< Back to Volume 8, Number 4
Volume 8: Number 4: Article 3
Applied Scientific Inference
P. A. Sturrock, Center for Space Science and Astrophysics, ERL 306,
Stanford University, Stanford, CA 94305-4055
The purpose of this article is to use the principles of scientific
inference to provide guidance in evaluating complicated issues such
as those raised by the study of anomalous phenomena. Specifically, the
article presents a formalism (a "protocol") for organizing and combining
the many judgments that must be made in the scientific evaluation of
the relevant hypotheses. All judgments are to be expressed as probabilities,
and the rules for combining probabilities are derived from Bayes' theorem.
Setting up a problem in a manner that permits such an analysis can be
helpful in imposing a structure and discipline upon the analysis, and
also in exposing relevant questions that might otherwise have remained
hidden. Furthermore, the introduction of probabilities makes it possible
to put on a sound numerical basis such assertions as "extraordinary
claims require extraordinary evidence." One finds that extraordinary
evidence can be built up from many (but not very many) items of unspectacular
evidence, provided the items are truly independent.
The proposed procedure permits a clear separation between a statement
of initial prejudice and an evaluation of the significance of considered
evidence. However, it would be even better to set out explicitly the
considerations on which the prejudice is based, and to view those considerations
as part of the evidence to be evaluated. The procedure also draws a
clear separation between the roles and judgments of data analysts (who
assign probabilities to specified statements, based on the evidence),
and those of theorists (who assign probabilities to the same statements,
based in turn on the considered hypotheses).
In order to reach a consensus on any topic, it is recommended that
probability estimates be made by teams of experts, all team-members
being presented with the same data but acting independently, and procedures
are proposed by which individual estimates may be combined to yield
a consensus estimate.
"Formality has its place in guiding one along suitable paths of argument:
and most of us need some guidance." (D. V. Lindley: see Lindley, Tversky
and Brown 1979, p. 177)
"Talking of a Court-martial that was sitting upon a very momentous
pubic occasion, he expressed much doubt of an enlightened decision;
and said, that perhaps there was not a member of it, who in the whole
course of his life, had ever spent an hour by himself in balancing probabilities."
(Samuel Johnson, 1780)
To purchase back issues contact Allen Marketing & Management: 1-800-627-0629
