Post date: May 11, 2017 2:3:5 PM
Scott wrote to Bill Oberkampf:
We've just started a collaboration about "humane algorithms" at https://sites.google.com/site/humanealgorithms/ that I think might interest you. The phrase is after William Kahan whom we met at the NIST meeting in Boulder. He's supposed to be the "father of floating point" or some such thing, and he said that the recent software disasters show that we need better software. I know that this is your own sentiment too. I'll make you an "editor" of the page so you can make comments and contributions to the collaboration yourself. No pressure, of course. But if something strikes a chord, feel free to vent!
Bill responded:
Thanks for the note about "Humane Algorithms". I certainly agree that more emphasis needs to be placed on robust software, as well as reducing the vagueness and ambiguity of interpretation of results. I think the primary root of the problem is poor specification of both software and simulation requirements. If you combine that with the (never ending) pressure of business to quote a low price on the software and simulation, then you have a recipe for frequent failures.
I glanced through the text on the "Humane Algorithms" website. I think that the endeavor is very worthwhile and I wish you and your colleagues success. To my taste, however, the text had too much flavor of "fairness, friendliness, and diversity" for my taste. I believe you can describe the technical requirements for software and simulations without resorting to "human traits" that are of such value to the PC crowd.