Handler beliefs affect scent detection dog outcomes (2011)

(link.springer.com)

110 points | by ColanR 1684 days ago

15 comments

  • daddylonglegs 1684 days ago
    Dogs are very good at reading people and if the handler wants an indication they will get it; they may not even realise that they're influencing the dogs. As others have said, this has been known for a long, long time. There was even a book written taking a wry look a scent dogs (and the plans for scent bees etc that were in vogue during the great war on terror) [1].

    That said, dogs can be fantastic when used for finding threats rather than 'proving' what the handler already believes. They have a combination of nose and brain that we can't replicate any other way and are great for search and rescue, finding explosives and some law enforcement searching.

    [1] https://headspace-the-book.blogspot.com/2008/07/acpo-police-...

    • sneak 1683 days ago
      ...and yet dog alerts (encapsulating handler biases) alone are still used to establish probable cause in the US to strip people of their rights to privacy, and has resulted in jail for thousands of people who do not deserve it.
      • dymk 1683 days ago
        It's better than absolutely no preventative measures at all
        • ksdale 1683 days ago
          That's very debatable. I, personally, would much rather that people in possession of drugs go free rather than the police using the dogs as an excuse to search people they don't like the look of.
          • Smithalicious 1683 days ago
            Yes, but I suspect you would also support lighter sentencing or even decriminalization of drugs. I don't think you'd think the same if it was concerned potential terrorists or child sexual abusers.

            Basically, this kind of comment usually encodes some belief on what the law should be rather than a belief about how it should be enforced. And that's just not the issue at hand here.

            • ksdale 1683 days ago
              You're correct about how I feel about drugs, but I think the general argument applies across the board. There are times that enforcement costs are too high, given the crimes they prevent.

              One of the most common criticisms of airport security in the US is that the harm it is preventing is nowhere near commensurate with the burden it imposes on the populace. That may or may not be true, but it's not a belief about what the law should be, it's a belief that at some point, the price paid for additional safety is too high.

              Likewise with children, as a father of 4, I also understand the desire to go to any lengths to keep our children safe, but keeping them in a bubble for their entire lives, while certainly protecting them from many tangible harms, will ultimately stunt their growth. Once again, the price paid for safety can be too high.

              I happen to think that drug sniffing dogs are on the wrong side of the line. They probably do alert the police to drugs that would not have been found otherwise, but in my opinion, they allow the police to have far too much latitude when it comes to searches.

            • novok 1681 days ago
              TBH I wish we wouldn't think about terrorists and other 'Very Rare Criminal Types that are less likely than being struck by lightning' and instead focus on public order obvious as shit crimes as a society. The stuff that people actually care about crime wise, like their bike being stolen, or their car being broken into, muggings, etc.
            • sneak 1683 days ago
              Well, the US has assassinated its own citizens who have been accused of terrorism (but not tried, just executed), and I think this is a bad thing, so let’s go back to the issue at hand: using subjective biases to strip people of their human rights is bad, regardless of what those biases suggest to you or anyone else about their alleged crimes.
            • salawat 1683 days ago
              > I don't think you'd think the same if it was concerned potential terrorists or child sexual abusers.

              Maybe he wouldn't, but I would be. In the land where witch hunts are illegal, I'm one of the three civic minded Libertarians.

              Civil rights revocation should never be treated lightly. Period. Pragmatism must take the backseat to ideals in this regard, lest we end up undoing the very thing we're trying to protect in the first place.

              As much as that tends to put me at odds with my own pragmatic side, I can't argue with the direction things tend to take when we cave to the urge to sacrifice liberty for safety.

        • jobigoud 1683 days ago
          I feel this entirely depends on one's stance on the old "Would you rather let a guilty person go free or jail an innocent?".
        • liability 1683 days ago
          What kind if prevention are we talking about? Nitrate-sniffing dogs, or cannabis-sniffing dogs?
  • seanwilson 1684 days ago
    Is this not fairly well understood? e.g.

    https://en.wikipedia.org/wiki/Clever_Hans

    > Clever Hans (in German: der Kluge Hans) was an Orlov Trotter horse that was claimed to have performed arithmetic and other intellectual tasks.

    > After a formal investigation in 1907, psychologist Oskar Pfungst demonstrated that the horse was not actually performing these mental tasks, but was watching the reactions of his trainer. He discovered this artifact in the research methodology, wherein the horse was responding directly to involuntary cues in the body language of the human trainer, who had the faculties to solve each problem. The trainer was entirely unaware that he was providing such cues.

    • kenny_r 1684 days ago
      This particular horse is mentioned in the very first sentence of the introduction of the paper.

      I think it's a worthwhile study, especially considering such dogs are currently still used in law enforcement. It's important to be aware of the shortcomings and biases of this methodology, and to do so in a sound, scientific manner.

    • gowld 1683 days ago
      Is there any proof that "the trainer was entirely unaware that he was providing such cues." Obviously the trainer has an overwhelming incentive to lie.
    • edoo 1683 days ago
      It should be easy to train the dogs in an environment where the trainer does not know the intended result and must investigate each claim from the dog. All the dog training vids I've seen have the trainer themselves knowing exactly what lies in the course.
      • hnick 1683 days ago
        There are two problems with this, but both can be solved by having two people.

        The first is you'd probably have to get someone else to hide the scent for you. Not so easy when training solo.

        The second is that when you want to reinforce a behaviour you generally give a trigger or marker as soon as possible after the event. If you wait too long, the dog may associate the marker with another behaviour they've done in the meantime and overwrite the correct learning. So a 'yes' or clicker as soon as the dog alerts is the best way to go. But if you yourself don't know where it is, you risk reinforcing a failure. This is easy with a helper so you can glance at them and they indicate if you are correct or not.

      • silveroriole 1683 days ago
        I don’t think this would be easy. For training, the reward needs to come as quickly after/during the action as possible. Most dogs don’t have a great mental capacity for figuring out that they’re being rewarded for something that isn’t happening right now. If the dog has to wait even 10 seconds for the trainer to figure out whether to reward or not, a lot of dogs are going to get confused.
        • logfromblammo 1683 days ago
          During training/testing a second person, who is not visible, audible, or smellable to the dog, and who knows the conditions of the course, observes the pair via cameras and microphones.

          When the dog hits on something, the observer presses a button, indicating which target or decoy was indicated, or that an unexpected other object was indicated. The observer may also indicate when the dog seems distracted, or especially attentive to the handler.

          A computer takes the buttons as inputs, and if its intermittent reinforcement program indicates it, the handler receives an instruction to reward the dog. The program may also reward the dog for misses, or just for paying attention to handler instructions, so the handler is not able to associate rewards with accurate hits. The handler then still has to investigate the hit to determine what the dog found.

  • praptak 1684 days ago
    I've heard this mentioned a few years ago in a TV report on dogs used in labs that test evidence for criminal trials. Their procedure was to let the dog alone into a special scent-free room that contained the sample being tested, plus a few control ones. The samples were placed into containers that only let the smell out (no visual cues, even though the samples were identical-looking cotton swabs).

    Also, I vaguely remember a high-profile case where the evidence was dismissed precisely because the procedures for gathering and testing the samples were not followed.

  • praptak 1683 days ago
    In this context it is ridiculous that a dog test is considered a probable cause for a police search: news.ycombinator.com/item?id=10942196
    • Smithalicious 1683 days ago
      Is it really? Probable cause doesn't need to be a high bar, in my opinion. Contrary to what people here seem to think, I don't think a police search is that invasive and I'd rather we perhaps search some people on shaky grounds than we let people with drugs/bombs/whatever walk free.
      • brokenmachine 1683 days ago
        I strongly disagree.

        Being physically searched by men with guns is not something innocent people should be subjected to with no, or just the thinnest of illusion of justification.

      • praptak 1683 days ago
        Then just do away with 4th amendment, rather than pretending that cops respect it.
      • htek 1683 days ago
        So you're cool with a cop that is biased against black people using their dog as a pretense to search them? Oh, and, yes, a police search is invasive and demeaning. I suggest you try it before you discount it.
        • Smithalicious 1683 days ago
          No I do not support people doing bad things. That does not mean that I am against everything that could possibly be used for evil. This is an absolutely inane accusation.

          Are you fine with black communities being ruined by crack cocaine because police can't search people? Of course not, that'd be a ridiculous thing to think.

          • htek 1682 days ago
            What you're saying is that it's okay to treat all black people as if they're criminals because a small subset of them engage in criminal activity. It is well established that black and white people use illegal drugs at about the same rate. Where's the call for stop-and-frisk of all white people? Or doesn't that fit your think-of-the-black-community veil over your racism?
  • ptah 1684 days ago
    does this imply that the dogs perform a similar role to dowsing rods in this situation?
    • 693471 1684 days ago
      This is the analogy I've been looking for.
  • the-dude 1684 days ago
    Up to 2006 'scent identification' ( linking persons to objects ) was used as evidence in Dutch courts.

    It turned out the test weren't done 'blind' and the practice has been abolished.

  • tiku 1684 days ago
    Well it is not far fetched that they can smell the excitement or anxiety of their owners as they approach the evidence they believe is the correct one?
    • anon9001 1684 days ago
      > For scent detection dog handlers, beliefs that scent is present might result in either sufficient inadvertent postural and facial cues so that dogs will respond regardless of the absence of scent

      I don't know how good dogs are at pose detection, but it's definitely plausible that they can tell when an alert is expected.

      • bayesian_horse 1684 days ago
        A less subtle cue would be how they treat the "suspect".

        It's less about the handler's pheromones and more about movement, speech and posture.

    • taneq 1684 days ago
      They can probably smell their handler getting interested well before the handler's consciously even aware of it.
      • bayesian_horse 1684 days ago
        I doubt that, at least in the literal sense of smelling.

        Dogs can however distinguish behavior patterns in their handlers, because that's mainly how they get rewards and are punished.

        Both the patterns and the rewards are much more subtle than handlers are generally aware of. A service dog is in some sense the worst case of an overfitting machine learning model you can imagine.

  • tanepiper 1683 days ago
    I've been doing detection classes with our Polish Hunting Spaniel and she's quite the natural, but the hardest part of it is training the human (i.e. me) to NOT give off any queues about where I think things may be.

    Of course, these are fun classes and in no way controlled environments - but the best thing you can do is to trust the dog's judgement.

  • your-nanny 1683 days ago
    I think the more likely explanation is that the handler's sensitivity to dog alerts changes with handler's beliefs, a possibility they raise in the paper's discussion.
  • mgarfias 1681 days ago
    would have been better to use SAR dogs.
  • katakuchi 1684 days ago
    I only read an excerpt of the article, but it seems to me that it might be tough to attribute the false detection to a certain variable within the animal's cognition. Is it perception of human behavior, mental state, tone of voice, owner's scent or maybe even some Pavlov effect on picking out cases that are certainly positive?

    However, this obviously has some interesting implications in court cases or situations where a traveller might be falsely accused of drug traffic, or when suspect X is 'randomly' selected for checks.

  • truths33k3r 1684 days ago
    how is this nonsense even getting published?

    tiny sample size, n=18 dogs 5 different breeds of dogs large variance of dog age

    • BlackFly 1683 days ago
      This shouldn't surprise you, but if the effect is large and real, you don't need large sample sizes to prove something. As an example, a loaded die that always comes up with 6. After 18 straight 6's in a row, you are more than assured that you have a loaded die. I imagine that a rational person is convinced long before the 18th 6.

      The point is: large samples are only needed to measure small effects. Most studies nowadays need large samples because the large effects are low hanging fruit and were noticed quickly (in the past). Meanwhile, finding a single black swan proves that not all swans are white.

      This study proves that in even a small selection of scent dogs, you are going to have dogs which are highly affected by handler signals. You might be able to use the methods of the paper to exclude dogs which pick up on handler signals and therefore improve the training regimen of scent dogs, there is no evidence to support this possibility however.

    • your-nanny 1683 days ago
      if the dogs were restricted to male Labradors age 5, then the range of inference would be male Labs age 5.
    • ShlomoS 1684 days ago
      The sample size is adequate for the conclusions drawn from the results. The breeds and ages are representative of dogs used for this work. Nothing in your post supports the idea that the study is "nonsense".
  • corodra 1683 days ago
    18 dog teams? What a thorough experiment. Sounds like the infamous chocolate weight loss study that was pushed to the media that made them look stupid.

    To a small degree I can agree to this. But I have k9 officer friends and I know their dogs. Good luck with this. It means a bad and impatient handler rather than an ineffective dog. But it's alright, police baaaaad. Evidence is evil.

  • jaclaz 1684 days ago
    The observation seems logical, but I wonder what is actually the relevance of the study in practice.

    I mean, there are three common situations where these dogs are used:

    1) airports, stations, traffic stops etc. where the influence of the trainer might be relevant (i.e. the trainer, even if involuntarily, may influence the dog if the person that is sniffed - or their belongings/looks/behaviours/etc. - look suspicuous to the trainer)

    2) search in buildings for drugs or explosives

    3) screening of mail, packages, containers, etc.

    I believe that among these only in activity #1 the trainer might have some pre-constituted suspect that would "trigger" the influence on the dog.

    But in any case the dogs are only used as a means of "triage" of sorts, so at the most this influence leading to false positives will cause some time wasted.

    Still in case #1 (and possibly in case #3) the issue might be that if there are too many of these false positives, while the attention of the trainers/policemen is drawn to the "false target" some other one (actually carrying drugs or explosives) will "go through".

    • hnick 1684 days ago
      I haven't read the article (it's a little late in the day for academic prose) but have done a lot of scent work with my dog as a hobby, and talked with a trainer who does it for a living (bombs, drugs, etc).

      Handler bearing absolutely plays a role if you aren't careful. One way our trainer taught us to fix this was to try to pull the dog away from the scent during training. If he follows us, we leave the room and stop the activity. The idea is to tell him that humans are stupid, they don't know where the scents are, so stop trusting them, because then your game and reward goes away. It worked very quickly and now he locks his legs and won't move when he's got the odour. This avoids accidentally pulling him off the odour when you're covering a large area at a brisk pace.

      Regardless, whenever possible one of us hides the odour and the other runs the dog. This helps us avoid any preconceptions (such as focusing on one corner of the room, missing potentional scent cones from cross-drafts). Humans can still mislead the dog like this so it's best to form good habits and have an open mind.

    • bayesian_horse 1684 days ago
      I have heard the term "search warrant on four legs".

      In many jurisdictions, including the US, police officers are not allowed to search anybody or their cars, even if they are black.

      So if the dog learns to please his handler by indicating contraband, they have a chance to at best inconvenience anyone they (dis)like.

      • pulisse 1683 days ago
        > In many jurisdictions, including the US, police officers are not allowed to search anybody or their cars, even if they are black.

        In many US jurisdictions it is settled case law that a search is justified if an officer simply avers that they detect the smell of marijuana, with no other cause or corroboration required.

      • jaclaz 1684 days ago
        >So if the dog learns to please his handler by indicating contraband, they have a chance to at best inconvenience anyone they (dis)like.

        Sure, but (of course it depends on countries and local Laws) it is only one of the possible ways (excuses) that can be used to target someone.

        Of course if in the car there is no drugs (or explosives) the only net result is a waste of time of both the policemen and the "target".

        It would seem to me much more relevant to know how many cars searched after a dog sniffed something ended up as being actually contraband free (false positives) and how many actually turned out as actually containing some drugs or whatever.

        If the dogs (in experiments without the influence of the trainer) have an accuracy of - say - 90% and with the influence of the trainer (or whatever other reason causes on field differences) this drops to - still say - 55% then there will be an issue as we would be nearing the flipism convenience, but if the overall accuracy remains in the 80 or 85% range then the effect - even if noticeable - would be not much relevant in practice.

        Or, if you prefer, I completely agree that a trainer can induce a dog to sniff something even when there isn't any, but the relevance of the fact depends only on how often this happens.

        • cbanek 1684 days ago
          For less reputable police, they might just say the dog alerted when it didn't. What's a layman to say? No it didn't? And courts usually give the benefit of the doubt to police, even if they technically shouldn't. Because police don't need probable cause to use a dog to search your car means it is the de facto way to escalate to a probable cause search, and even if a dog isn't present, bringing one there usually makes people uneasy, leading to claims of "suspicious behavior", and allows the police to detain someone until it arrives.

          Sadly, I don't think the false positive alert rate is something the police are looking at or trying to reduce. In my cynical opinion, they are just looking for more reasons to find you a criminal, even if you aren't one. If they view you as a criminal to start, that's a huge burden to overcome.

          • bayesian_horse 1684 days ago
            I believe that malice is optional for the conduct of police officers to be problematical.

            Perfectly honest or understandable motivations and beliefs are sufficient.

          • jaclaz 1684 days ago
            >For less reputable police, they might just say the dog alerted when it didn't. What's a layman to say? No it didn't?

            Exactly.

            • whenchamenia 1683 days ago
              Odd, I raised that objection and a warrant to search my car was still issued. There are many, many shady cops. My guilt had alredy been assumed, and they were just hunting for justification.
            • trotawa 1684 days ago
              I've seen body cam footage of the "less reputable" category of cops.

              The likelihood that drugs get planted in your car during a search (whether legitimate search or not) screws up all of these collected stats and just like most things in this country, there is a racist component to this that helps more easily target the poorer and browner.

              Terrorizing black people is a national pastime and somewhat tradition... some cops seem to want to continue that tradition and many white legislators are glad to give them the opportunity. For them, ignoring this research is a no-brainer.

        • lultimouomo 1684 days ago
          > Of course if in the car there is no drugs (or explosives) the only net result is a waste of time of both the policemen and the "target".

          Google for "civil forfeiture dog car stop". They don't need to find drugs to take the car or any cash in it, as the dog signalling is deemed enough of a cause for forfeiting, though not for prosecuting criminally.

          • jaclaz 1684 days ago
            Sure, I know about that. It is part of the (IMHO completely off) ways the Law is applied in the US, I don't think that the same exists (or is used to the same extent) in other countries.

            But the base issue is still about reliability of the dog, the integrity of the trainer handler and how it is considered "acceptable" see:

            https://www.washingtonpost.com/news/the-watch/wp/2015/08/04/...

            • bayesian_horse 1683 days ago
              Intent on the part of the handler is optional, I think.

              Common Biases on top of good intentions would be enough to produce the "desired" false positives.

    • anonymfus 1684 days ago
      >so at the most this influence leading to false positives will cause some time wasted

      That "time wasted" feels more like a psychological torture or at least a day ruined, and you ignore that cops stop mostly the same people every time, so sometimes it's a day ruined every day.

    • praptak 1683 days ago
      4) Fabricating a probable cause for a search when there is none: news.ycombinator.com/item?id=10942196