Stay connected.

Earlier posts by category

Entries in behaviorism (3)


When positive is a bad thing


We sometimes create unnecessary confusion when we mix terms of art with everyday language. Take the word "positive." Those of us who shelter under the wide umbrella of "positive training" mean "positive" in the way that most people do: we like to exchange good stuff for good behavior. Unfortunately, we get tripped up when we then try to explain the principles of operant conditioning that buttress our methods: "Well, positive punishment isn't actually part of positive training..." Aargh.

We might have an easier time if we substituted "addition" and "subtraction" for "positive" and "negative" in our description of the consequences that condition behavior. As awkward and ungrammatical as "addition punishment" might be, it would at least have the advantage of common sense. But for now we have "positive punishment," and I want to examine some of the fundamental reasons that it's bad not only in the moment for the trainee but in the long term for the trainer.

Sad to say, there are people who get a charge from punishing other creatures; we can define a sadist as someone who finds punishment (colloquially here, the infliction of suffering) reinforcing. These are often people whose insecurity runs so deep that they require cringing submission from dogs or children or spouses to reassure them of their power. The New York Times Magazine recently ran an article by Charles Siebert describing a shift in the attitude of police and others in law enforcement toward animal abuse: as they have come to recognize its strong correlation with other, human-directed forms of violence, they have begun to take it more seriously.

The fact that many serial killers begin by torturing animals is well enough established to have become something of a cliché in film and fiction, but I hadn't known how often animals are used as the levers of pressure in abusive family dynamics. According to Siebert, abusers will often threaten violence against a pet in order to bend other family members to their will. This kind of emotional blackmail has the horrible side effect of eroding empathy in the victim: a child who is helpless to protect a beloved dog or cat can only defend himself against the pain of identification by numbing himself to the animal's suffering, even to the extent of participating in the abuse.

Many (I want to think most) of us with pets use punishment more "judiciously," and yet it's difficult to make an indelible distinction between abusive and "constructive" punishment. On the question of what motivates abuse, Siebert quotes Randall Lockwood, the ASPCA's senior vice-president for forensic sciences and anticruelty projects: "I've spent a lot of time looking at what links things like animal cruelty and child abuse and domestic violence. And one of the things is the need for power and control. Animal abuse is basically a power-and-control crime."

All social engagement requires negotiation. When we share our lives with other creatures, we often find that our desires clash. When compromise seems impossible, we may resort to force to impose our will. If we are not sadists, if our sense of compassion is strong enough that we feel the pain we inflict ("this hurts me more than it hurts you"), we punish because we're convinced that nothing else will work-- we don't know how else to interrupt or eliminate behavior we find unacceptable. In the heat of frustration or anger, we're often unable even to imagine other possible responses, let alone consider their relative efficacy.

Regardless of the soundness of our reasons and the resilience of our capacity for empathy, we punish because we can. We are only able to use pain as an "instructive tool" if we're at least momentarily in a position of superior power (real or credible): we either don't expect retaliation or are prepared to escalate our force if the other party fights back.
Committed positive trainers reject that contract. They recognize the fundamental imbalance of power that exists when one creature is dependent on another for its sustenance and many of its pleasures, but refuse in principle to exaggerate that advantage through the use of force. Indeed, many of our methods were developed in situations where coercion wasn't practicable, with wild animals in open spaces.

We reject punishment in principle, but as impulsive animals with many bad habits (maybe I should speak only for myself here!) we may sometimes find it difficult in practice to eliminate it from our training. Fidelity to positive reinforcement requires self-discipline, and so in the process of training other animals, we discover that our first, most important (and most challenging) task is to train ourselves.

Photo by Bob Pearson.



Behaviorism and Desire

We want what we want and we want it now. (Humans are better than most animals at deferring gratification, but not always and not by much.) Any deliberate manipulation of another creature's behavior requires that we become attuned to that creature's desires, and these may be almost as idiosyncratic among dogs or dolphins as among people. To paraphrase Sam the Eagle (of Muppet Caper fame), we are all weirdos. This is where behaviorism goes productively amok.

In operant conditioning, one doesn't create behavior per se, one merely increases or decreases the likelihood that a given behavior will be performed, and one does this by controlling the behavior's consequence.

Consequences fall into four categories, defined by two binary oppositions (positive/negative, reinforcement/punishment): positive reinforcement, negative reinforcement, positive punishment, and negative punishment. "Positive" in this context refers to the addition of some thing or force, "negative" to the removal of some thing or force. "Reinforcement" names anything that increases the likelihood that a behavior will be repeated; "punishment" names anything that decreases that likelihood. More simply, reinforcement tends to the yummy and pleasing, punishment to the nasty and fearsome ("aversive" in behaviorist lingo).

So "positive reinforcement" is the introduction of something good (chocolate cake, a belly rub, a game of tug, a shoulder massage), "negative reinforcement" the removal of something bad (pressure on the bit, a parent's screaming, a scary dog or mailman): whatever I did to create either consequence, I'm more likely to repeat it. "Positive punishment," which sounds like a contradiction in terms, is the introduction of something nasty (leash jerk, skunk spray, burned fingers), while "negative punishment" is the removal of something we like (attention, bones, freedom): whatever I did to earn these consequences, I'd like to avoid repeating it.

There's a wealth of complications buried in this simple schema, but the most significant concerns the vagaries of desire. We all (human and non-human animals) like different things, and we like them with varying degrees of intensity. Our desires are fluid and changeable, shifting with experience, mood, and context. Once upon a time, I loved bananas and (very briefly) the voice of Suzanne Vega, but both now make me queasy. Conditioning wouldn't be possible if our preferences were forever fixed, but our fickleness makes us slippery subjects. And that seems very much to the good. I have learned to love Skinner only because his account of behavior remains forever incomplete; the "laws" of behaviorism, while they are powerfully, empirically predictive in the aggregate, get wonderfully complicated when they tangle with the rebelliously singular individual.

photo credit: texascooking via photopin cc


Learning to love Skinner

What I found most remarkable in Karen Pryor's book Reaching the Animal Mind is how seamlessly and matter-of-factly she enlists the insights of behaviorism in a project of creative collaboration between species. Once they are taken out of the laboratory and into the world at large, the conditioning techniques that B. F. Skinner and others developed for scientific purposes become powerful tools for the achievement of warmer, fuzzier ends. They can help elucidate animals' integrity as individuals, and (not coincidentally) foster positive emotional bonds within and across species. Pryor's subtitle-- What Clicker Training Teaches Us About All Animals-- hints at her (sadly radical) proposition that humans can and should be in dialogue with other animals. Trainers may be more strongly inclined to listen (though many are not), but we all have as much to learn as to teach. More.

In order to describe Pryor's neat sleight of hand clearly, I first need to travel back to Behaviorist Psychology 101 for a quick primer in classical and operant conditioning. All conditioning involves the establishment of novel associations, and the two types are not in every situation distinct, but for simplicity's sake let's say that classical conditioning promotes reflexive, involuntary responses to a given stimulus, whereas operant conditioning engages a creature's will.

If I say "Pavlov," does the image of a drooling dog spring immediately to mind? Is it still there if I tell you "Don't think of a drooling dog"? If so, you are well conditioned to associate both "Pavlov" and "dog" with more or less specific representations of domestic canines, though the dog you imagine when I say "Pavlov" may be more slack-jowled and blank of expression than the one that "dog" conjures in another context. My point here is that there exists no intrinsic relationship between the word "Pavlov" or "dog" and any actual dog (or even the category of dogs), but if you are an English speaker with some knowledge of psychology, these unlike things have been paired often enough in your experience that one invokes the other without any deliberate effort on your part. That's a form of classical conditioning.

Operant conditioning asks a little more from you: intent and action. In this case the salient association is created between a behavior and a consequence. If it becomes strong enough (if the consequence follows consistently from the behavior), it takes on the flavor of a causal relationship, though the connection may be an arbitrary one. Pavlov's dog is the icon for classical conditioning; Skinner's lever-pressing rats embody operant conditioning at its most basic. Rat pushes lever, and out pops kibble. Far out. If you want to preserve the association but don't want fat rats, you can add a cue. Green light on, push lever: kibble. Green light off, push lever: nada, niente, SOL. As with the example given for classical conditioning, there's no ready-made relationship between lights, levers, and kibble. It's dreamt up by the scientist and taught to the rat, simply through temporally close, predictable association.

However little we might like to believe that we resemble rats, humans are subject to the same tendency to perceive causal relationships where none may exist. Like all animals, we seek coherence and control in a stubbornly chaotic world. Nothing undermines our well being more disastrously than a sense of helplessness, so we are apt to exaggerate our agency and influence. Sports fans seem especially susceptible to delusions of this kind: when I am watching my beloved San Diego Chargers on television, a thousand or more miles from the field of action, I become temporarily (absurdly) convinced that my shouted "Come on, D!" or my failure to wear my Quentin Jammer jersey could make or break my team's chances at victory. Athletes themselves take superstitious behavior to comical extremes, baseball pitchers to an art. This is where the line between classical and operant conditioning begins to blur: when a stimulus (or set of stimuli) triggers us to act automatically, even compulsively, will and intent disappear from the mix.

photo credit: San Diego Shooter via photopin cc