Next: From the probability of
Up: A probabilistic theory of
Previous: Learning from observations: the
Contents
People very often think that the only
scientific method valid in physics is that of Popper's
falsification scheme.
There is no doubt that, if a theory is not capable
of explaining experimental results, it should be rejected or modified.
But, since it is impossible to demonstrate with certainty
that a theory is true, it becomes impossible
to decide among
the infinite number of hypotheses which have not been falsified.
This would produce stagnation in research.
A probabilistic method allows, instead, for a scale
of credibility to be
provided for classifying all hypotheses taken into account
(or credibility ratios between any pair of hypotheses).
This is close to the natural development of science, where new
investigations are made in the direction which seems the
most credible, according to the state of knowledge at the moment
at which the decision on how to proceed was made.
As far as the results of measurements are concerned,
the falsification scheme is absolutely unsuitable.
Taking it literally, one should be authorized only to check
whether or not the value read on an instrument is compatible
with a true value, nothing more.
It is understandable then that, with this premise,
one cannot go very far.
We will show that falsification is just a subcase of
the Bayesian inference.
Next: From the probability of
Up: A probabilistic theory of
Previous: Learning from observations: the
Contents
Giulio D'Agostini
2003-05-15