Causes and Randomness

[Jonathan Siegel <jmsiegel@yahoo.com>]

I just wanted to mention that it's entirely possible
that when we look at events individually, each and
every specific failure may be attributable to an
identifiable cause. And yet, looking at things over a
long period of time, it may nonetheless be more
profitable to regard the system the individual
failures are part of as being random. 

For example, people ordinarily do not die out of the
blue. They get sick or involved in some sort of
accident or crime that creates an identifiable root
cause of death. And yet insurance companies, who look
at the overall system of mortality in a population, do
their work by treating deaths, taken as an aggregate,
as essentially random phenomena. 

Seeing events as random is far from helpful to people
immediately involved in a crisis, who need to make
medical decisions or are recently berieved. They need
to focus on the peculiarities of each individual case.
But seeing events as random can be quite helpful in
preventing deaths: looking at the overall system helps
identify behavior, like dieting, eliminating smoking,
wearing seat belts, and reducing other risk factors,
that can be quite helpful in lowering the probability
of death and thereby increasing the likelihood of
prolonging ones life. The two ways of seeing events --
as individual phenomena each wholly deterministic, and
as an essentially random aggregate, are both true: it
depends for what purpose one wishes to se things. 

The reason why both are true is that the whole is not
the sum of the parts. Randomness is only a property of
systems, never a property of individual events or even
any finite subset. Each individual event can be
wholely deterministic, and yet together they can be
(at least approximately) random. This systems concept
of randomness - this notion of randomness being a
property of systems and of systems not being the sum
of their parts -- was one of Shewhart's key
contributions to statistics.

Jonathan Siegel