FOM: Book Reference

Dean Buckner Dean.Buckner at
Mon Apr 22 13:51:46 EDT 2002

I have found the reference to the book on Risk Compensation theory that I
mentioned in earlier posting, namely *Risk* by John Adams, 1995.  It may of
course be *completely* unconnected with Shipman's problem, but Adams' book
is worth reading anyway, full of subtle and interesting observations and
clever examples.

Also worth reading is a shorter paper, URL here

For example:

"Britain's chief medical officer of health (Sir Kenneth Calman) says "It is
possible for new research and knowledge to change the level of risk,
reducing it or increasing it.  This view sits uncomfortably alongside the
Royal Society's view of risk as something "actual" and capable of "objective
measurement".  If risk is "actual" and subject to "objective measurement",
how will further research modify it?

"This phenomenon might be described as the Heisenberg problem.  The purpose
of measuring risk is to provide information that can guide behaviour.
Statements about risk are statements about the future.  Accident statistics,
the most commonly used measures of risk, are statements about the past.  To
the extent that the information conveyed by accident statistics is acted
upon, the future will be different from the past.  The act of measurement
alters that which is being measured.

... What Calman perhaps meant when he said that new research might change
the level of risk is that the probabilities intended to convey the magnitude
of the scientist's uncertainty are themselves uncertain in ways that cannot
be expressed as probabilities.  He should have perhaps have said that a
scientific risk estimate is the scientist's "best guess at the time but
subject to change in ways that cannot be predicted"."


It's a long time since I watched Star Trek, but remember being v. impressed
but also puzzled by the way that Mr. Spock would invariably predict the
chance of success as some many digit number to one.  Puzzled, because the
Enterprise crew would, against these usually tremendous odds, invariably
succeed!  Were they just lucky I wondered?  Or was Spock just very bad at
estimating probability?  And if he was so bad, how could he be so accurate
about the probability figure itself?

"If a guy tells me the risk is 1 in 10^5 then I know he's talking
c--p" (Feynmann)

The idea of "Risk Compensation" comes out of research by  a University of
Chicago economist, Dr. Sam Peltzman, in the 1950's.


Dean Buckner
4 Spencer Walk
London, SW15 1PL

Work 020 7676 1750
Home 020 8788 4273

More information about the FOM mailing list