Linked by Thom Holwerda on Sun 2nd Jul 2006 21:26 UTC
People say I rant too much. I complain and complain, but never seem to really like anything. As I promised a few weeks ago, I will talk about things I love about computers. After explaining why I like to complain and rant, this column will solely deal with fluffy bunnies, green meadows, blue skies, and shiny, happy people. I promise.

Member since:
2005-08-22

>Where, precisely, is the bias in that?

All right, fair enough. We're getting off topic, but I'll still answer the question since the thread is dying anyway.

There are many biases here. Why are you asked to judge? Is it for a game of dice, or is it for scientific explanations of probability. If it is for a game, is it for tons of money, or just for fun? You can keep going down the line of questions until you become very precise. My bias is that you are being too vague, because in MY past, I've answered vague questions, and gotten in trouble. I never like to answer questions that are vague. Others may not mind.

Then you come up with a "statistical test." #1 comes up on the dice 16.1% of the time. Is it rigged? #1 should come up 16.666666% of the time. Is it close enough? Depends upon the underlying bias of the examiner. A statistician may say no. I may say yes, depending upon the situation. Did we roll the dice enough times? Maybe it was just an anomaly. What does "failed test" mean? Is 15.2% failure? If so, what about 15.3%? Where is the line? Is there a margin of error? If so, what is it? All of this will depend upon your personal beliefs about statistics and the potential variability. i.e. your preexisting biases.

Additionally, I might disagree with the test itself because of my personal biases. It's Monday, July 3rd. Dice should never be rolled on July 3rd, because of Gamblor, god of gambling, alters the odds. Sounds stupid? It's more reasonable to me than some religions. That's just my bias.

You may be thinking that I'm just being stupid and petty, but that's again a bias by itself. Biases vary as much as opinions. Your question implies that you trust statistics. I don't. I usually distrust the methodologies of data collection and analysis, and therefore, the results.

Finally, ask yourself the question, why do you have to judge to begin with? That implies there was disagreement ahead of time. Why is there disagreement? Why does anyone disagree with anything? I'll give you a hint; it has something to do with bias.

My point is that you shouldn't deny biases. Reveal them, and speak about them within the context of the situation. Then you and your audience can find a common ground to agree or disagree. If you deny any bias, then you can never find common ground with anyone, since you don't know what it is yourself.

Edited 2006-07-04 01:28

Member since:
2006-02-15

None of the biases you describe apply to my action of applying the test to the device though.

In science, this behavior is called objectivity. If the result of an observation is independent of the observer, then the observation is objective. Here is a sequence: "HTTHTH". There are many questions I can ask about that sequence that have objective answers:

which characters appear between the quotation marks? How many characters are there between the quotation marks? How many are 'H'? How many are not 'H'?

There are unbiased answers to all of those questions: 6, 2, 3, and 3.

In this particular instance, I can ask questions about the sequence that do not have objective answers: Was it generated at random? Was it meant to represent a series of coin tosses? Does it, in fact, record such a series?

These questions are not objective because I was alone when the sequence was generated and recorded, so I alone know the answers. They do have unbiased answers (yes, no, no) but you cannot verify that the answers are unbiased, or even if they are accurate.

I can even ask questions for which there are no unbiased answers: Why 6 characters instead of 7 or 5? Why did you pick 'H' and 'T'? Why only two characters?

The key here is measurable empiricism. If I can measure it against an unambiguous agreed upon standard, then the measurement is objective and without bias and if you apply the same unambiguous agreed upon standard, you will achieve the same measurement.

The opposite of objective, of course, is subjective. Any action which is objective is unbiased. Subjective actions may or may not be unbiased: measurably empirical actions are.