Monday, November 30, 2009

Estimating empathy

This post has been on the "Set of possible upcoming posts" for over an year now. I have been wanting to write this post since last election campaign. I just had become a big fan of Obama and his stress for inclusive politics to improve the way DC operates the US (and the world :). Then I joined Waisman center whose mission coincidentally focuses on foundations of life, developmental disorders and Neurodegenerative disorders. A few goals essentially are about "understanding" psychological disorders and especially empathy deficits that can produce dysfunctional families!! Obama had mentioned about empathy deficits in Washington while people at Waisman focus more on a social, personal level. Obviously there's empathy involved at many levels of human life.

The main reason I wanted to write about empathy was that although mentions empathy many times in his speeches and which underlies his inclusive attitude, rarely did he say how others can acquire that skill. If you think about it you quickly realize that empathy is the most basic psychological quality (instinct) that creates any value in the society! Think about stock markets, music, movies, any art for that matter, even intelligence, quest for survival etc. So it makes sense to attempt to quantify such an important quality keeping such a scale of human civilization functioning so that there are guidelines for people to try to acquire such ability in an effective way. As Scott always says the meat of non-trivial reasoning is typically in quantifying, since it captures the complexity of the task without taking about trivial (im)possibilities in the rationale-extremes.

So what tools can we rely on! My favorite is obviously computational. As I mentioned before too the notion of computation is really unifying many spheres of knowledge since it tries to model the though process of human beings that underlies every sphere of knowledge. Its impact on all fields is so insuppressibly real that if there is no impact of computer science on a field then its realisticness can be questionable! Studying emotions like empathy and qualities like intelligence are a tricky thing. But as long as we have end goals for these computational thinking can help ask real questions. A few successful examples are the quest for artificial intelligence and computational game theory which ties up computer science and economics which helps design good societal games to keep the society stimulated in a healthy way. Hence for psychologists to ask truly relevant questions about empathy or other types of emotions its important to be able to take the machinist approach.

For e.g. we can rely on tools from interactive proofs (IP) where we can convince others of the truth of the statements using some communication protocols. The IP class is very powerful that means there are lot of things can in theory be communicated with others in effect creating empathy. Of course finding the protocols itself might be exponentially hard based on what we want to communicate. Usually experience in survival (survival instinct) seems to be the closest causal reason for intelligence which can be efficiently verified (polynomial time verifiability). Isomorphisms across spheres of knowledge and across time actually indicates that and actually realizing those isomorphisms actually fascinates me and indicates robustness of our human survival instincts.

The key to pick on survival instinct as the cause is the assumption (axiom) that everyone wants to survive. But how do you communicate with those who don't view this axiom like suicide bombers etc.? Well that's why we need some sort of enforcement of axioms otherwise there can be no basis for reasoning. Usually the best way to enforce axioms is to show the benefits of having those axioms like let's say proving non-trivial theorems (consequences) of those axioms and hoping that at least one of the consequences impresses them to fall for the axioms. For e.g. people fall for America for various reasons, like money, liberty, luxuries etc. etc. Empathy can only be verified after achieving it. Actual way of achieving is like coming up with theorems. So estimation of empathy is equivalent to theorem proving which means it's NP-hard! We can only hope to achieve practical approximations assuming P!=NP.