difference between what a dog can do, versus what a dog actually does, and this is often dependent on the training received.  To date, there are no required national standards for certifying a dog as a narcotics/bomb sniffing dog, or anything else.  Each company that engages in training dogs can set their own standards and certify according to those.
...reports that only two states have standards for dog certification. In Canada, only British Columbia does, according to Barry, a private trainer and handler of three drug-sniffing dogs. He says that he and others are working with the Department of Labour and Safety to change that, but “at the moment, it is possible to advertise that any dog is a detection dog, charge people and companies a fee, and show up and provide a service with no documentation whatsoever to indicate that the dog and handler are competent!”
http://www.slate.com/articles/news_and_politics/jurisprudence/2012/10/drug_sniffing_dogs_the_supreme_court_will_hear_argument_on_whether_they.html
Furthermore we must consider that regardless of how effective or ineffective a dog may actually be, it is also a creature that is capable of acting on its own behalf and subject to the same type of good/bad attitudes a human might feel.  Would one consider a dog that is abused by its handler to be in the best condition to respond to commands and perform tasks?  Yet, despite the more romantic notions about human/dog bonds, this is much more prevalent than many would like to believe.
A member of the town council in a small village in Ohio told another tale of woe. His council hired a dog handler and dog team and paid more than $7,000 for additional training, only to be told that the dog could only do search and rescue, because the handler had mistreated the dog rather than training him properly.
http://www.slate.com/articles/news_and_politics/jurisprudence/2012/10/drug_sniffing_dogs_the_supreme_court_will_hear_argument_on_whether_they.html

Ten (10) K-9s have been killed in over-heated law enforcement vehicles this last summer. This alarming trend must stop. A K-9 handler cannot ignore the ramifications of killing a K-9.
http://www.nndda.org/nndda-library/82-news/324-overheat
As a result, without truly understanding the psychology of the dog involved, it becomes impossible to ascertain the motivation for any particular response.  Certainly we would like to believe that these animals are selfless and devoted to satisfying their handlers, yet that in itself could be problematic.
The overwhelming number of incorrect alerts identified across conditions confirms that handler beliefs affect performance.  Further, the directed pattern of alerts in conditions containing a marker compared with the pattern of alerts in the condition with unmarked decoy scent suggests that human influence on handler beliefs affects alerts to a greater degree than dog influence on handler beliefs.
http://www.springerlink.com/content/j477277481125291/fulltext.pdf?MUD=MP

http://esciencenews.com/articles/2011/01/31/explosive.and.drug.sniffing.dog.performance.affected.their.handlers.beliefs
One of the most contentious elements of the legal argument is the lack of reliable data regarding the error rates for these canines.  Despite a long history of using canines, there is a remarkable lack of scientific data to back up our assumptions.
The court has recognized the fact that “false positives” occur and dogs can be as low as 62% accurate.
http://www.asctk9.org/id55.html
A recent study of three years of data from an Illinois police precinct found a 56% erroneous alert rate. Only 44% of the alerts resulted in discovery of drugs, and the error rate was even worse for Hispanic drivers (suggesting that human handlers have some influence over the process).
http://www.stanfordlawreview.org/online/war-on-drugs-privacy-law
Similarly, when Robert Yell was accused of intentionally burning down his home, killing one child, and severely burning another, probably the most damaging, prejudicial evidence was that PJ, a sniffer dog, “alerted” at six locations in Yell’s burned-out trailer. According to PJ’s handler, this behavior meant PJ had detected accelerant at each and every one of the six locations. Using state-of-the-art science, the lab was unable to confirm PJ’s opinion, and the six samples all came back “negative” for accelerant. Despite this, the Court in Yell found there was a sufficient showing of reliability for PJ the dog. PJ’s opinion that accelerants were present, and the fire was no accident, was presented to Yell’s jury. The jury convicted Yell of intentional arson, ...
 http://apps.dpa.ky.gov/library/advocate/pdf/2008/adv030108.pdf

At trial, Cannon testified that after arriving at the fire scene on the morning after the fire and calibrating PJ, PJ went into the trailer and alerted to six locations for the presence of ignitable liquids.   Contrary to his earlier ruling, the trial court allowed Cannon at trial to identify the locations of the six alerts in the trailer.   As stated above, PJ again alerted to the presence of accelerants in the six separate samples taken from the scene, but did not alert on the control sample containing no accelerants.   However, the testimony of Kenneth Rider from the Kentucky State Police forensic lab established that all six samples tested negative in the laboratory for the presence of ignitable liquids.
http://caselaw.findlaw.com/ky-supreme-court/1167613.html
As I stated at the beginning, none of this is intended to disparage the widespread use of such dogs when the purpose is to assist humans as supplementary assets.  However, when the dog's behavior becomes evidence in the legal realm and assumptions are made about what the dog "meant", or the mythical belief in canine abilities prevails, or where issues of reliability are simply ignored because of handler bias, then we have a problem (2).

For a discussion regarding the legal issues involved in using police canines.
http://cjr.sagepub.com/content/35/4/438.full.pdf+html

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2116389

Analysis of Probable Cause

===========================

(1)  For those that don't think this presents a problem, consider that the issue in front of the court is to determine whether a K-9 unit can simply walk up to the door of your house, sniff around, and based on the interpretation of the dog's alerts, establish probable cause and obtain a search warrant.  In addition, if challenged in court, neither the dog nor the handler are required to establish the reliability of the dog's performance nor provide documentation regarding the dog's training. 

For those still not convinced, instead of the dog consider if the police had parked in front of the house with a thermal imaging device [without a warrant].

(2)  Consider this telling court exchange defending the K-9.
"Q.  Do you keep records as to the effectiveness of your dog?"
"A.  Yes, sir, I do."    
"Q.  Do you know how often your dog gives false positives?"
"A.  He doesn't give any false positives.  We're just unable to verify the alerts at the time."

Transcript of Record at 7-8, Commonwealth v. Fens, No. CR03F01831-01 (Chesterfield Cir. Ct. Jan 06, 2004)
http://heinonline.org/HOL/LandingPage?collection=journals&handle=hein.journals/nyujlawlb1&div=43&id=&page

It is also worth considering the following link where one can't help but feel the sense of disdain for people that insist on due process.  As stated previously there is no problem regarding canine reliablity if the purpose is simply to supplement human investigators, however it becomes an entirely different matter when the dog's sense of smell becomes evidence in it's own right and cannot be cross-examined or validated in a court of law. 
http://positivepolicedogs.wordpress.com/tag/sniffer-dog-reliability/