Posts tagged ‘learning’
Maybe. Maybe not.
An article in the February 15 edition of Scientific American by Ferris Jabr reported that pharmacologist John Hepler and associates recently discovered a gene in the brains of mice that codes for a signaling protein that “significantly boosted brainpower with seemingly no negative consequences.” Jabr writes:
People have this gene, too, and it is active in the same brain area. In other words, we may have a gene in our heads that is actively making us dumber.
Hepler et al were studying the CA2 section of the hippocampus when they discovered that neurons in that area appeared to be blocked from participating in learning and memory processes when they were saturated with signaling protein RGS14.
When they bred mice who lacked the gene for RGS14 the CA2 neurons were no longer blocked and the mice exhibited some interesting differences in learning and behavior. Scientific American reports that:
The genetic tweak affected more than physiology—it changed how the mice performed on memory tests, too. The experimenters presented two identical objects to knockout mice, which lacked the RGS14 gene, and to normal mice. Four hours and again 24 hours later, the researchers switched one of the objects with a new object. The knockout mice spent far more time exploring the new object than the normal mice did, indicating that the altered rodents had a better memory for distinguishing familiar and strange objects. Knockout mice also learned to navigate a water maze and locate a submerged platform faster than normal mice did. The scientists observed no detriments from removing the RGS14 gene.
“Why would we have a gene that makes us dumber?” asks Serena Dudek, a neuroscientist at the National Institute of Environmental Health Sciences and a co-author of the study, which was published in the September issue of the Proceedings of the National Academy of Sciences USA. “We don’t know. But if the gene is conserved by natural selection, there must be some reason. Intuitively, it seems there should be a downside to having this gene knocked out, but we haven’t found it so far. It may be that these mice are hallucinating, and you just can’t tell.”
And this is where I get frustrated that I don’t have unlimited access to journal articles (or an unlimited budget to buy it with) because based on what they’re saying here it sounds to me that while the gene may have something to do with spatial learning, it might also have evolved as a safety mechanism controlling neophobia – not stupidity.
I suppose that one might be able to test this by looking at the action of CA2 in young animals during different stages of development when neophobia typically ebbs and wanes, but since mice reach maturity within weeks of being born, I’m not sure they’d be viable candidates for this.
Either way, I’m sad to say that I don’t think they’ve discovered a genetic cure for stupidity.
There is a not necessarily equal and typically 10% faster reaction. At least neurophysiologically speaking.
The February 3, 2010 issue of the Proceedings of the Royal Society B includes an interesting article. In The quick and the dead: when reaction beats intention; Andrew Welchman, James Stanley, Malte R. Schomers, R. Chris Miall and Heinrich H. Bülthoff (Welchman et al.) write about the different neurophysiological pathways followed by intentional acts and reactions.
The mythology of the American West is shaped by liquor and Hollywood (Brown 1995). Inspired at least by the latter, the Nobel laureate Niels Bohr considered why, during a gunfight, the man who drew first was the one to get shot. He suggested that the intentional act of drawing and shooting is slower to execute than the reactive action in response (Cline 1987), an idea grounded in the everyday trade-off between stimulus-driven behaviour and intentional, planned actions.
Welchman et al. didn’t actually study gunslingers, they studied identical movement sequences in conditions where participants either chose to initiate the movements themselves or reacted to an opponent. Their work demonstrated that reactive movements were, on average, 10% faster than intentional movements.
They propose that this occurs because intentional and reactive sequences follow different processing routes in the cortex. After all, a system capable of producing quicker movements in response to threats in the environment makes perfect evolutionary sense.
This is nifty stuff – but what’s it got to do with dog training?
Well… no matter what the nice sales brochure tells you, a four-week course will not magically turn you into a real dog trainer (especially when that course also covers sales, business operation and franchise rules).
Dog training isn’t a rote task (like being a census taker) that you can pick up in a few days – it’s an art. And like classical guitar or dressage, it takes years to master it.
As Welchman et al. conveniently point out, an inexperienced handler has to rely on those relatively slow intentional movements when he works with a dog. While this may only be 10% slower than the reactive movements an experienced trainer relies on – that 10 % can mean the difference between a bite and a crisis averted.
With experience, arts like dog training or sumo wrestling get mapped into the fast, implicit pathways in the brain and they go from being skills we work very hard to acquire to becoming an integral part of who we are. The work is transformed into an autotelic experience.
There are three stages in this process – learning, practicing and mastering.
In the learning stage we need to explicitly parse out actions in our mind as we perform them and our work is almost entirely intentional. In this phase our skills tend to be slow, unnatural and awkward. We know where we want to be, we just don’t know how to get there.
In the practicing phase our actions come naturally, but we aren’t explicitly aware of what we’re doing. This is the “I know when I have it right but I can’t explain why” phase. Running on fast, implicit pathways our skills are quick and natural – but the parts of our brains that allow us to explain what we’re doing can’t keep up with them yet. As William James put it; “We are aware then of nothing between the conception and the execution … We think the act, and it is done; and that is all that introspection tells us of the matter.”
As we master a skill, we hone our actions and our mental processes. Eventually, we reach a stage where the work and our awareness of it meld. Our intentional and reactive processes work together to put is in a state of flow. And the real beauty of this stage is that the chase never ends. As Mihaly Csikszentmihalyi writes; “You can approach it. You can home in on it. You can get really, really, really close to it. But like Cezanne, you can never touch it. Mastery is impossible to realize fully.”
I feel sorry for dog trainers who graduate from corporate training programs thinking they’ve been given all the tools they need to pursue a successful career. Being a dog trainer is the kind of career that should take over your mind, your heart – your very being. Like being a writer, it’s not a job where you’re likely to make a lot of money or get a lot of recognition. You need to do it because you love it. And you need to be willing to spend a decade or more working just to begin to master it. You should to it because being and working with dogs is something that’s so deeply and strongly and intensely etched into your soul that you cannot imagine doing anything else.
And worst of all, like Cezanne, you need to realize that being as good at it as you really want and need to be is an illusive target that you’ll never truly achieve.
Today I came across a fascinating article in Seed Magazine on dopamine, artificial intelligence and social learning. There’s so much good stuff in there that it may inspire a few posts. Here’s the first one.
Dopamine is great stuff. It doesn’t just let us take pleasure in our world — it’s also key in helping us understand it. Cambridge University’s Wolfram Schultz studied dopamine’s role in triggering Parkinson’s disease. Schultz recorded dopamine levels in monkey’s brains to study how dopamine neurons died in the part of the brain that controls movement. As he collected data for the study he noticed that dopamine neurons fired just before the monkeys were rewarded for moving.
Stunned, Schultz realized that he had just discovered the brain’s reward mechanism — and he’d done it by accident. According to Seed:
His experiments observed a simple protocol: He played a loud tone, waited for a few seconds, and then squirted a few drops of apple juice into the mouth of a monkey. While the experiment was unfolding, Schultz was probing the dopamine-rich areas of the monkey brain with a needle that monitored the electrical activity inside individual cells. At first the dopamine neurons didn’t fire until the juice was delivered; they were responding to the actual reward. However, once the animal learned that the tone preceded the arrival of juice — this requires only a few trials — the same neurons began firing at the sound of the tone instead of the sweet reward. And then eventually, if the tone kept on predicting the juice, the cells went silent. They stopped firing altogether.
The data were fascinating and utterly perplexing. Schultz knew that the dopamine neurons were a key part of the learning process, he just didn’t know how they were doing it.
Meanwhile, over at the Salk Insitute, computer scientists Read Montague and Peter Dayan were working on the temporal difference reinforcement learning (TDRL) model of artificial intelligence. Their goal was to create a “neuronlike” program that could learn simple rules to perform goal-oriented behaviors. The Seed article notes:
The basic premise is straightforward: The software makes predictions about what will happen — about how a checkers game will unfold for example — and then compares these predictions with what actually happens. If the prediction is right, that series of predictions gets reinforced. However, if the prediction is wrong, the software reevaluates its representation of the game.
Montague and Dayan were holding the key to Schultz’s dopamine release as learning mechanism conundrum. And Schultz had collected neurological data that supported Sutton and Dayan’s TRDL model. Dayan discovered the link in 1991. Seed quotes him as saying:
“The only reason we could see it so clearly,” Montague says, “is because we came at it from this theoretical angle. If you were an experimentalist seeing this data, it would have been extremely confusing. What the hell are these cells doing? Why aren’t they just responding to the juice?”
The truly fascinating thing about the model is that it’s based on expectations. Predictability, correlation and contrast are the basis of learning — not reward and punishment. I may just be a dumb dog trainer, but this isn’t news to me.
A predictable world makes sense. Predictability builds confidence. When you work with a shy or insecure animal, maintaining a high degree of predictability can build the confidence. Correlation is the foundation of predictability. Contrast tells us what things don’t correlate, it teaches us what things don’t work (and what works better than we expected). It also works neatly in conjunction with correlation to teach animals how to generalize. Add a little contrast to a lot of correlation and you’ve created the perfect proofing exercise.
These three processes work together to help an animal make sense out of the geopbytesof information that inundate it in day to day life. And much of the beauty of the system lies in the fact that it’s a dynamic one. The first time that we notice that A correlates to B we are surprised. The neurochemistry of surprise makes novel events memorable because bursts of dopamine are emitted in the wake of unexpected rewards. After a few repetitions of the pattern, we store that bit (or byte, as the case may be) of correlative information away. If, at a later time, the correlation fails, we’re surprised by the failure in our expectations and our neurons readjust to the new situation as dopamine production declines when we expect a reward that we don’t receive. Sometimes those rewards are explicit ones like food treats or payroll bonuses — but they can also be implicit rewards like the excitement you feel when your team wins the big game or the thrill your dog feels when he finds his bumper in that tuft of grass where he expected it to be.
But rewards alone won’t create learning. We need the contrast that comes from errors too. Again from Seed (bold mine):
“The accuracy comes from the mismatch,” Montague says. “You learn how the world works by focusing on the prediction errors, on the events that you didn’t expect.” Our knowledge, in other words, emerges from our cellular mistakes. The brain learns how to be right by focusing on what it got wrong.
And there you have it. It’s impossible to learn unless you make mistakes. The freedom to make mistakes is vital to intellectual growth. This is why raising a dog in an over-managed way does irreparable harm to his intellect. A dog can only learn to think for himself if he’s given the opportunity to do so. Making mistakes and coping with their consequences is a key part of learning. It helps the dog evaluate the contrast he needs to generalize his learning to new situations — and to reject previous learning when it doesn’t fit a novel situation.
I think that it’s also interesting to note that the model appears to imply that the implicit dopamine rewards of making correct, correlative predictions are stronger and more long-lasting than explicit rewards like food. This may help explain why the associations created by negative reinforcment (a concept most trainers don’t really understand) are very strong and long-lasting.
Young Audie is a very conscientious dog. A thinker. Things matter to him.
As we continue with his education this attitude of mindfulness becomes more apparent. I can often tell when he reaches a new stage in understanding a concept because he will begin to practice (or rehearse) it on his own.
In practicing he repeats an action I’ve taught him on his own. He’ll often practice an action a few times in a row, then go lie down to process what he’s taught himself. The calm, mindful demeanor he expresses as he practices is utterly different from the bounding exuberance he is prone to much of the rest of the time.
Here are a few recent examples of his practicing:
- I’ve been working to teach Audie to pick up his and Zip’s metal food bowls and bring them to me at meal times. They’re somewhat heavy and oddly shaped which makes them more difficult to pick up than most (though not all) of the items I’ve had him fetch to me. If he picks a bowl up in what seems like the simplest way – by gripping the rim closest to him in his mouth – the bowl is not only difficult to balance in his mouth, it also obstructs his view as he tries to walk with it. So, I’ve been coaching him to work against his first instinct and pick the bowl up by gripping the rim of the bowl at the point where it is farthest from him. About a week after we started to do this I noticed that Zip’s bowl (the smaller of the two) seemed to be randomly moving around the house. I didn’t initially see him pick her bowl up and carry it, but it would *miraculously* appear in a new place three or four times a day. About four days after Zips’ bowl begin its journey, Audie’s bowl started to wander as well. The morning after he had apparently started to carry his bowl around on his own, he made the leap. When I looked down at him and asked “Would you like your breakfast?” he grinned, ran across the room and very carefully picked up Zip’s bowl – correctly, by the far side of the rim – and brought it to me. I took it from him and told him “Find the other one,” and he turned around and picked up his own bowl – correctly again.
This is still not a simple chore for him and he still practices it a few times a day. I’ve caught him at it recently and it’s fascinating to watch how he experiments with the task. He’ll often still sometimes take the bowl by the near side – then stops with a “hmmm, this isn’t right” look on his face – sets it down, tries again and then visibly goes “Aha!”
- We’ve also been working more on off leash heeling skills. Since my dogs spend most of their lives with me and off leash – heeling is one of the last “standard” obedience skills I teach them. Because my own dogs have a strong foundation in other skills (recall, send out, stationary commands, directionals, yielding) before we begin to work on the heel, we do most of our heeling work off leash right from the start. I start out working my dogs on short bits of off leash heeling with lots of turns and stops. The initial goals are to teach them to stay on my left side and to pay close attention to where I am going. My kitchen has had a central island surrounded on 3 1/2 sides by counters. It was a perfect place to practice beginning heeling skills as the narrow aisle between the counter and island restricts his ability to move out of correct position to, mostly, forward and backward errors. We started out working on short bits of heeling there, then as his skills improved moved inside the training building. There I set up traffic cones to create a smaller working area inside the 50×50 training room and we worked on random weaving patterns. From there, we moved on to working in amongst the seveal large hardwood trees in my front yard. As we moved from one training environment to another, I kept the patterns much the same. And I used the same body language to coach him.After I while he started to sometimes come and sit at my left on his own. He’ll usually do this either when I am standing still in a place where we’ve working on heeling – or when he wants my attention. And he adopts a rather formal posture as he does it. Without giving him a verbal command, I’ll crook my left arm, lead with my left leg and move forward. He looks up at me with a huge grin and heels along with me. He does this voluntary heeling in a very cheerful and animated way. Like a it’s dance, or a game.
It’s fascinating to watch this young dog learn and grow. I’ve learned to watch for these periods of practice and use them to establish the rhythms of our work together. If he seems to be having difficulty with a task, we keep working on it – but don’t add a new layer of complexity to the task until after he masters it in his own practice.