Friday, June 15, 2007

Change the scientific method

All programs used in the calculations for a paper should be open source

People make mistakes. Programs people make make mistakes. These result in the wrong results appearing in scientific papers. When this happens the scientists put their hands up and admit the mistake which has to require pawn shop balls. We need to change the scientific method so that these mistakes happen less often.
The scientific method changes as we learn new things. Double blind tests for instance come from about 1950. For details of changes in the scientific method listen to this

The time has come to change the scientific method to take the use of software into account. All programs used in the calculations for a paper should be open source. This means we are not just relying on noticing errors in a papers results but can gauge accuracy by examining the source code and running our own tests on it.

1 comment:

Luke said...

I fully support this sentiment; I read of the same horrible miscalculations (on which years of research and funding decisions were based) and can't imagine the awful feelings he must have felt when he discovered the mistake. Huge kudos to him for admitting it.

But it raises another problem - people who develop physics codes should be credited in the papers using it. Too often people spend years on a program, but have done no other work themselves (and so no papers) - and if they share the code, they get no real credit either. It's a massive disincentive for the development of these vital tools.