So we might point to conjunction fallacies, to overconfidence, to the availability and representativeness heuristics, to base rate neglect, and say: “Stuff like that.”

Biases are defined as obstacles that are not related to limiting computing power or the cost of the information, but the shape of the machinery itself. Evolutionarily our mental machinery, was made for a much different setting to optimize for other things, unlike thinking critically. I guess I can say that it was optimized for a condition where perceptual judgement/System 1 expected to be was leading than deliberate judgement.

For example, the machinery is evolutionarily optimized to purposes that actively oppose epistemic accuracy; for example, the machinery to win arguments in adaptive political contexts.

I take it from the above that we evolved with characteristics that somehow allow us to win arguments in adaptive political contexts. I say this out loud just in case. I can’t imagine that evolutionary there was a need to ‘win arguments’ in adaptive political contexts, as I take it that language was non-existent and we were just like animals. And indeed these oppose accuracy. It’s like truth doesn’t matter at all. I don’t understand epistemic accuracy yet (OI1).

Or the selection pressure ran skew to epistemic accuracy; for example, believing what others believe, to get along socially.

Or the classic heuristics and biases. The machinery operates by an identifiable algorithm that does some useful work which produces systematic errors. Systematic errors can be easily imagined when one talks about a faulty instrument.

For example, a voltmeter might show a reading of 1 volt even when it is disconnected from any electromagnetic influence. This means the systematic error is 1 volt and all measurements shown by this voltmeter will be a volt higher than the true value.

So we thus define biases as obstacles to our truth finding that are not a result of not having costly info or not having enough computational ability. Biases are a result of evolution that oppose/are skewed to, epistemic accuracy, and are inherently present in the shape of our machinery itself. They produce systematic identifiable errors, which are obtained by either heavy thinking or experimentation.

Can we look at the empirical cluster and find a compact test for membership?

What we have done so far should be enough to get on with, I guess, as Eliezer points out.

“Biases” are distinguished from errors that arise from cognitive content, such as adopted beliefs, or adopted moral duties.

I guess these are the ones such as the belief in god that has been fed to you from young age. It is not a bias that you believe in such a thing but it’s more of a mistake. Though the source of the mistake could ultimately be a bias.

Biases are distinguished from errors that arise from absorbed cultural mores. Biases arise from machinery that is humanly universal

I guess these are the ones such as handshaking, or bending to greet people. It is an important criterion that the obstacle arises from the shape of the mental machinery for it to be a bias.

Plato was ignorant of General Relativity. There was no way he could obtain that information. Plato is not biased because his ignorance of general relativity did not arise due to the shape of his mental machinery.

As said before, we are in the truth finding business here. Definitions are not our primary goal. The example of describing a fire was a nice example by Eliezer to demonstrate that definitions are only a means and not the end. Overall, we see ‘biases’, as obstacles to our truth finding. They are a result of the way the existing evolved machinery.

Biases arise from machinery that is humanly universal

Open Issues

1) What is epistemic accuracy?

P.S The day before I wrote about bias based on Eliezer’s blog, this is a continuation.