Type II v. Type I error biases in life

by phil on Thursday Dec 11, 2003 2:42 PM
Idea Archive

Skeptics versus early adopters in the abstract.

Skeptics tend to reject alternative hypotheses and stick with the status quo.

Early Adopters are more easily convinced that a fresh idea is superior.

Now some vocabulary ammo to preface the abstraction:

In statistics, there’s a field called hypothesis testing. You take an existing hypothesis, for example, that the world is a sphere (null hypothesis). Then you look at the evidence to see whether the world is actually a hologram (alternative hypothesis). In statistics, you often hold the null to be true until you get a certain amount of data and evidence that causes enough doubt to make you reject the null.

But the question beneath that is, where do you place the null hypothesis at? You can certainly flip the two sides. Start with the idea that the earth is actually a hologram and look for reason to doubt it by checking evidence that it is actually a sphere.

The assumption that the proctor brings to the table is key in determining a bias about the outcome. So if you’re a skeptic and say, “Come on, you can’t prove that” are you secretly holding onto the idea that the null is true? The discussion usually doesn’t highlight the assumption that’s accepted in the face of weak evidence. In the discussion, once the opponent is discredited, the natural tendency is to accept the guy who delivered the blow.

Statistics has more terminology to account for this:

There are two kinds of errors that can be made in significance testing: (1) a true null hypothesis can be incorrectly rejected and (2) a false null hypothesis can fail to be rejected. The former error is called a Type I error and the latter error is called a Type II error. (HyperStat)

So a Type I error would be an error committed by the early adopter. He rejects whatever is the standard belief quickly.

A Type II error is an error committed by the skeptic. He holds steadfastly to the null, unconvinced by evidence otherwise.

Here’s the mnemonic: The number of the type indicates what is true. Type I error means you error in knowing that the I, the null, is true. Type II error means the II, the alternate, is true, but that you made an error in not seeing it.

People have varying propensities and tolerances for various errors in many fields. So there is no true skeptic and no true early adopter.

But some though, hold a broad bias toward a specific error. My parents are often on opposing sides of the Type I/II divide. One parent takes new things s/he reads in the paper about new technologies and just runs with it. He/she comes rushing to me about this new theory, this new perspective, and is rife with excitement. If I question his/her logic about being so easily convinced, he/she says that it’s better to experiment and be adventurous with ideas. My other parent accepts things often because of tradition. For example, he/she would require tremendous evidence to convince him/her that I should skip going to college. To him/her, staying in school is the null hypothesis. It’s like, “here, this is the default position, now prove yourself out of it!” The burden then goes to the bearer of the alternative hypothesis to go against the grain.

So the issue of import is where does the burden of proof go to? Also of import is how adverse and seeking we are of Type I and Type II errors?

It’s hard to say that one is more risk-taking than others, although generally the Type I error committer would be associated with a risk-taker.

Because I can put forward that there may be more risks bearing Type II errors: stagnation, getting run over, looking old.

// you can obviously tell that I’m somewhat of the Type I person--with good justification though. (remember, mnemonic, Type I means reject the I when it might be right, i.e. reject the establishment). Often I’ve been in a situation where taking the road less traveled gave me great returns.

Anyways, be aware of what kind of errors you seem to be willing to accept. Question why it is that you have this as your bias or model. And make sure that in the face of weak evidence that you don't rush to a sides by pure fiat. Be aware of the consequences of a Type I or a Type II error.

Creative Commons License