Wednesday, July 20, 2011

Balony Detection Kits

Check out written versions of:

Lets think about this for a bit. It seems to me that these guys are thinking about big questions -- "is mankind causing global warming" or "is evolution the best explanation of the web of life". Does the approach work for all questions of what you should believe? How about really little questions -- should I take an umbrella with me today?

I have a little box that predicts the weather, a daily newspaper that does so, and several weather channels on television. I could spend a lot of time making my own observations, but it seems likely that I should accept the authority of these sources. I could spend a lot of time comparing their predictions over history with what actually happened. But the cost of making the decision should not be greater than the penalty for making the wrong decision.

It also seems to me that belief is a continuous variable, not a binary one. I would rather say, it is probably true that it is going to rain rather than have to say "I believe it must rain" or "I believe it can not rain". Better yet is the weather service's probability that it will rain. (Note, however, that you have to know what that probability means.)

So, let me think of a kit for making little decisions about what to believe.

1. How much is it worth doing to decide what to believe? There are distinctions without a difference. There are also distinctions that are either so unlikely to have costs or to have major costs that only a small amount of time and effort are justified in their research.

2. Refer to authority. Don't reinvent the wheel, look up the answer or ask someone who knows. On the other hand, be sure that the authority you source is based on knowledge, not some extraneous source such as access to the media or political office. Check to see that the source is really credible.

3. Break problems into their component parts, and plan an attack to find the best answer expeditiously.

4. Don't investigate extraneous issues. Physicians use the guidance that they should not order a test if the treatment decision will not depend on the results of the test; you too should limit your search for answers to those answers that will influence future actions.

5. Use your eyes and ears. If its raining, you don't need to check the weather reports to know whether to carry an umbrella; just look out a window.

6. Use your reason. Does the answer fit with other things you know? Does it make sense.

7. Investigate in ways that help you think. If you are number oriented, quantify. If you are oriented toward graphs, plot information. If you think in words, do so.

8 Be willing to accept ignorance? It is usually better to know that you don't know than to assume you know something which is not true.

9. Recognize degrees of credence. Ideally, one might say that a proposition is true with probability xx percent. Lofti Zadeh developed a fuzzy logic to recognize that, in contrast with traditional logic, it might be useful to consider that a proposition could be more or less true, more or less false, or more or less of undetermined truth value. Bayesian statistics provides not only a statistical means of adjusting probability estimates with new information, but a conceptual framework for doing so heuristically.

No comments: