August 3, 2018

The Anna Karenina Principle

Leo Tolstoy's book Anna Karenina begins with the line:

Happy families are all alike; every unhappy family is unhappy in its own way.

This line has inspired the "Anna Karenina Principle" which states that a deficiency in any one of a number of factors can doom an endeavor to failure. As Wikipedia then continues, "consequently, a successful endeavor (subject to this principle) is one where every possible deficiency has been avoided." I believe that in many ways, this has perfectly described my ability to continually and consistently push out content for a personal blog.

For a number of years, I have tried on and off to write a blog about my life, or the going-on's that happen in it, yet consistently, I find that something keeps me back from doing it. Either I'm "too tired to write today" or "I don't have anything interesting going on in my life" or something else keeps me from becoming sufficiently motivated to sit down by my computer and type out a handful of words about how my week is going, or something shockingly cool which has happened.

Thus, my blog has been an endeavor doomed to failure - yet, I don't think it has to be that way. Thus, this week I'm going to take a look back at the Anna Karenina principle through history and talk a little bit about where it came from so that you might be able to use it in your daily lives.

To start to understand why I'm so interested in this, and why I'm going to talk about it, we need to understand how I learned about the principle in the first place. I came across this principle while reading a paper on Facial Expression Recognition for computers (which I can't mention here because at the time of writing it's still under review). The authors claimed that the reason that humans and robots fail to interact in meaningful ways is due to this principle - there's always something that is a bit off, thus, we are unable to achieve a happy interaction with the robot. This is, in a way, a similar idea to the Uncanny Valley, the principle of robotics that says "Until it works perfectly, it's just going to be creepy" (Paraphrased)

[From a Gizmodo article here, the image above is the classic example of the Uncanny Valley in our modern age, Grand Moff Tarkin from Rogue One]

Thus, perhaps we can consider the Uncanny Valley as a corollary of the Anna Karenina Principle - until we understand all of the things that we're doing wrong, and fix all of them, we won't be able to achieve a positive outcome.

The Anna Karenina Principle can be traced back all the way to Aristotle, in his book, "Nicomachean Ethics" (Book 2, for those who are interested). He writes:

Again, it is possible to fail in many ways (for evil belongs to the class of the unlimited, as the Pythagoreans conjectured, and good to that of the limited), while to succeed is possible only in one way (for which reason also one is easy and the other difficult – to miss the mark easy, to hit it difficult); for these reasons also, then, excess and defect are characteristic of vice, and the mean of virtue; For men are good in but one way, but bad in many.

Of course, it was not Aristotle who gave the principle its popularity. That would be Jared Diamond, and his book "Guns, Germs and Steel" - a book that has been on my reading list forever, and I haven't had the opportunity to get around to reading it. He uses this principle (according to Wikipedia) to discuss why so few wild animals have been successfully domesticated. He argues that successful domestications come, not from one or two positive traits, but from a lack of negative traits. (Again, you shouldn't take my word for this, and read his book, which is arguably much more interesting than my own musings, so maybe just put this down now and go read it).

In a 2001 paper, psychologist Roy Baumeister and his group concluded that "bad is stronger than good", meaning that bad emotions, bad parents, and bad feedback have more impact than their good counterparts [1] (OMG. A reference? in a blog post? Oh no!). This is interesting to me because it gives weight to this theory in a way - we find that a single negative trait can overwhelm any of the good traits that we can perceive. The question then, is why is that the case? Why is it that negative reinforcement is so much more powerful than positive reinforcement?

One reason that might be the case (pure speculation here) is that throughout the evolution of human history, a negative detractor is much more closely linked with death, thus, the reward signal for something bad is much stronger than the reward signal linked to something good. It is essential not to die, but living "better" isn't as strong a priority. This can be reflected in what economists have found with respect to the relationship between income and happiness, they find that happiness increases up to a threshold (About 65-95K annually), and then stays relatively constant or even decreases ( Thus, there may be some truth in getting stuck in the "local minima of life", that is, once we achieve a result we're happy enough with, then we cease trying as hard to improve.

But this discussion of happiness has lead us to digress from the main point, the evolution of the Anna Karenina principle. We've seen how one bad thing can truly offset the good - and thus, argued that a combination of merely good things (and not doing one thing particularly bad) can lead to success. This is not, however, a widely agreed upon principle. In his book "Zero to One" Peter Thiel argued that business operates in the opposite way to such a principle:

Business is the opposite. All happy companies are different: each one earns a monopoly by solving a unique problem. All failed companies are the same: they failed to escape competition.

While this is an interesting thought, I don't think that this truly claims that business is opposite - as he would like to say. While the market for each company may be different, the strategy to success is the same, in essence, "Being Different" becoming a mantra, leads to companies that are generally the same, and operate in similar ways.

Yet, there is also some truth to the matter. It is clear that no two people are the same, and to be happy, you don't have to be the same. That is, where I believe, that the principle falls short. In a way, we are all the same, we have all followed our own recipe for happiness. But that doesn't make everyone the same - we may be unique, and happy, but when we are unhappy, it is often because of just the slightest imbalance, even just one thing falling through the cracks can shatter the happiness that we have.

So, in the end, what does this mean for robotics? What does this mean for this blog? For robotics, I think it means that we have to focus on the failure points of our unified systems, and until we create unified systems, we won't know which parts are failing. This is why I respect the Rogue One directors for their choice of CGI, by showing people what they could do, they received the invaluable feedback they can use to improve their system. For this blog, it means that I'm going to try to work a bit harder to push out content in the future. I can't promise that they'll all be as engaging as this one, or as long, but I hope that in the end, at least a few of them will be worth reading.

If you want to read more on the Anna Karenina Principle and it's relation to science, try this:

[1] Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001). Bad is stronger than good. Review of General Psychology, 5, 323-370.