Sometimes you gotta let "theory" trump "fact". Example: in modern weather forecasting, before actual observations can be put into numerical models and used to make predictions, the measurements have to be adjusted, smoothed, and corrected until they're consistent with the equations. Otherwise, tiny errors result in ridiculous physical instabilities --- artificial sonic booms that propagate around the globe, or abrupt discontinuities in other parameters. Even if perfect readings of humidity, temperature, pressure, etc. were somehow possible, they couldn't be used in models. Observations have to be bent to fit the discrete computational process.

When a theory is good enough, in other words, for some purposes it takes over reality. We know that the theoretical situation is "wrong" in one sense --- but it's so convenient that we prefer to live with the error. Maybe it's like our system of justice. In order to have clean, consistent rules, we accept less-than-perfect fairness ... and then add further escape-clause inconsistencies (e.g., executive pardons) to break out of the troubles that result.

(See Lewis Richardson's 1922 vision of a meteorological Forecast Factory, , and ^zhurnal 12 October 1999, LearningInconsistency, or 6 October 1999, MereAnarchy.)

Friday, June 08, 2001 at 06:00:23 (EDT) = 2001-06-08

TopicScience - TopicPhilosophy

(correlates: NumberNineDream, Red Patch Now, WeatherMan, ...)