Tuesday, April 17, 2012

Mailvox: considering self-correction

Azimus is interested in the possibility that science is not, in fact, self-correcting.
Experimental replication, in the very rare instances it is actually performed and is successful, is nothing more than auditing. There is no substantial difference between one scientist re-running another scientist’s experiment and one accountant re-calculating another accountant’s books. In other words, science isn’t self-correcting in any meaningful sense even in its ideal form."

Interesting thought. Tilting a little in the direction of a "let's have a definition war" argument, but an interesting thought. By that yardstick would you call the market, or engineering self-correcting?
Very well, we can certainly do this the methodical way. Rather than risk a definition war, I will first ask Azimus for his definition of "self-correcting" before I answer his question about the market or engineering being self-correcting. I'm not avoiding his question, it's only that as I've pointed out before, depending upon how one defines "self-correcting", science is either NOT self-correcting or else it is TRIVIALLY self-correcting in the same manner that practically every human activity is.

To which Azimus responded:
As I read your post, it struck me that the definition of "self" is scaleable. In your accountant example, accountant #1 may not be self correcting, but if accountant #2 audits #1 as part of a departmental auditing system, the accounting department is "self correcting." In the same way an engineering firm has a green-horn doing most of the design work, which is then reviewed by a 5yr+ experienced PE who examines the work and makes corrections. The greenhorn is not self correcting, but scaling the word "self" to be the engineering firm, would the firm not be "self correcting"?

A marksman firing at a target makes allowances for distance, elevation distances, windspeed, etc. His first shot misses. He interprets the fall of the round and hypothesizes the wind was stronger than he allowed for and he adjusts accordingly in the second shot hitting the target. Is this self correcting?

Since there is no argument on the word "correctiong", The battle line seems to be drawn along the word "self". I see it as scaleable and will define the term thus:

Self correcting: an entity is self correcting if it contains a mechanism by which error is identifed and eliminated.
Very good. So, Azimus has chosen the option by which we must ultimately conclude that science is TRIVIALLY self-correcting. He is correct, and in his examples given, the auditing department, the engineering firm, and the marksman would all be considered self-correcting.

But from both his definition and his examples follow three obvious questions. They are:

1. What is the entity of science?
2. If there is no successful replication of a scientific experiment, and therefore no self-correction, is the experiment still science?
3. Since scientific reliability and authority claim is based on its self-correcting mechanism, how is science any more reliable than any other entity that possesses its own mechanism for self-correction?

I'm sure we shall all await his answers to those three questions with interest. In the meantime, I owe him a direct answer to his previous question: yes, the market and the engineering discipline are both self-correcting by his definition provided. The market self-corrects incorrect corporate valuations. Engineering self-corrects technologies that do not work and structures that do not stand.

Labels: ,


Post a Comment

Rules of the blog
Please do not comment as "Anonymous". Comments by "Anonymous" will be spammed.

<< Home

Newer Posts Older Posts