Gregory Bateson's Levels of Learning

An exciting chapter in an otherwise exciting book, Steps to an Ecology of Mind, is concerned with Gregory Bateson's application of the idea of logical levels to the concept of learning. His premise is that the idea of logical levels is applicable to processes of learning, and that the ignorance of this fact leads to confusion in the behavioral and social sciences. What I want to focus on in this post is simply an overview of the different logical levels of learning. I hope in future posts to proceed from there, because this seems to be a topic worthy of exploration.

I imagine each successive logical level of learning as a process of systemic activity in which the learning (understood as specific changes in the system over time) comes to occur in ever larger parts of the system. I will begin by describing the first level.

Learning 1 is commonly understood as simple Pavlovian or operational conditioning. An organism is exposed to a specific environmental event which repeats itself, and this has the result of familiarizing the animal with the contingencies of their behavior in this instance. For example, an animal that repeatedly barks and is given a treat during one session with its owner has learned at level 1 the specific barking>treat contingency.

Learning 2 involves a more substantial type of learning. Rather than learning how to act in a specific situation, an organism that undergoes learning 2 learns to approach novel situations and act in a certain way that has been rewarded in other situations in the past. In other words, learning two is the process whereby an organism learns to act certain ways in certain types of situations, rather than just act in a certain way in a certain situation.

According to Bateson, level 2 is the general ceiling for most people's learning. Despite this, learning at level 3 is attempted with some success-and some destruction. The relationship between levels 1 and 2 is the same as between levels 2 and three. Beyond this level three is (and Bateson admits this) difficult to describe.

It helps to go back down and describe level 2 in more detail, and then scaffold off of this for level three. So, level 2 involves the development of the ability to approach a novel situation and categorize it (and then act towards it) in a certain way. Although the situation facing the organism is novel, the organism still retains its acquired ways of acting. Learning at level 2 gives the organism an intelligent basis upon which to choose a certain alternative.

In contrast, learning 3 involves learning to intelligently choose between sets of alternatives themselves. The choice of alternatives is no longer at the level of the individual situation (as it was at level 2), or at the level of the individual action (level 1). The choice that must be learned to be made intelligently occurs at the level of the sets of alternative interpretations.

So, someone who has learned at level 3 would, upon approaching a novel situation, say to themselves "what set of interpretations offers the most promise in this instance?" Once a decision has been made, the activity occurs on level 2-which interpretation within the selected set will I make-and then on level 1: what is the appropriate action to take in this context?

Linguistic Structure and Overcoming the Barriers of Autism

Bullshitters rely on, perhaps more than any other thing, the structure of language. If someone doesn't know the exact meaning of a particular word, a vague meaning can be constructed if the word has been heard before in conversation because such a context often serves as a pointer arrow directing inquiring minds towards an unknown word's meaning.

Can this be taken a step further? What I mean by this is can one figure out what a whole part of a language (that refers to a specific domain) is simply by making inferences about its structure as it relates to the structure of other parts of the language? This may be confusing, though I think I can ground it by giving the example that got me started on this whole line of thinking.

I work with autistic adults as my job. Some (like Simon Baron-Cohen) have theorized that autism is the absence of an understanding of mental states which leads to problems with communication, social interaction, and imagination.

In line with this theory is the fact that many autistic people have very inhibited language abilities, though some come to have an impressive command over language. Several of the people that I work with fall into this category, and I often find myself wondering what sense they make of those parts of language that refer to mental states. In my experience they don't use these parts of the language outside of situations in which they've been explicitly taught to do so by rote.

Perhaps it's possible that the structure of language could serve as a guide towards the (admittedly artificial) construction of mental concepts in high functioning autistic people who have well developed language abilities. The existence of the linguistic domain of mental terms would not be initially understood, but this is no reason to believe that its structure would escape a careful mind. This structure would exist both within this domain (intrastructure) and between this and other domains (interstructure).

In the coming days, I hope to find out more about how this structure may aid in the acquisition of mental concepts, or something like them...

continuation of the lower post

I've been continuing along the line of thought outlined in the post directly below, and have come to the following conclusions concerning the"realization" method of overcoming undesirable (but still powerfully influential emotions).

I was wondering why the realization the the emotion was "wrong" was so ineffective in the instances I talked about (with the alcoholic not wanting to obey the urge to drink, or the obese person not wanting to follow the urge to eat more). I contrasted these cases with cases in which a certain realization can lead to the abolition of an emotion. As an example of this, take a person who suspects that, at this very moment, his spouse is cheating on him. This person knows where his spouse is, and wants to break in and confront her in the act. Doing so, he goes to the room where she is, breaks down the door and says "AHA!," only to immediately realize that she was indeed not cheating on him. At this moment, the powerful emotions of jealousy and fear that he had probably been feeling likely disappeared, perhaps leaving a residue of anxiety, but largely evaporating. These emotions were declared to be wrong or inappropriate given what was actually happening in the situation he arrived at.

How can it be the the knowledge that the emotions are wrong in this instance immediately lead to the emotions' demise while with the drunk or the obese person, the emotions persist in the knowledge that they are wrong? The reason for the difference, I reasoned, has to do with what I will call the logic of emotions. All emotional reactions are in some way logical. There are certain conditions under which they arise, and other conditions under which they are not activated. The ways that any given emotional reaction obey this is called their logic.

With the suspecting husband, the emotion disappeared immediately after he saw his wife alone because this sight changed the conditions that led to the emotion, making its presence no longer appropriate. With the alcoholic and the obese person, the realization that their pesky emotions were inappropriate did not violate the logic of the emotional response. Instead, their ideas about the drink/food desiring emotions being wrong was not part of the logical system governing the activation of these emotions.

The desire to drink operates in accordance with the conditions of a variety of factors in the alcoholic. First of all, drinking is equated with drunkenness, a positive state which, for the drunk, is preferable to the average non-drunk state. When the alcoholic is not drunk, the system governing the desire to drink is aware that the drunken state is currently unfulfilled, and puts pressure on the system to change this. If this particular alcoholic wants to stop drinking, this urge is, on an abstract logical level in the alcoholic's mind-the level that realizes the futility of drinking and the problems it causes-, subordinate to the desire to not drink. But this abstract logical level does not have complete control over the system.

The emotional logic that led to the desire to drink is unaffected by the knowledge that drinking is not a good solution. This does not change its conditions for activating the emotion. You might say that these are two logical systems speaking a different language.

At this point, it is now appropriate to bring in the alternative explanation for the emotional discord: the competition model. In the competition model, the outcome of competing and contradictory desires is control by the desire (the emotional-logical system) that is strongest at a particular time. While the realization model accounts for how the logical contradiction is not resolved, the competition model accounts for which logical system has the upper hand. In the case of the alcohol/food addict who cannot stop eating or drinking to excess, the emotional logical system governed by the satiated-drunk/non-satiated-sober awareness is stronger than the emotional logical system governed by the doing the right thing/not doing the right thing awareness. Presumably, their relative strength is a factor of the potency of the conceptual duality (drunk/sober, or "do right/do wrong").

Overcoming Drug Addiciton

The topic of overcoming a drug addiction has risen to the center of my interests in psychology in recent weeks because it is such an interesting and illustrative example of the nature of our embodied existence as human beings. By "overcoming drug addiction," I'm referring to the process by which a person attempts to stop using drugs, while at the same time deals with the feeling of wanting to use drugs.

The coexistence within one person of these two competing desires is revealing of what I have referred to in previous posts as the constraints on the adaptive action of living systems. To briefly state this, the possibility for competing and contradictory urges to coexist in a single living system is a likely problem that arises from the fact that the system is autopoetic, and can never exercise completely adaptive control over its whole being. The last point is not a complex one, it's actually an extremely obvious observation about how physical systems have to work. This state of affairs doesn't have to happen, but there is nothing to prevent it from happening given the right conditions.

With drug addiction we can see, in society, the manifestation of this state of affairs in a certain type of way. The drug user is the victim of these competing urges that send contradictory signals, each attempting to have the body controlled in a certain way. For the drug addict who wants to stop using drugs to actually stop using drugs requires the development of control over the body by a value system that recognizes the futility of drug use. In order to attain this control, the value system must be more powerful than the value system that seeks to continue drug use. In other words, the anti drug value system must have greater leverage over the affective processes of the body than the pro drug value system.

This is actually just one view of how this might happen. Another possibility is that the user who formerly "believed" in the emotions that led to drug use, learns to stop believing in these emotions; he/she learns that they are produced by the body in error, from a bodily system that does not "get" the overall picture of what has to be done.

I will refer to these two views as the competition view (the first) and the realization view (the second). Which is correct? Are they compatible?

If we look to our own experience, some aspects of the realization view are very reasonable. Even if one is not a drug addict, it's easy to think of a situation in which one has an emotional response that is undesired (e.g., a dieting person wanting food). Our own experience shows us that we can learn to respond to such emotions by saying "that is exactly the kind of affective reaction that I want to stop happening." In spite of our awareness of its being the problem, the emotional reaction continues to hold sway over our decisions, and so, in the example, the dieting person gives in to eating a cheeseburger, despite thinking what a terrible decision this is.

So, it seems that we can evaluate our affective reactions, but in many cases, this evaluative capacity does not hold sufficient sway over the rest of out body to suppress the emotion's power (though it does in some contexts).

I'm out of time for today, but I will continue to explore this issue in the next post (as the task I set out for myself today remains unresolved). I want to finish up looking at the "realization" method of overcoming emotions and then look at the possible merits of the competition method.

Motivation for acting

When we talk about what people do, especially in the third person, we tend to treat actions as the products of a unified decision-making machine that is supposedly what the human body is. In reality, our lives and our actions are more complicated than this.

Actions are the product of what is called in artificial intelligence a value system. Basically a value system is a system that makes use of a network of action concepts and places value on them, in many cases the value is contingent on a certain place and time. A primitive value system might make use of the concepts of hungry/full and get something to eat/rest, linking these up in a way that leads the creature to go and get food when its hungry and rest when it's not.

Human beings are complex creatures, but we too have value systems like the simple one described above. The difference between us and a simple creature is that we have several value systems existing side by side which compete with each other. At certain time, certain conditions may lead to a certain value system being given precedence over others, and with changing conditions through time, this may change, leading to a different value system to assume control. The experiential product of this is our continual feelings of ambivalence about whatever activity we're pursuing. When we sit down to study for a test, the value system that values scholarly achievement has taken control, and is pushing other value systems (such as the "have sex" system into submission.

The description above was meant to illustrate what I mean by coexisting value systems, and it overlooks a few things. First of all, the value systems in our lives are not completely discrete. Different value systems work together in such a way that the activities that are pursued under the guidance of one value system may be rationalized by another (if they initially produce conflict with the goals of that second value system). In some cases, of course, two value systems maybe be unable to resolve their differences and one will eventually be pushed completely into submission.

The focus of this post is not on these intricate interrelationships, but on the broad fact that we are subject (in our daily pursuit of activities) to guidance from an array of different value systems. When trying to understand another persons actions, it is therefore incorrect to try to create a unified schema for why they act as they do. Instead, it is beneficial to try to understand another's actions as the product of several competing value systems.

Experience and the organization of neural networks

One of the main areas of focus in this blog has been the relation between phenomenology and biology. I've explored this theme from many different angles, and, I'm afraid from the same angle many different times in an attempt to organize my thoughts and create a coherent foundation upon which they may develop. Today, I'm revisiting this topic again from a slightly familiar angle, but I believe that this time the ideas are more coherently organized than in the past, and that a new reader of this blog can find the core ideas well expressed in this post, without having to reread all previous posts.

The issue I want to talk about today is the question of the place of phenomenology (experience) within science, specifically psychology and cognitive science. Traditionally, phenomenology has been the classic example of the kind of domain that is unsuitable for science. It's been my position that although some phenomenological data are inappropriate for scientific work (especially clear scientific work that others can understand easily), there are certain kinds of phenomenal "facts" that are appropriate and actually very helpful for the progress of science.

In a project last year, I focused on exploring the connection between the phenomenal experience of bodily control and the anatomy of the nervous system (and the body as a whole). I thought, and still think that exploring connections such as these is a fruitful way to explore the way that biological systems are associated with consciousness.

A major conclusion of this work was the emergence of the idea that biological systems can never be completely adaptive. By virtue of their existing in the physical world, and being bound by the constraints therein, no living system can ever be completely adaptive since being completely adaptive would require having complete control over its own functions. In a physical world, such an idea makes absolutely no sense. A system can have control over its parts, and can use this control (with the aid of perceptual systems and the interpretation of sensation) to act adaptively, but the operations of the entire system cannot all be controlled adaptively since every mechanism of adaptive control is itself uncontrolled until an additional structure controls it. The quest for complete control ends up with an infinite regress of additional control structures.

Now, a central feature of our embodied existence is this conscious divide between control and no control. We have no direct control over the environment, and no control over many parts of our body, but we of course have control over some. What I'm trying to say here is that the constraints on physical systems that I talked about in the last paragraph show up conveniently in our own embodied experience. This comes as no surprise if one accepts the view that there is a complete correlation between phenomenal states and biological states in any given living system.

Recently, I've been thinking about the extent to which (and the ways in which) we have control over our mental processes. While we can of course control our arms and legs, we can also control our thoughts-but there are limits to this. While we can bring images into our mind at will, we cannot control other aspects of our mental states-such as awareness, otherwise people like me would never suffer the embarrassment of falling asleep in class.

What this means is that our phenomenal awareness of those mental processes that are and aren't under our control is a reflection of the organization of the nervous system. The fact that I can bring certain images into my mind at will reflects the fact that my nervous system has mechanisms of control over the production of mental images. In contrast, the fact that I can't always make myself fully aware or bring myself into deep concentration when I want to is reflective of the fact that full awareness is not amenable to conscious control-or very much conscious control (maybe biofeedback could change this).

By exploring the issue of control with regards to our mental processes, we can make valuable use of phenomenology to make sense of the biological organization of the brain at a very high level. By understanding our phenomenological states in terms of these constraints we are in fact creating an understanding of the underlying neural organization. This is especially helpful given the complexity of the nervous system-a complexity that bars us from seeing the kinds of organization I'm talking about by looking at the neurons themselves (this would be a herculean task).

How much control can we get over our thinking?

In today's post I want to talk about the kinds of influence that we can have over our thoughts. This is by no means a conclusive look at this subject, but merely a first glance intended to clear up very obvious dilemmas, as well as to see how far can be gotten by just thinking some issues out.

What I'm concerned with here is the extent to which we are able to influence the kinds of thinking that we do as human beings. This subject came to interest me after car ride that I took last week. During the car ride, I had what I considered to be a really interesting thought. The content of the thought itself is not important here, suffice it to say that the thought was interesting, and I was excited that my mind was working in a way that led to the thought emerging in my mind. Reacting to this nice surprise, I began to wonder if my positive reaction to the thought might in any way lead to more thoughts of that type in the future. How nice it would be, I thought, that by appreciating a style of thinking, one could guide ones self toward more of that type of thiking, or influence the future probability of it.

Since that day, I've thought about this a good deal, and have made some important progress in understanding it.

First of all, finding that what one is thinking about is interesting does encourage us to think the same thought again. The result of this is that it becomes easier overtime to think about a particular idea, and, I would wager, to think about ideas that are similar in structure to that idea. This is just a matter of classical conditioning: Actions/thoughts lead to either positive negative reactions, and these reactions influence the future probability that we will execute a particular thought/action.

As soon as I began thinking about this, a new issue came to dominate my mind. Although a positive reaction to a certain thought may lead us to rethink that thought in the future (thereby strengthening the connections resposible for that and similar thoughts), a reaction to the processes that led to a thought ("oh that was so clever!) would not be able to influence the future occurrence of those processes. In the example of my thinking in the car, I was excited not just by how interesting I thought the idea I had was, but also by the fact that it was a really creative idea: it was a spontaneously emerging idea that was grounded in concepts I had been thinking about. The question is: By reacting positively to the fact that the idea was creative, could I support the emergence of more creative thoughts in the future?

Operant conditioning accounts for how a positive reaction to one idea can lead to rethinking it and similar ideas more easily. But I'm not sure that it could help us to think more (e.g.) creative thoughts if that's what we want to do. The conditioning mechanisms above worked because the idea itself led to an affective reaction, and therefore, the idea itself was strengthened. The creativity of the idea did not lead to the positive reaction, though it can be held accountable for why the idea was interpreted as creative.

Therefore creativity can not be linked to the affective reaction, it remains untouched by the conditioning process. Broken down, this is what this looks like: A creative process leads an idea to pop into a persons head>>they interpret the idea as the outcome of a creative idea generating process>>>This is something they value, and consequently they are happy to have had the idea>>>They have good feelings about the situation and the idea.

This sequence shows that, although a person can value creativity, as well as interpret ideas as being creative, there is no direct link between the actual creative processes themselves and the positive feedback-Unless the creative processes are reactivated every time the (originally) creative idea is re-thought-which doesn't make logical sense.