Fascinating research summarized in The Economist. Basically, detecting dogs respond to unconscious cues from their handlers, and generate false alarms because of them.
It makes sense, as dogs are so attuned to humans. I’ll bet bomb-sniffing bees don’t make the same mistakes.
EDITED TO ADD (3/14): Research paper.
Dilbert • March 4, 2011 7:35 AM
Hmm, Russian spammers are out early today. Or maybe it’s late in their time zone? 🙂
Clive Robinson • March 4, 2011 7:42 AM
Let me be the first to get it out of the way,
To bee or Not to bee that is the question?
Wether it bee nobler in the hart to suffer
the stings and harrows of outrageous fortune.
I’ll let somebody else do the “let lose the dogs of war”.
Clive Robinson • March 4, 2011 7:47 AM
On a more serious note.
Dogs are pack animals they see humans as either a pack leader or something to be subordinated/attacked.
Most humans also have a “pecking order” where those lower seak to please those high to hope by asociation to improve their ranking and thus can “please” to the point of obsequiousnes.
Why should we expect dogs to be any more independant of the boss than any other subjugated salaryman?
Andre LePlume • March 4, 2011 8:06 AM
The problem lies with the humans. I am sure cadaver (or rescue) dogs do not alert falsely as much precisely because their humans are much less likely to be biased. One tree is as good as the next for the lost hiker to be unconscious against, and no tree has beady eyes and glances furtively. Where the humans think they already know the answer, you get trouble.
Moreover, we’ve had way more time to selectively breed dogs for “real” sniffing work in environments dogs understand – outdoors searching for something. Whether that skill is transferable to finding contraband/chemical traces in a noisy indoor urban setting packed with other smelly mammals I’d be suspect of.
Prescott • March 4, 2011 8:07 AM
“Drug-Sniffing” dogs are an easy way around American 4th Amendment legal protections for ‘probable-cause’ in police searches.
Police often bring in a K-9 unit when they want to search a particular vehicle after a routine traffic stop, but the driver rightfully declines permission for a warrantless police search. Walking the sniff-dog carefully around a vehicle is somehow considered not to be a police search procedure.
If the dog “signals” its handler that it has ‘detected’ illegal substances — the courts generally consider that sufficient “probable cause” to then conduct a full police search of the vehicle without any other indication of wrongdoing.
And if the dog signals … and nothing actually turns up in the full search– there’s no penalty to the dog, cops, or supposed validity of the dog-search procedure. Of course, the hapless citizen-driver has no real legal recourse either.
The real issue of ‘false-positives’ by the dogs, subtle sub-conscious signals given to the dog by the handler, or outright lies, falsifications, or ‘planting’ of evidence by the cops in the conduct of the sniffing/search/reporting of the incident — is never addressed by the courts.
Police (and military) dogs are always honest & trustworthy before the law (?)
So if these sniffer bees ever escape into the wild we’ll get honey made from Semtex? Somebody warn the TSA, time to ban honey packets from the 3-1-1! Remember, it’s never too soon to overreact!
@ The Inestimable Clive Robinson
Couldn’t have said it better myself Mr. Robinson. Fine writing indeed. You look different somehow, have you been working out? Have I ever told you how much I enjoy being a co-poster on this blog with you? Say, that’s a nice tie…
Ian Mason • March 4, 2011 8:18 AM
For anyone who prefers original sources rather then second hand rehashes of them, the original paper is at http://www.springerlink.com/content/j477277481125291/fulltext.pdf . And unusually for URLS that include the word “springerlink” it’s not behind a paywall.
Clive Robinson • March 4, 2011 8:35 AM
@ mcb,
“You look different somehow have you been working out”
Hmm as I type this from my hospital bed (sadly true) if I was your boss I might suspect you where not being quite sincere 😉
However thanks for the post it has raised a smile on my face 8)
@ All
Have others thought about this animal behaviour and training.
It all appears to be based on “rewards” of some kind (often food) and in some cases (the original BF Skinner) by actually starving the creature.
Now I don’t know about you guys but my best work is rarely done on an empty stomach and even a person of my known astronomical abilities in this area can only eat so many “wafer thin mints” before I start to feel nauseous or people head for cover 😉
Why should we expect creatures to be any different.
So if you have illicit substances perhaps it would be best to time your arival at some point where the beast has been just rewarded or fed.
No One • March 4, 2011 9:22 AM
There’s a two-word answer to anyone that claims their animal can do something extraordinary: Clever Hans.
If you can’t prove your animal isn’t just as clever as Hans then your animal-based method should be ignored.
Unfortunately, within a statistically zero probability, no one understands statistics so why should we expect lawmakers and law evaluators to?
uk visa • March 4, 2011 9:59 AM
@Prescott
Good point… I’d never considered that as an aspect.
@Clive
Obedience: for so work the honey-bees.
@ Clive Robinson
Sorry to hear you’re laid up. I hope you have one of those self-serve morphine buttons with which to while away the hours…
Get well soon and please keep your hospital gown tightly tied in back.
Petréa Mitchell • March 4, 2011 10:10 AM
Actually it’s not clear from the Economist’s article whether the dogs are responding to handlers’ cues, or the handlers are interpreting ambiguous behavior as a “hit” or “miss” based on what they think it should be.
Petréa Mitchell • March 4, 2011 10:12 AM
And if it’s the latter, bomb-sniffing bees will be subject to exactly the same problem.
mabbo • March 4, 2011 10:23 AM
Knowing only the false positive rate when the trainers are given false information is interesting, and of some use, but I would have preferred if these tests also included true positives.
Consider: what if the same tests were done again, but now half the time, in some part of the search area, there was a bag of cocaine or PETN hidden. What kind of true-positive rate would we see? Would the real drugs and explosives be skipped over for the red tape?
If the dogs ignore real targets in preference of handlers’s ideas of fake targets, that would be a far more significant result, and would truly be a blow against the use of dogs. If the dogs ignored fake targets despite handler’s false information, then the study would instead support the use of dogs.
Incomplete science like this leaves me wondering about the intention of the research, and what goals they wanted to find.
NobodySpecial • March 4, 2011 10:26 AM
I don’t think anybody appreciates the amount of training that these dogs undergo.
Long hours spent late into the night in the library studying 7th circuit appeal court judgements on what constitutes an illicit drug. The chemistry knowledge to decide if an isomer of MDMA constitutes the same legal structure if it has the same metabolic effect.
And learning not to pee on the police cars.
Steven Hoober • March 4, 2011 10:42 AM
I thought this was already well-known. Users of dogs who are serious about making a case segregate everything as much as possible. For example, check luggage and other goods or personal belongings in a group, without the owners nearby. If inspecting a locker room, make sure there are no personal identifying characteristics visible to the handler. AND, the handler is not the investigating officer. As little contact between the two as possible beforehand.
Sure, it can be abused, but eventually case law will catch up and it’ll get thrown out when stupid. Then, they’ll stop doing it the stupid way as there’s no point.
Ryan Cobine • March 4, 2011 10:54 AM
I am not a dog trainer, but the ones I have worked with clearly understand that what this research shows in a controlled way is absolutely true. They emphasize this in their handler training. They take pains to point out that the notion of a scent detection dog as an “explosive detecting machine” is a misleading misunderstanding, and that any evaluation of canine scent detection must treat the dog and handler team as the unit of study. They try to impress upon the handlers that the dogs are tools in the hands of humans who must take steps to use these tools appropriately and correctly, and must also maintain the sensitivity of the tool through ongoing training. The ability of dogs to detect scents in the parts per trillion range is not the issue, as with most security issues, it is the human psychology and behavior that is the most significant factor–and the trickiest to get a handle on!
scottnottheotherscott • March 4, 2011 10:57 AM
This just in – Stunning scientific revelation indicates dogs are adaptable to their environment.
Early research indicates that canine behavior can be modified in accordance with human behavior. Internet Dr. scottnottheotherscott says that laboratory results conclusively prove that this ‘training’ effect can alter a dog’s behavior.
“Dogs can be taught to respond in a predictable manner to both cues from a handler and to their external environment.” says the Internet Dr. “My research on such things as the ‘rolled up newspaper’ and ‘Do you want a treat, Boy?’ effects indicates that they are very adaptable indeed.”
Tongue out of cheek, it’s always good to remind ourselves that training is fallible and dogs themselves are fallible.
The old saying comes to mind, for dogs both talking and sniffing. It’s not that they do it well, but that they do it at all, that is remarkable.
boog • March 4, 2011 11:12 AM
@mabbo
I don’t think the study should be discarded as “incomplete science”. It’s certainly reasonable to go in with the intention of finding flaws. Otherwise, how would we ever improve existing processes?
Do you think they are using this information to discredit or eliminate bomb-sniffing dogs? If that’s the case, then yes, the study is incomplete. But I don’t have any reason to believe that’s what they’re doing.
I believe they are using the research to identify ways to improve the method. At least it seems that way to me; they not only identify that there are false positives, but also try to understand why. Now they can take proper steps to improve their methods and reduce false positives. Surely you’d agree that reducing false positives would be a good thing, right?
BF Skinner • March 4, 2011 12:51 PM
@Clive
“Hath not a rat, pride? If you prick us, do we not bleed? If you tickle us, do we not laugh? If you poison us, do we not die?”
BF Skinner • March 4, 2011 12:57 PM
“incomplete science”.
Since when is science ever complete?
You can have good science and bad science but complete? Nah.
It’s A study. Science says reproduce the results. Look for the flaws.
What I don’t understand is why lawyers are not contesting the validity of searches with dogs in court?
(though when your hearing for a drug felony is 20 minutes long probably not that much time to challange the long held notion that dog hearing is better and dog noses are better)
You all had me convinced a couple of weeks ago when you pointed out dog signal to handler is a secret channel. That channel can’t be questioned in court. We can’t cross examine the dog for his motive or impeach them for lieing to please their handler. Let alone the handler’s concious and unconcious bias.
Clive Robinson • March 4, 2011 1:32 PM
@ BF Skinner,
“hath not a rat pride?”
Have you seen my whiskers recently?
The amount of grooming I’ve put in I should get “first in class”.
And yes the nurses are proving atleast three times a day that I do indead blead when they prick me, I’ve more pinholes in my paws than a second hand pincussion.
As for being tickled this current crop of nurses don’t know me that well 8(
But they do have me on the rat poison(warfrin) but I aint dead yet (just overly anti-coagulated)
Prescott: Not only that, but IMO dogs also violate due process and the confrontation clause. They are a travesty.
@mabbo, I’m guessing the dog would be more biased to detect the real contraband has it has had more repetitions to do that.
but I suppose a strong enough message might change the dogs mind in the hear and now
“Now they can take proper steps to improve their methods and reduce false positives.”
Impossible unless they can breed out the ability for dogs to respond to subtle cues from their handlers.
Richard Steven Hack • March 4, 2011 7:00 PM
Clive: You’re only pricked 3 times a day? I once volunteered (when broke) for a medical study of blood pressure reduction medicine. My blood pressure was already so low they thought I was already on it. Every time a nurse has ever put a cuff on my arm, they get a puzzled look, then ask me if I’m on low blood pressure medicine.
But once in the study, they put a catheter in my arm and drew a vial at least four or five times a time or more (I can’t remember the exact number, might have been six). And also at 4 in the morning, too. Would go in and out of the hospital a couple times a week staying for a few days. This went on for six weeks. Made $1700 out of it, though.
Good times…not.
Roobie-roo • March 4, 2011 8:32 PM
@EH: Prescott: Not only that, but IMO dogs also violate due process and the confrontation clause. They are a travesty.
Ruh-roh, a Rontirutional rawyer.
Beta • March 4, 2011 11:27 PM
@mabbo: real explosives are very rare (and real narcotics pretty rare I suppose) so a false positive rate this high will swamp the true positives no matter what. This is enough to show that this method should not be called “probable cause”.
The false negative rate you ask about doesn’t matter unless we are willing to use this method regardless of the violation of civil rights and wasted {time, effort, expense}. And if we’re willing to disregard those things then the accuracy of dogs becomes kind of academic.
@Beta, I don’t think they would weaken the accuracy of the dog by teaching it false negatives on the job, when in situation it could make a real difference.
If the dog wasn’t any use it wouldn’t stand up in court or be allowed(some researcher would have a document stating that fact,accuracy)
Clive Robinson • March 5, 2011 5:26 AM
@ Richard Steven Hack,
“But once in the study, they put a catheter in my arm”
I hope not that would mean you had an “existing portal” and you realy would have been in poor shape 8( unless of course you have a stomata of saintly proportions 😉
Catheter / cannula are words script writers often get mixed up in films etc especially as once in place a cannula is “technicaly” a catheter (so pays your money takes your choice which is why they tend to get called “lines” these days).
Further when you look at the modern verisons they are both (usually soft and flexible) plastic tubes that are used for transporting liquids in or out of the body (some medical staff often think incorectly “cannula for in” like an IV line, “catheter for out” such as for urine out of your bladder but that is not correct).
The name all revolves around history and how the line gets to where it is supposed to go…
Cannulas are the new boys on the block which you “guide from outside” and are a much later invention than catheters which “follow an existing path” supposadly “without guidence” just gental pressure.
Usually a “cannula” has a “trocar” (the exception being the nasal cannula) which is a sharp tip or these days what looks very much like a syringe needle down the middle that is used to make the portal and guide the end of the tube to the desired site of drainage etc. And in modern non reusable cannulas it’s removed after the line is placed. This is what they should have used to get your blood out 😉
Up untill quite recently the business end of the line that was in your body was a metal tube that could be (if you were lucky) not just sterilized but sharpened as well…
A catheter is a much older invention (the Romans are known to have used them), it lacks the trocar and must go in to an existing presumed natural opening into the body or be inserted by surgury ie incise with a knife or lancet place the catheter and apply preasure to seal around it.
These days the catheter is most often seen used for the collection of urine or other bodily fluids etc the body would normaly expel if things where working normaly 😉
As already noted the “nasal cannula” is the “fly in the ointment” and gives the real clue as to the historical name usage. This is because although it goes in a natural openning and follows an existing path like a catheter, it needs to be “guided” into the right sinus etc as an integral part of it’s placment that makes it a cannula even though it does not have a sharpend trocar.
How do I know this arcane knowledge that few others including the medical fraternaty know? “bloody mindedness”.
Few nurses or doctors “know” as they “absorbe” the names during training and the difference is seldom if ever said as I suspect those doing the training don’t know either such is teaching by rhote….
It is only by researching the history of medicine you find out… Oh and many explanations up on the internet including the wiki entry have got it either wrong or only partialy (ie trocar) right.
And the reason for my “bloody mindedness” is when I asked people in the medical proffession they gave a “hand waving” answer and that’s like a red flag to a bull with me it means they either don’t know which brings their abilities into doubt or they are fobbing you of which also brings their professionalism into doubt.
And lets be honest when the seed of doubt has been sown who wants a poorly trained non proffessional working on them?
A few issues regarding the article “Handler beliefs affect scent detection dog outcomes” published in the journal “Animal Cognition” (January 2011). http://www.springerlink.com/content/j477277481125291/
Service dogs (e.g. SAR, bomb/drug detection) work as a team with their handlers who, based on their years of expertise, guide their dog to the most relevant areas and items to search. There is no comparison with Clever Hans who responded to inadvertent cues from his trainer.
Half the testing conditions were marked by an 8-1/2 x 11 sheet of red construction paper. The experimenter told the handlers falsely that it marked the target scent. If a suitcase in an airport was similarly marked, irrespective of the presence or absence of target scent, shouldn’t we expect a dog handler (or possibly the dog itself) to give that item heightened scrutiny possibly eliciting a false-positive dog alert?
It’s also possible, in this study, that a handler may accept the veracity of the experimenter’s falsely stated linkage of the red marker to the target scent, and indicate an alert regardless of dog behavior. The study was designed to accept the handler’s call without independently verifying the dog’s alert behavior.
Evaluating human and dog influences on alert locations relies on the assumptions that the 1) incorrectly defined red sheet of paper causes false-positive human input and 2) the hidden Slim Jim sausages cause false-positive dog input. It may be that 1) both the handler and dog are influenced by the marker, and that 2) the Slim Jim sausages did not interest the dog. If the experiment hid an open can of dog food or a piece of juicy meat, the result that “human more than dog influences affected alert locations” may have been exactly the opposite.
The four testing rooms may have had alert locations contaminated. On each of the testing days, the experimenter brought to the church in a closed metal box, separate from the briefcase that held the decoy scents and paper markers, 12 half-ounce samples of marijuana and 12 half-ounce samples of gunpowder sealed in plastic bags. The experimenter left the closed container of contraband at the door of each of the testing rooms before placing the decoy scents and paper markers.
Trace amounts of the contraband could have been carried by the experimenter and contaminated the rooms with valid target scents. If so, then some of the reported false-positives could actually have been correct positive alerts. In addition, regarding the experiment’s design, the metal box of contraband was “never opened inside the church” (page 3). What then was the reason for bringing real contraband? The experimenter could have reduced the possibility of contamination by bringing into the church instead an empty box, or a box of some innocuous objects.
Finally, the premise of the study that “Handler beliefs affect scent detection dog outcomes” is overbroad because the experimental design examines only false positives. The main objectives of service dogs and their handlers is to identify true positives and to prevent false negatives.
Slack • March 5, 2011 12:49 PM
@ Clive
I hope your recovery is both swift and complete, sir. I’m glad to see that being restricted to a smart phone hasn’t restricted the length of your posts, though I’m now of the opinion that you must have hands like a six year old girl to write all that in polynominal time.
Slack • March 5, 2011 12:51 PM
Whoops…I meant to say ‘Polynomial’. Guess I picked the wrong day to stop sniffing glue.
Clive Robinson • March 5, 2011 4:18 PM
@ Slack,
“you must have hands…”
Sorry they find UK size 12 Gents gloves tight, an I should think put the fear of god into even the largest of herbivours should I ever become a vet…
Believe it or not I actually type with the corners of my thumb nails so some of my odd spelling (but by no means all) where you get an S instead of a W etc is simply due to being unable sometimes to see the key I’m actually pressing….
But hey I’m told by a small but select audience it adds to my charm (but she’s just being nice 😉
Anonymоus • March 6, 2011 8:19 AM
@mw:
Thanks for that analysis. I have for the last 5 or 6 years adopted the policy of not trusting any research paper until I have read and “peer reviewed” it myself; it may be outside my own field, but principles of experimental design and statistical analysis are the same everywhere, and standards are declining nearly everywhere.
(And that’s not just me being a crusty old fart; editors of Nature have also noticed the decline in quality and reliability, and started trying to find out why.)
@Beta:
“real explosives are very rare (and real narcotics pretty rare I suppose)”
The explosive they were searching for was smokeless powder, which is not at all rare. The “narcotic” was cannabis, which, while used regularly by only a minority of the population, is hardly rare.
More to the point, the scenario simulated in the experiment was that the officer had obtained a warrant on the basis of detailed information from a highly credible witness (a university academic, no less!) That significantly changes the odds.
@boog:
“I don’t think the study should be discarded as “incomplete science”. It’s certainly reasonable to go in with the intention of finding flaws.”
That’s not what was incomplete about it. It was incomplete in that it was not fully controlled and so leaves open the possibility of confounding factors that lead to false conclusions. And when you look at the study design such confounding factors are not only possible, but probable.
“Do you think they are using this information to discredit or eliminate bomb-sniffing dogs?”
No, I didn’t think that. But the possibility that they are using this information to discredit drug-sniffing dogs … yes, that definitely crossed my mind. Once I saw the flaws in the design of the study, I googled to find what others have said about the FP rate for detection dogs, and discovered that there seems to be an active campaign by the drug culture to discredit drug-sniffing dogs. Of course that is hardly proof of intentional bias by the authors, but, well, yes, it did raise a little suspicion …
Oh, and what is the rate published by others? Well, one long running study by Chicago PD found that 44% of indications (they call them alarms) by drug-detecting dogs lead to an arrest for possession. However, it would be erroneous to describe the other 56% as the FP rate because some proportion of those cases would have involved drugs that were disposed of before they were found.
An interesting data point is the certification protocol by the NNDDA for certifying explosive detecting dogs. The NNDDA is independent of both law enforcement and dog training agencies. Tested dogs are tested in 5 distinct indoor locations, 10 to 20 pieces of luggage, 5 to 10 vehicles, and an outdoor area of 1,000 – 3,000 square feet. The number of targets in each site is varied randomly but can never be either 0% or 100%, and must be “consistent with” the number of hide locations. What “consistent with” means is not defined, but I would take it to mean something like “a large fraction of.”
There is no specific time limit but the assessor can terminate the test if it is clear the dog “no longer shows competence.” Dog and handler are assessed as a team, i.e. a wrong response from the dog, or wrong interpretation by the handler, both count as a fail. A true positive is also a fail if indicated at too great a distance, or by the wrong method.
Out of all those tests, the team is allowed no more than 1 “miss” (i.e. false negative) and no more than 1 false response (i.e. false positive.)
This protocol is by no means flawless, but it isn’t awful. A dog team would be unlikely to pass if their false positive rate, under these controlled circumstances, was more than 10%. With the most aggressive interpretation of the parameters, the team’s FP rate would need to be below 5%.
“I believe they are using the research to identify ways to improve the method.”
Why would you think that? It certainly isn’t what they claim in their own abstract. And they have never done any previous work on improving police procedures; before upping and starting on a study of drug- and bomb-sniffing dogs, Oberbauer and Lit were geneticists. (Dr. Lit, with Schweitzer, has done one previous study involving dogs, but it was unrelated to police work.) As for Schweitzer, she is psychologist, whose previous papers include a lot of work about addictive drugs and the people who use them. Notably, some very well regarded work on cocaine-dependent violent males. But nothing about police procedures.
“Surely you’d agree that reducing false positives would be a good thing, right?”
No. Maximising total utility is a good thing. You do this by simultaneously optimising the payoff matrix times matrix of FP, FN, TP and TN. Reducing FPs may increase utility, but in some circumstances, it may decrease it, through creating other costs or missing other opportunities. For example we can trivially reduce FPs to zero by laying in our beds and waiting for sweet oblivion to carry us away, but total utility tanks, too.
If we are talking about sending an innocent man to prison, the cost of a FP is extremely high, and it is worth doing a great deal to lower the FP rate. But in the scenario tested in this paper, the only cost of a FP was spending a few more minutes searching an empty room, in a building that was already being extensively searched under warrant. The cost of an FP in this circumstance is very slight, and only the most trivial fixes would be worth considering.
Petréa Mitchell • March 6, 2011 10:19 AM
w:
“If the dog wasn’t any use it wouldn’t stand up in court or be allowed”
We’d all like to think that, but the truth is that for years, many things have been allowed in court with no scientific validation. DNA analysis was the first technique to be introduced which had all the analysis done before it became standard. Since that came in, researchers have been trying to go back and apply the same rigorous standards to other forensic tools, and have been finding that some of them are not the slam-dunks everyone used to think they were.
(There was a great article about one of these, bullet-lead analysis, in Science News a few years ago. It’s not available online to non-subscribers, but if you’re in the US it’s probably available at your local library. See volume 165, #13, March 27, 2004; it’s the cover story.)
Clive Robinson • March 6, 2011 1:12 PM
@ Petréa Mitchell,
“We’d all like to think that, but the truth is that for years, many things have been allowed in court with no scientific validation.”
Just about everything to do with “forensic science” lacks any kind of validation scientific or otherwise.
Take for instance “eye witness” statments and “identification parades”, every time somebody gets it into their head to test it the results are dismal with most witnessess geting less of the details right than you would expect from chance, as little as two hours after an event and it goes down hill from there.
Which if you think about it is actually quite amazing. The reason that has been put forward is fixation on a single detail.
That is if the “perp” had an unusual detail like a broken nose or promenant scar, bushy eybrows or some such the witness concentrates on that and that alone and actually magnifies it out of proportion compared to any other feature. If the feature was actually within reason “normal” it becomes problematical.
That is thee result in a line up of six people if one person has that feature (even if it was not them) they get identified. Even when two people have the feature the one with it most prominantly gets chosen even if the other person is actualy the perp…
The result is eye witness testimony is in reality a compleast waste of court time, unless the perp is well known to the witness…
Likewise fingerprint evidence has been called into question successfully so many times you have to wonder what science there is in it.
The root cause of the problem is “m’learned gentalmen and ladies” their lives revolve around paper trails not the actual meaning of the words on the paper.
That is as long as the paperwork is on the correct form presented in the correct manner can be shown to have been unmodified etc etc it is taken as being valid. The actual content is rarely questioned for it’s validity under science as “M’learned friends” belive such cross questioning will confuse the jury…
The result every few years in the UK we get a scandle where a supposed “expert witness” preeminent in their field (acording to the court) has repeatedly and without any doubt presented as valid some pet theory that has absolutly no validity with others in the field of endevor and in some cases they have deliberatly mis used statistics to show there was no doubt of the defendants guilt (infant sudden death syndrome being a very very notable example and shaken baby syndrome going the same way).
Many of M’learned friends seem to have this quaint notion that science will provide answers without doubt, error or any room for a differing view point. I guess because of the traditional “burden of proof” being “beyond reasonable doubt” in criminal cases. Which might account for rumblings for the civil burden of proof of “reasonable probability” (whatever that might mean 😉
Clive Robinson • March 6, 2011 2:24 PM
@ Anonymous,
Further to your comment,
“Nature have also noticed the decline in quality and reliability, and started trying to find out why.”
There may be other reasons for the Likes of AMS IEEE, see Mat Blaze’s comments on their highly restrictive copyright policy,
http://www.crypto.com/blog/copywrongs/
If high quality authors decide they neither need nor want the likes of thes averacious organisations then it leaves the significant question of who will submit themselves to being riped off…
The answer being the inexperianced and the terminally mediocre who cannot stand in their own right.
Thus Pay walls and other copyright stealing policy are going to work against such organisations.
And personaly I would urge any who think of paying the devil’s price to do it sparingly, lest you convince the Devil he is just in his dealings and stealing of souls.
Beta • March 6, 2011 3:42 PM
@w: “I don’t think they would weaken the accuracy of the dog by teaching it false negatives on the job, when in situation it could make a real difference.”
If the handler gives the dog praise or attention, even subtly, after it “points”, that’s a reward. And because it’s immediate, it’s much stronger reinforcement than punishment that comes later, for no clear reason, when the officers find no contraband (even assuming the dogs are punished for that– are they?). And if the officers are not punished for false positives (which they don’t appear to be) then there is no “real difference”.
“If the dog wasn’t any use it wouldn’t stand up in court or be allowed.”
It is “useful” for allowing warrantless searches, the results of which have been admitted in court.
“…(some researcher would have a document stating that fact,accuracy).”
Some researcher, like, say, Dr. Lisa Lit at the University of California?
@Beta, if the dog handler ,nods there head, and the dog takes it as a signal to start running through the car/building looking for something and can’t find it, the dog could easily take that as (nod head, find/don’t find).
Saying two things that contradict each other to a person(without much knowledge of the whatever) creates a lot of confusion, to a dog…
“Some researcher, like, say, Dr. Lisa Lit at the University of California?”
“”If the dog wasn’t any use it wouldn’t stand up in court or be allowed.””
something is wrong or has just been realized
@Petréa Mitchell, yeah I’m not in the states. The discovery channel had something on dog training awhile ago, and they do have a good sense of smell, bad + good ?..
Doug Coulter • March 6, 2011 8:23 PM
A long time ago in a city far away, I worked for a defense contractor who got hired to do the signal processing for mice and rats trained to smell HE.
This was in the ’70s. The idea is they would train them in the usual way, with rewards and shocks, and we’ll look at their brainwaves and decide if the animal was responding — gets rid of all the issues with the dogs tripping on their handler being suspicious.
But it didn’t work as planned, and it took awhile to figure out. (This is kind of like Feynman’s story on rats) Turns out, they are so insanely sensitive that after a lab had been used to train one batch, it couldn’t be used again — there was still enough in the air to confuse the next batch! This was eventually sorted out, but at first of course they tried to blame our signal processing. Nope, we did it right, they failed to correctly train the animals, who (like in the Feynman story) just clued on something else. When that wasn’t in the briefcase we put them in, they never worked right.
But LEO’s will never willingly give up the dogs, because yeah, instant probable cause anytime they false trigger the animal. They do find things often enough (not all the lawmen are stupid or lacking in instinct, after all) so the courts continue to allow it. And hey, getting paid to have a very high class pet — where do I get a job like that? And hey, if someone shoots your dog as a cop, you can shoot them in return. Can’t do that if a neighbor shoots my dog!
All that stuff comes down to trust. I know enough signal processing to make a fake wiretap tape that god himself couldn’t detect the trickery in, and it’s now easy to do with easy to get software if you know how to use it correctly — but they allow it in court anyway — along with a ton of other things easily faked (or just done wrong to give a false positive).
I’ve met some criminals who were astonished at being in jail for the one crime they didn’t do! Maybe it kinda evens out? I think it comes down to what Bruce calls “detecting Hinky” in the various LEO’s and Judges involved.
BF Skinner` • March 7, 2011 6:58 AM
@Clive “most witnessess geting less of the details right than you would expect from chance, as little as two hours after an event and it goes down hill from there. ”
Some of this is how our brains process, store, retrieve and re-store memories. It seems that we don’t actually just report out our memories but replay them and once replaying them they change as we store them again.
How many times have we been on the wrong side of an argument and worked it in our heads again and again until we were right and they were wrong? Part of the same thing I think.
” as long as the paperwork is on the correct form presented in the correct manner can be shown to have been unmodified etc etc it is taken as being valid. ”
At the moment we have a section of our citzenry who believe our president is illegitimate because he was born in . . . anywhere but here. Of course it would be a more interesting argument if they hadn’t started out believing he was illegitimate and siezed on whatever proof they could. Their version of the scientific method I guess. R.A. Wilson called it what the Thinker Thinks, the prover proves.
But if the records are challanged, and some should be, then what’s the proof that this man now is the same infant born in Hawaii? He could have swapped out early in life, or stolen from another family, or switched at birth (all have been reported on in the last year’s newspaper) Without biometrics taken at birth, and stored for later questioning, there is no strong proof. My prints, and footprints weren’t taken when I was born. Am I the guy whose name is on the Birth Certificate? All the people who were there are long dead.
I have heard there are moves afoot to require DNA sampling from all new borns.
Clive Robinson • March 7, 2011 9:20 AM
@ BF Skinner,
In reverse order as I feel a little perverse today (I’m stuck in hospital and my body wants to be out in the rare London sunshine but my brain knows better unfortunatly).
“Am I the guy whose name is on the Birth Certificate”
The simple reply and in reality the only one of relevance is,
“Why should it matter?”
There is some very real stupidity that comes from the top that believes that serial numbers on people are important, yet they reserver the right to change theirs at will…
The simple fact is you chose to call yourself “BF Skinner” on this blog does it matter if this is your real name or not to me or the other posters on the blog?
No, and the next obvious question is why not?
And it boils down to good old fashioned risk and trust.
I have no reason to trust you, not because you are untrustworthy or anything like that it’s just there is no reason for me to trust you. The reason I don’t need to trust you is because there is no real risk to me.
Even if we met in person again with certain provisos (stop fingering that war axe lovingly 😉 I still have no reason to trust you, nor you I, because we are not likley to put ourselves in a risk situation where trust is required.
We have no reason to trust the government in fact the very opposit they put us in danger one way or another almost every moment of our lives and they steadfastly refuse to balance the risk/trust as we would expect in ordinary life.
Yet they also expect to be able to fully identify us at any place at any time so they very obviously don’t trust us and never will trust us so won’t take any risk from us.
The situation is patently absurd and contary to all other forms of social existance…
What the Government fails to realise, or just plain ignors is that this lack of trust in the people actually destroys society from the inside out.
As social animals we have to know what risk is and what trust is and how to balance the two otherwise we are a danger to ourselves and from others such as socio/psychopaths.
So why does the government want to destroy this very important lesson an the balance it gives to society?
Your guess is as good as mine unless they are trying to build a dependancy culture which even in the short term is a doomed culture…
So backwards to your previous point,
“Some of this is how our brains process, store retrieve and re-store memories. It seems that we don’t actually just report out our memories but replay them and once replaying them they change as we store them again.”
Actually this is usefull because the chances are we once would have focused on a significant threat which might just have lasted scant tenths of seconds in real time. But the nonlinear nature of a mind that can revisit and expand the time and in effect explore different outcomes would be a very valuable survival trait.
Are honeybees better than dogs at detecting explosives or narcotics?
Bruce opines that “I’ll bet bomb-sniffing bees don’t make the same mistakes.”
In an applied setting (i.e. non-lab), honeybees may not be exposed to potentially inaccurate handler input. However, if bees are going to replace dogs in the field, comparing their accuracy at detecting target scents isn’t sufficient. It is also necessary to determine/compare how easily bees and dogs can be trivially thrown off track by interfering agents.
If someone inserts a cloth soaked with pesticides into a suitcase will that block target scent detection by honeybees? If someone creates a decoy that’s drenched with sugar syrup or bee pheromones would that be sufficient to distract bees from their task of detecting an IED?
Beekeepers use pheromones to trap and control the behavior of bees. A mock swarming pheromone can be made by mixing common grocery, or health-food store, ingredients: lemongrass and germanium oils combined with lemon or lime juice. http://www.ehow.com/how_7555977_make-honey-bee-pheromones.html
In a lab, however, I believe that using honeybees to identify scents such as cancer will be very productive.
Richard Steven Hack • March 7, 2011 4:01 PM
“I’ve met some criminals who were astonished at being in jail for the one crime they didn’t do!”
So have I! Not including me. My case was open and shut.
I think most of the deliberately faked evidence is relatively rare because most of the time the arresting officer can simply lie under oath about what evidence has been collected – and it passes.
I spoke to one inmate who told me he was arrested (on the street in front of a drug dealing house contrary to warrant) based on a warrant which had a judge’s signature forged by the arresting officer (he had it analyzed by a handwriting expert – which I suppose in itself is suspect!) and based on a testing labs report on substances found which, when he wrote to the lab post-sentencing, proved to have been forged because the lab could find no record of that test.
I was in a Federal building holding cell when a suspect in a case came back from court laughing. It turns out the Magistrate was listening to the evidence to charge a man with some drug related offense. The arresting DEA officer was not impressing him. The Federal prosecutor told the Magistrate, “Your Honor, this man is a Federal agent! He wouldn’t lie under oath!” The Magistrate burst out laughing and told the prosecutor, “Don’t come into my courtroom and tell me a Federal officer wouldn’t lie under oath!” Obviously he’d seen way too many cases where that was subsequently demonstrated to be true.
I met another inmate who claimed that while he was an armed robber, he was not a BANK robber. But he was arrested and convicted of a bank robbery when the prosecutors at his trial had him remove his shirt and show the jury his numerous tattoos – thus proving he was a criminal and therefore had to be guilty of the specific crime.
“Justice is served”. Right.
Dirk Praet • March 8, 2011 8:06 PM
Although I’m not entirely convinced that the research methods used make for good science, I think the conclusion is actually kicking in open doors. It stands to reason that whether it be dogs, bees or elephants their olfactory capabilities can and will be influenced by other factors. These can be manyfold. The only question that actually matters is to which extent such a factor or combination thereof – deliberately introduced or not – can sufficiently subvert the method to the point that it can be shown unreliable to serve its purpose.
As to the bond that LEO’s have developed with their dogs, trained dogs can be great friends and the chances they will turn against or cheat on you statistically are way smaller than the odds of your wife doing the same. On the shopfloor, they are just as much used for intimidation purposes as for their other capacities. Even if it were positively proved tomorrow that dogs totally suck at detecting stuff, chances of any human member of current K9-teams applying for a partnership with an elephant, rodent or bee hive would probably be infinitesimally small.
Vince whirlwind • March 8, 2011 8:08 PM
I’ve walked past the dogs at Sydney airport more than a few times, including several times a few years back when I was returning unkempt and long-haired from some backpacking aroun Thailand – although the customs people were interested in my bags, the dogs never were.
Two years ago, however, cleanshaven and with angelic children in tow on our way back from an overseas trip, the Sydney airport beagle went straight for our luggage trolley, and specifically, one of my children’s pieces of hand-luggage and the contraband fruit she had in it. There is absolutely no way this was anything other than a well-trained dog doing its job and finding something nobody else could possibly have guessed was there.
Another reason for using service dogs instead of honey bees in the field as scent identification animals: Dogs thrive in snowy conditions, but honey bees become immobile when the temperature drops to about 10 degrees centigrade.
Subscribe to comments on this entry
Sidebar photo of Bruce Schneier by Joe MacInnis.
Andrew • March 4, 2011 7:33 AM
Even if the evidence that sniffer dogs (or whatever animal) are no better at detecting contraband than chance, I predict authorities will be very unwilling to give them up. In fact, I expect such evidence will make authorities even less willing to give up such methods for the simple reason that it means they get on-demand probable cause. “Honest, your honour, I searched his luggage because the dog reacted to it.”