A few weeks back, I was showing “Avatar – the Last Airbender” series to my daughter.  It took only a little more than a week for us to cover all three seasons (or Books, as they’re called). In addition to cultivating in her a refined taste for high-quality cartoons, I hoped that it would help her learn English. Not only was I not disappointed with my daughter’s learning, watching her learn also gave me some very valuable insights into associative learning and the root of stereotypes and prejudices

My daughter had very less prior experience listening to English. One of the first words that she picked up, and one that was reaffirmed in her mind throughout the series is the word “attack”. After all, “Everything changed when the fire nation attacked” is part of the introduction to every episode. After watching an episode or two, she noticed this word, and asked me what it meant. The easiest way for me to explain English words to her is to give their meaning in Malayalam. In this case, however, the Malayalam word for attack was equally new to her and wouldn’t help. So she had to learn this word in the way children naturally learn languages – by associating it with something that they see.

“Attack, is what the Fire nation is seen doing” – that was roughly the explanation I could give her, in which Fire nation was the group of guys in red uniform (yes, she knows red and can identify the fire nation). But fire nation soldiers were doing a lot of different things – they were marching, “firebending”, destroying buildings, hurting people, and more. So it is not easy for her to identify which of these things that they do is the actual attack – she might have formed the association between all or a few of these actions and the word attack. Next time she sees another instance of attack, that association will get refined. Over time, she will get a fairly consistent notion of what that word represents – one that is consistent with the examples she has seen, and also with what other people take that word to mean. That is when she’d have learnt that word, and this is how we learn a language.

Associative learning is not limited to learning languages. In fact, it is by forming associations, later to be refined, strengthened or discarded, that human beings and animals learn. This is also how we form notions, prejudices and stereotypes.

At the end of the series, when evil is finally vanquished, and good Prince Zuko is crowned the Fire Lord (King of the country that had started the war and had been shown as evil till that point), my daughter asked innocently – “So has Zuko turned evil now?”. It was only then that I realized how she had come to associate the Fire Nation insignia and headpiece with evil. This is no different from how many people come to associate the Swastika or similar symbols with their objects of fear or enmity. Even today, use or display of Nazi symbols is outlawed in Germany.

While I could explain to my daughter that the symbol itself was neither good nor bad, and it depended on the person who wore it, I am sure that many of my own opinions, prejudices and suspicions could be similarly misplaced. Often, this cannot be helped because only a fine line separates the processes of learning and of forming a prejudice. They both begin by forming an association. In learning, one is willing to refine this association in the light of more information, but in a prejudice one is often emotionally attached to the association and thus unable to change it even when later observations disprove that association. It is easy for me to say that Fire Nation insignia is not in itself evil; it will not be as easy for those who were crippled or lost their loved ones to the war to change their mind about that symbol.

So there are two fundamental requirements for being a good learner – the ability to form associations, and that to improve them over time. When we stop forming new associations, we stop learning. Not discarding incorrect associations even when proven wrong leads to incurable prejudices. This process of learning is what makes human learning versatile and empowers them take decisions even with incomplete information. It is on these very same counts that modern general purpose AIs like Google’s Deep Mind are superior to  human intelligence, enabling them to beat humans in their own forte, such as in a game of GO.

With the immense computational power at its disposal, an AI can process huge amounts of data and form millions of associations (possibilities) in no time, resulting in better conclusions. It also lacks any emotional attachment to any of these associations, allowing it to discard or refine these as required. In the manga/anime series Hunter X Hunter, there is a character named Komugi who is shown as being unbeatable at a highly complex (fictional) board game. In one of her games against a highly competent opponent, when the latter makes a very intelligent move, she hesitates for a minute, before countering it and defeating him. She later explains how she had herself come up with this move earlier and used it to win an important contest, and how beating that move made her feel like she was having to kill her own child. A “cold” AI will have no such weakness.

An important aspect of intelligence, be it human or artificial, is deriving conclusions from limited, incomplete knowledge. This involves generalization, extrapolation and approximation, which are more valued in practical, applied, fields of learning than in abstract ones such as mathematics where focus is more on rigor and absolute surety. There is this story of an astronomer, physicist and mathematician who, while travelling through Scotland in a train, see two black sheep and arrive at three different conclusions. “The sheep in Scotland are black!”, exclaimed the astronomer. “We can only conclude that at there are at least two black sheep in Scotland”, interjected the physicist. “All we know is that there are at least two sheep in Scotland with at least one side black”, said the mathematician, rebuking his co-travelers for their lack of rigor.

Most of the things we “know” and use to build the world as we see it today are based on models built from limited information that is available to us. When presented with a counter example, these models can either be reworked to account for exceptions, or the laws can be continued to be used while we “acknowledge” the exceptions or inaccuracies involved, and still ignoring them for all practical purposes. A classical example would be laws of classical mechanics, which have been proven to be inaccurate by results from the theory of relativity, but still remain a very reliable tool for most practical calculations in day-to-day life. The more complex equations are reserved for cases that demand a high degree of accuracy or when dealing with speeds comparable to the speed of light in which classical mechanics fails to give useful results.

It is seen from these example how practicality lays stress on arriving at a useful conclusion even when we cannot be one hundred percent sure about it or, when we know that we are only making an approximation. It is this very same principle, or human tendency, that has helped the species to survive and progress that also leads to crude generalizations and stereotyping that so many people are up in arms against these days. In my opinion, people who want to break all stereotypes are those who fail to appreciate the value that they provide, most likely because they themselves find it hard to fit into some of those, or consider some of those as disadvantageous to themselves.

The truth, however, is that breaking a stereotype does not destroy it as many think. Breaking and forming are only natural steps in the evolution of stereotypes and, if anything, it makes them stronger. A stereotype is only replaced with another that is more accurate (and thus harder to break) – it is seldom discarded altogether. If a group of people are disadvantaged by one stereotype, replacing this stereotype with another will be at the expense of another group; completely discarding that stereotype will be against the interest of a still larger group, or the society at large.

Those who worry too much about stereotypes are being concerned about how a particular group that they belong to is being viewed, and how that negatively affects them. There are exceptions for every stereotype and every person is likely an exception to some of the stereotypes that are associated with them. It is a common perception that “Indians love cricket”. I don’t love cricket, but when I am in a group, it is assumed based on my nationality that do, and the subject often comes up for discussion. The way this stereotype helps people is that it gives them some guidance as to what to talk about when they meet a stranger from India. In this aspect, it also helps most Indians who do conform to the stereotype. I don’t confirm to it, and am thus disadvantaged by being party to conversations that don’t interest me.

My not liking cricket does not mean that the stereotype is wrong. It still holds for a vast majority of Indians and is useful. Crying out loud that I don’t like cricket and that the stereotype should thus be discarded doesn’t help. By doing this, I can change the perception about myself, but not the truth about the group I am part of. I am only one in a billion. While there is no real reason for me to be disturbed by the existence of such a perception which does not hold true for me, if I really want to change it, my effort should be directed first at changing reality and only then towards influencing others’ perception of it. If I like Karate, I could work towards promoting it in India and if I succeed in replacing the popularity of cricket with that of Karate, I may finally replace the stereotype with a new one that says “Indians love Karate”. There would still be many who are inconvenienced by it – perhaps even more than by the cricket stereotype, because being challenged to a dual is usually of more serious consequence than being invited to a game of cricket!

You might say that I have chosen a very convenient, unoffensive stereotype to prove my point, and that argument is not entirely without value. For example, if somebody was still ignorant enough to think of India as a land of snake charmers and were to begin a conversation on that subject with me, I might be offended – not at the stereotype itself as much as from the shock of being exposed to such an extreme level of dumbness. This is not to say that there is anything bad about being a snake charmer, but to accept the point that people whose perceptions are far from reality can be quite irritating. Again, this is a problem with specific individuals and should not influence our general opinion that stereotypes are okay and can be quite useful as long as they are not taken to be sacrosanct.

Finally, on a concluding note, a good learner needs to draw conclusions from whatever they know, as well as throw them away or build on top of them as he learns more. If you have some thoughts on this subject, feel free to share them below so that they serve to strengthen or weaken my convictions on the subject, either way helping to refine my understanding and that of other readers of this post.