Thursday, December 7, 2006

What's the difference between you and your dog?

For a very long time human beings were defined as the "rational animals". In other words, is has been supposed that what makes us human is that we are very apt problem solvers, that we are able to recognize patterns and find the hidden regularities (and "laws of nature") existing beneath the empiric surface of the world.

However, probably due to modern art, this kind of definition isn't that popular nowadays. Imagine advising a contemporary poet to be a rational animal! Moreover, animals are themselves very good problem solvers and thus, from such a viewpoint, the animal/human difference is only quantitative at best. But is the human ability to change the environment so dramatically just a more extensive beaver-like behavior?


Few people would still define today human beings as the "rational animals". It's more likely that one would hear them defined as the irrational animals! The focus has changed on claiming that we humans are intentional beings, we have purposes and goals, while the animals are mere survival machines driven solely by causal factors. We act in certain ways because we want certain things. The animals behave in certain ways because they are driven by their instincts. In the same way as a bullet does not have the power to choose its target and it is simply bound to go in a certain direction, the animals are supposed to be bound to obey their genetically-inscribed "nature".

"When the plant, the rock, the animal start their existence they already are all they can be and thus all they will be," wrote Ortega y Gasset in The Interpretation of Human History. "Man on the other hand, when it begins to be it doesn't have a predetermined path, it becomes what it will be out of a vast universe of possibilities. Man therefore has the power to choose, and doesn't have the power to abstain from choosing. No matter whether one likes it or not, one is engaged in every single minute in the act of deciding what exactly to make out of one's life. As a consequence, this freedom to choose, which is man's privilege in the natural world, also takes a tragic character because one is condemned to be responsible for what one becomes. One is responsible about oneself, something that doesn't happen to the rock, plant or animal, which exist in innocence having an enviable irresponsibility. Due to this condition, man is that strange creature that wanders around carrying in itself both an accused and a judge, both of which are oneself." (my translation)
In this passage, Ortega y Gasset puts forward the classic existentialist doctrine about freedom and responsibility. According to this view one isn't responsible just for what consequences one's actions have upon others, but one is also responsible for what consequences one's actions have upon oneself. This idea is used for the purpose of denying the legitimacy of statements like "It's not my fault (that I did that awful thing X), I did what I did simply because that's how I am, and I haven't designed myself to be this way (my genes designed me and/or my education and environment designed me)". According to the existentialist point of view, one is in fact designing what one becomes, so there is absolutely no escape from responsibility. "My genes + environment designed me to do this" is seen as just a cheep trick.

I don't wish to argue against this existentialist point of view, as I generally agree with it, although I think it needs more elaboration, but I want to remark that the first step in the argument, the claim that plants and animals are virtually identical to rocks, is deeply flawed. Plants and animals might be similar to rocks from a moral point of view, but not from a physical point of view. They certainly don't have a predetermined path in the same way as a typical causal system has. Even the simplest bacteria is an intentional system, in the sense that its motion (its behavior) is always aimed towards something which is to be found in the future and it strives toward that thing in spite of the random forces acting against it.


If the first definition of the human being doesn't go well with the artists,
this second definition doesn't go well with the scientists: A scientist believes a certain theory only if that theory helps him or her make some real predictions. In this case, a good theory on the inner workings of animals must allow us to predict something about the animal behavior. But the causal theory of animals doesn't predict anything – because the animals are supposed to be "very complex" causal machines. Moreover, as anthropologist Michael Tomasello has noted, saying that some behavior is instinctive (or innate) doesn't really explain anything, it only attaches the label "instinctive" to that behavior. In discussing the shallowness of the nature-vs.-nurture dispute he wrote [in The Cultural Origins of Human Cognition]:
"... this discovery [that something is instinctive rather than learned] should not stop the process of inquiry — we should not simply say that X is innate and so our job is done — but rather it should lead us to ask other questions ... The search for the innate aspects of human cognition is scientifically fruitful to the extent, and only to the extent, that it helps us to understand the developmental processes at work during human ontogeny, including all of the factors that play a role, at what time they play their role, and precisely how they play their role."
Similarly, just saying that some animal behavior is instinctive leads nowhere. One wants to know the details. In this endeavor one has to start not from some ad hoc philosophical assumptions about animals (such as that they are entirely causal complex machines), but from what actually works in describing their behavior. And one can predict the behavior of animals only by assuming they have goals - by assuming that they strive to achieve certain things in their future. One cannot predict their behavior just by assuming they are entirely determined by what has happened in the past.

For example, one can assume a bird has the purpose of surviving, and thus it builds a nest, and thus it goes around gathering small tree branches and leaves. It is virtually impossible to describe such a behavior in causal terms, i.e. starting from the opposite direction: to assume there is some cause that determines the bird to gather a leave and that there is a cause that constraints the bird to put all its gatherings in the same place in the shape of a nest and so on. One would have to assume countless unseen "causes". It is much simpler to assume the bird has purposes. Of course, the explanation doesn't stop here either - the biologist now has to ask where those purposes have come from. This, however, is an entirely different question than the one Ortega y Gasset (or the biologist inspired by him) would ask. There's one thing to wonder about causes and another to wonder about goals.

Only once we have accepted that the difference between humans and animals does not lie in the contrast between our intentional nature and the animals' causal nature, we can truly start to understand the real difference. The Oxford zoologist Richard Dawkins noted that we, unlike most animals, have not only goals, but also the ability to decide what goals to have. For us the goals are not just given. According to Dawkins, this is precisely what provides us with the ability to attach meanings to various things or events. The meaning of something is simply the goal to which it serves. In the most familiar case, the meaning of my words is the information I intend to transmit to you with the help of those words - the words are the means and the transmission of information is the goal.

In the absence of the ability to choose one's goals the world and everything that happens in it would be meaningless (or the meaning would simply be imposed from the exterior). The animal that has a predetermined set of goals, all of which are subordinated to its ultimate goal of passing on its genes, cannot be said to have a truly meaningful life (at least not in the usual sense of the word "meaningful") because it is not really free. Consequently, nor is it responsible.

So, Dawkins successfully maintains the "enviable irresponsibility" of animals, their amoral nature, while acknowledging that they have goals. While Ortega y Gasset has claimed that responsibility is the outcome of having an intentional nature, i.e. it comes from that fact that we have goals, Dawkins claims that responsibility (as well as meaning) comes from our ability to choose what goals to have. A cheetah that hunts antelopes for a living cannot be said to be immoral, but the Canadians that hunt baby seals for a living can.
(One can wonder whether an orca, which apparently is self-aware, can be called immoral for playing with its prey in such a cruel way. Or are chimps immoral?) Most animals choose only what means to use in order to arrive at the nature-given goals, while humans (and probably also, to a lesser extent, other animals such as chimps or dolphins) choose even the goals. But why does this difference occur?


To find out try the following experiment. If one plays hide-and-seek with his or her dog, one finds out a very interesting fact. At first, it seems that the dog understands the game, it understands that it has to find you and it is able to adopt this purpose. Its behavior is indeed goal-driven: it has the goal of finding you. However, when you play again and again, you can notice a curious fact: when the dog starts searching for you it always goes first to the same place where you have hidden in the previous game. The dog does not understand that you have the goal of hiding from it! It thinks that if it has found you in one place in the previous game you are probably there again. In other words, the dog acts like a Bayesian robot that uses its previous experience to guess the best course of action for the present, assuming that the present will mimic the past. It adopts the goal of finding you, it does not act like a strictly causal entity, but nonetheless its reasoning machinery assumes that you are such a causal entity!

Imagine that you are set to pick mushrooms in a forest. At first you know nothing about where the mushrooms are most likely to be found, except that they are somewhere in the forest. This lack of knowledge means that, from your perspective, the plausibility of finding a mushroom in any given place is equal to the plausibility of finding one in any other place. In mathematical terms, the probability distribution of finding a mushroom in any location inside the forest is a constant (this is the so-called Laplace principle of insufficient reason which tells you to assign equal probabilities to epistemologically equivalent options). If we divide the whole forest in N locations and there are n mushrooms in the forest, the initial probability (plausibility) of finding one mushroom in any given location is n/N.

But when you search for mushrooms again and again, and as you develop a certain experience at this job, you notice that in some locations, or types of locations, you find mushrooms more often. So you change the initial assessment of the probability distribution accordingly, giving more weight so some locations and less to others. This process of altering the initial probability distribution and obtaining a new, more complex probability distribution, is described mathematically by the Bayes formula. This formula describes fairly well how you learn from past experience.

The experiment with the dog shows that it treats the game of hide-and-seek in the same way as we treat the game of finding mushrooms in the forest. However, there is a very large difference! The people who hide have the purpose of not being found, while the mushrooms "hide" wherever the right causal conditions for their growth are met. A mushroom grows wherever its spore finds the proper environment. (Arguably, natural selection has imprinted in the mushrooms, to some extent, the goal of hiding, but this isn't the point I'm trying to make here. The point is that one finds the mushrooms using causal reasoning - it is a good strategy to wonder about the conditions that favor the
mushroom growth; on the other hand, one finds people hiding using intentional reasoning - a kind of reasoning the dog fails to adopt.) So, a dog playing hide-and-seek treats you as if you were a mushroom!

I don't intend to argue here that we don't use the Bayesian reasoning. [Actually, we do employ other kinds of reasoning. For example we don't seem to generate metaphors by using a Bayesian mechanism. Moreover, empiric studies of doctors have shown that they make gross errors in estimating the probability that somebody is sick, errors one would not expect if we were actually using a Bayesian induction mechanism.] The point is that the only inputs a dog can use are about what has actually happened in the past. Our reasoning machine however can also have a different kind of input - we can use assumptions about goals. How do we infer what goals one has from one's behavior?


This is a subtle difference: there is a difference between having goals and being able to attribute goals. In other words, a living being can have unconscious goals - in fact, most goals are unconscious. Most animals have only unconscious goals. We on the other hand, besides having such goals, also have some conscious goals. And we are also able to asign goals to others.

The simplest empiric test of detecting consciousness is the mirror test: if an animal manages to recognize itself in the mirror, it means it has a certain awareness of itself (whatever that means). Insofar, the self-aware animals are: humans, chimpanzees, orangutans, gorillas, dolphins and orcas (killer whales).

Using the things said above one can understand why some animals manage to recognize themselves in the mirror while others don't. One does not recognize oneself in the mirror because one recognizes the behavioral similarities between oneself and one's reflected image – this is virtually impossible to do because one cannot really observe oneself (we have a very improper perspective). One recognizes oneself in the mirror because one ascribes goals to the person in the mirror and ascribes goals to oneself, and then one can realize that the image has the same goals as oneself – therefore it is the same person!
As a funny side note: The case of Douglas Harding. Douglas Harding was a man who became famous by declaring that he doesn't have a head. Harding realized that, while he could observe that other people had heads, he could not observe directly that he himself had a head. Thus, he concluded that the idea that he had a head was in fact the product of his imagination, and not a real empiric fact that could be trusted.

Apparently, his eccentric belief was accepted only by a Zen group and it became "the headless way". Oddly enough, Susan Blackmore also presents this idea in her book about consciousness in a chapter about "awakening". She criticizes Dennett and Hofstadter for being too harsh on poor Harding.

But did Harding really believe what he claimed to believe? Blackmore missed this: if Harding had truly believed what he was saying he would have failed the mirror test! The mirror test consists in painting the animal on its forehead, without it noticing it, and then placing it in front of a mirror. A self-aware animal, such as a chimp, wipes its own forehead, while most animals, such as e.g. a gibbon, try to reach the animal behind the mirror to wipe its forehead.

Of course, probably nobody did the mirror test on Harding, but I doubt he would have failed it. I think he would have wiped his own forehead! But even if he would have failed it: this would only
have shown that he was no longer self-aware, not that he had waken up. He would have reverted to a state similar to that of a gibbon. Or maybe this is insulting to gibbons - after all, they do know that they have heads!

The point of this story is only to show that we don't recognize ourselves in the mirror based on the similarities between our motion and the motion of our image. My speculation is that we recognize ourselves in the mirror because we attribute goals to the person in the mirror and we realize that it has the same goals as ours.
Based on this interpretation of the mirror experiment, it follows that we are the set of all our goals. And this set has a structure - our identity is thus structured. Not all our goals have equal importance and some goals are subordinated to other, more general goals. As Gasset and Dawkins have noted, our goals, as well as their hierarchies, are not set in stone by nature, we are free to choose them. But why are we?

The answer is that we are free to choose them because, and only because, we are aware of them. In fact, being aware of something is synonymous, from a functional point of view, to being capable of changing it (or preventing it from happening). A fish for instance is aware of a predator if it is capable, at least in principle, to avoid being caught. You are aware of the visible part of the light spectrum in the sense that you can avoid seeing that kind of light - if, on the other hand, you are illuminated by X-rays or radio waves you are not aware of this - you cannot avoid these rays
without using artificial detectors (and you become conscious of the existence of these other parts of the light spectrum once you have access to such detectors).

So, the idea is that, from a purely functional point of view, being aware of something implies the ability to avoid that thing and having the ability to avoid something implies that you are conscious of that thing. In other words, being aware of X and having the ability to avoid X are equivalent concepts.

What happened during our biological evolution
was that one of our ancestors had developed the ability to understand the goals of other animals. This is a very useful adaptation as it makes one better at playing hide-and-seek, and many animals in nature play hide-and-seek on life and death. And when this ancestor then applied this new ability to itself it became self-aware. Being self-aware means being conscious of one's own goals - as one's identity is defined by one's hierarchy of goals. And being conscious of one's own goals means being able to change them. Thus, this was the birth of responsibility, morality and meaning.

So, at least to a first approximation, the functionalist approximation, awareness is the ability to attribute goals (either to oneself or to others). Whether one has such an ability can be detected by empirical means such as the hide-and-seek game (which is a little bit more general than the mirror test).


One might be let to believe that having the ability to choose our goals means that we can choose any goals and any hierarchy or goals. This however, doesn't seem to be true. You might not like it, but it appears that there are certain laws that describe and predict the way we decide what goals to pursue based on what abilities (skills) we attribute to ourselves. Actually, without such laws we could never guess what goals others have, and thus we would have never become self-aware.

The most important such law is the law of diminishing marginal utility. This law incorporates the observation that people tend to assign a higher subjective value to things that are less accessible to them. Where do you value water more, in the desert or in the city? What are you trying to acquire, what you already have or what you don't (yet) have?

This law can be used to discover what goals somebody has. One first needs to assess what abilities X has. As a first approximation, one assumes that X believes to have the same abilities as we believe him or her to have. (Young children work only within this first approximation.) Applying the law of marginal utility we derive X's hierarchy of values.
(I have described this process in more detail in this article.) Then, we assume that X strives for the most valuable things it can, given its skills, reasonably acquire. Thus, based on how the environment is and on what skills we think a person has, we manage to estimate what goals that person might have.

We then observe what X actually does. Based on this observation, and the difference between our prediction and what has actually happened, we can modify
either our assessment of X's true abilities or the assessment of how well X knows himself or herself. This is how we learn about other people's beliefs about themselves (and this is one door to mind-reading in general). This is how you get a feeling of what somebody else thinks about himself even if s/he doesn't tell you what s/he thinks about herself/himself. Young children, up to around 5 years old, seem to be incapable of making the second kind of assessment change. This is why they are so gullible about what other people can actually do. (Nonetheless, in other aspects children seem to be much saner than adults; for instance, in a study about what children think of unseen entities it was found that they are more likely to believe in the existence of microbes than of things like God.)


Should we get beyond the functionalist approximation? Or maybe this isn't an approximation at all, but all there is? Getting beyond the functionalist approximation means seeing consciousness not only in purely behavioral objective terms (what X can do, i.e. consciousness = the ability of attributing goals to others and to itself), but also in subjective terms. When I feel about something I feel in a certain way. What is that feeling all about?

A more precise way of defining the subjective aspect of consciousness is to say that there are some things in regard to which you have a first person authority - i.e. if you feel they are true then they must be true. For instance if you feel that your back hurts, no one could prove to you by any objective means that this pain does not really exist.

The problem of functionalism is then to show how such subjective (first person) aspects of human life can be constructed from objective (third person) phenomena. There are some philosophers (most notably Daniel Dennett) who argue that this is possible but insofar nobody has actually put forward a model that describes how this can be
literally done - a model that could be used for instance to actually building a robot with subjective experiences.

But before starting to build such a robot one must face the skepticism that one can never tell whether a creature has subjective experiences. This is the so-called problem of other minds and it is much easier to solve than philosophers tend to believe.
This issue however, will be treated in another essay as it is too remote from dogs - although one can wonder how come that we know that dogs have subjective experiences.

So, insofar we can maintain that what makes us different from dogs is that we not only have goals but also have the ability to perceive that others have goals too, and consequently, as we switch the focus from others to ourselves, we are also capable of becoming aware of our own goals, and thus capable of changing them. This brings morality and responsibility to the stage. Moreover, by comparing our assessments of others with what they actually do we can become aware not only of their goals but also of what they believe about themselves. Finally, due to the fact that people generally believe what they think it is useful for them to believe, by knowing what X believes about itself one also gains access to X's beliefs about other things. However, it's important to note that this isn't the only path to other people's beliefs - there is at least one other path that goes via emotions and happens thanks to the mirror neurons.

No comments:

Post a Comment