[SystemSafety] Practical Statistical Evaluation of Critical Software

Peter Bernard Ladkin ladkin at rvs.uni-bielefeld.de
Mon Mar 2 15:17:29 CET 2015


On 2015-03-02 14:15 , Derek M Jones wrote:
> If there is little or no empirical data to test theories against then
> theorizing should be flagged as speculative at best and treated as such,
> or as philosophy.

Another simple but mistaken metaphysical view.

Do you test whether 2+2=4? How? What do you do when you find 2+2=5? Do you reject your count as
mistaken, or do you announce to people you've discovered arithmetic is wrong? Why do you do
whichever of those you do?

Arithmetic is surprisingly resistant to "empirical data". When the person at the check-out gets my
sum wrong, then heshe does it again until it's "right". Do you call arithmetic "speculative at best
and treat.. it as such, or as philosophy"?

My guess is that you don't. You just use it, whatever "using it" may consist in, and if the
"empirical data" doesn't fit your preconceptions at the cash register, just like I do you ask the
check-out person to redo the arithmetic until it does.

I wonder if you can give a coherent account of that practice, which according to you is classified
as "speculative at best and treated as such"? I doubt whether you really treat arithmetic as
speculative. Nobody does. So much for the coherence of your announced principle above.

The unexamined practice is not necessarily trustworthy. But monkey-do with respect to arithmetic
does seem to yield people passable results.

I once had a discussion with a renowned critical-systems person who maintained that Popper's
criterion of falsifiability was "obviously" the right criterion to distinguish science from other
enterprises. Except that 99% of what any scientist does - especially empirical scientists - is math.
And math is not falsifiable except in the case when it is wrong. So apparently math isn't "science",
despite it being most of what many scientists do, especially the arithmetic part.

I say this to people who study what scientists do and they say "of course, we've known this for
decades". I say it to said critical-systems person and he goes ballistic and leaves the discussion.

> Once we have data we can start doing engineering or science.

Newton and Einstein were obviously both capable of doing enormous amounts of influential science
without touching "data". Indeed, Einstein said that it was the data that needed to fit his theory,
rather than the other way round. He was right. And it did.

> I have found comical remarks (or flippant, depending on your point of
> view) a useful way of curtailing speculative theorizing that is taking
> itself too seriously or philosophical expositions.

If you think "theorising that is taking itself too seriously" is a meaningful phrase, then I have
some colorless green ideas to sell you.

PBL

Prof. Peter Bernard Ladkin, Faculty of Technology, University of Bielefeld, 33594 Bielefeld, Germany
Je suis Charlie
Tel+msg +49 (0)521 880 7319  www.rvs.uni-bielefeld.de






More information about the systemsafety mailing list