Defending ‘Theory’ and Rationalism

Lib Nicholas Nassim Taleb takes another shot at Dawkins.

As for the passage itself, there’s room for both theory and empiricism, as I explain in an earlier article Falsifiability, String Theory, and Policy:

1. Algorithms save time, versus having to do everything individually. Whether it’s movies or video games, computer simulations have become so realistic as to be almost indistinguishable from reality. Although some problems such as turbulence remain unresolved, algorithms do a reasonably job of approximating reality. Why waste time with heuristics if an algorithm is 99% accurate and can be done instantly?

2. Scientists who create theoretical models understand the limitations of their models, and also have ways of measuring the plausibility of said models based on various preconditions without having to explicitly test their theories. Many physicals laws, for example, must obey conservation of energy, must have various invariance properties, and not ‘blow up’ under various limits and extremes and give reasonable answers otherwise. All of this can be verified mathematically without experiment. Once a model passes the ‘smell check’, tests can be run on it. However, as in the case of theoretical physics, constructing appropriate models is hard enough, let alone testing them.

3. Theory can proceed verification, not the other way around. General and special relativity were later verified with tests. Same for Maxwell’s equations, which predicted the speed of light, and later verified.

4. Scientists are aware that trajectory models cannot predict whether someone will throw a baseball or why he would want to throw it, yet Taleb assumes that scientists are unable to make this categorical distinction between a closed-system and an open one.

Liberals like Taleb want to promote ‘leveling’ – the belief (also shared by other pop psychology charlatans) that humans (especially smart ones) are irrational. ‘If smart people occasionally make mistakes, we can’t trust smart people to make decisions’ To the left, the ideal state of man is irrational, egalitarian, and primitive, not civilized.

Unless someone is clinically insane, everyone acts rationally in their own best-interests (homo economicus), but IQ determines outcomes, with smarter people having more fortuitous rationalizations. A stupid person may rationalize buying lottery tickets, for example, which it’s why it’s appropriately called the ‘stupid tax’. In the strictly mathematical sense, playing the lottery is irrational (due to the negative expected value for the participant), but less intelligent people rationalize playing it. These people act irrationally due to the availability of a priori knowledge (such as the odds of wining the lottery, which are available to the public), but other instances are not so obvious such as whether or not the stock market is in a bubble, in which case there is much more ambiguity. To settle this, in Defending Rational Markets, I argue there are two types of rationality: ‘real time’ rationality (based on what is happening now) and post-hoc/a posteriori rationalism (after the fact). For example, if someone observes that a stock has risen linearly from $40 to $60 in twenty minutes, would it not be so irrational for him to assume it will rise to $61 in the 21st minute? So he buys. But then the SEC halts trading of the stock and fraud is revealed. He loses everything. In retrospect he acted irrationally.

But if there’s ambiguity, doesn’t that invalidate certain models, particularly financial models? Not necessarily. The theory of Rational Expectations implies that even if everyone is uncertain, the aggregate of guesses will lead the ‘best’ answer, or in other words, the best guess for tomorrow it today’s forecast. For example, in predicting GDP growth, some economists will predict a number that is too high; others, too low. Since financial markets involve thousands of agents (stock brokers, hedge funds, day traders), the ‘committee’ is very big, meaning that it’s unlikely that everyone will be wrong. Regarding #1, financial models (such as Brownian Motion and jump diffusion) based on Rational Expectations do a reasonably good job approximating real-life financial markets. Even ‘fat tails’ and most ‘black swans’ can be accounted for if the models are sophistical enough. For the market to be irrational would imply that there is some sort of arbitrage. Option pricing formulas forbid arbitrage, and in real-life there is no arbitrage observed in option pricing, in agreement with the underlying theory.

Related:

Irrational Investors Don’t Exist
Defending the EMH