Resistance to New Ideas (5)*

[David Kerridge <dfkerridge@mac.com>]

In 1931 Walter Shewhart wrote:

"We cannot do what we want to do, and cannot know why we
  cannot do it."

This deceptively simple statement is a challenge to centuries of
scientific and popular thinking. There may be limits on what we
can do, in practice, but limits on what we can know? From the
old logical-deductive view of science this is meaningless. Yet when
we understand operational definitions, what we can know and what
we can do must not be separated.

The important thing is not so much that Walter Shewhart said
it, but that he realised its fundamental importance, and went on
to investigate the nature of these limits.

It may not look at first sight like a law of nature, but that is
because the popular idea of a natural law is of a simple
mathematical rule. In fact the really fundamental laws of nature
are seldom like that. Karl Popper put it this way:

"Every testable theory can thus be put into the form 'such and such
 cannot happen.'"

Popper's work was not published in English while Shewhart was
active, but he was very familiar with "negative" laws like this.
That entropy cannot decrease is a good example. Heisenburg's
uncertainty principle, published at about the time of Shewhart's
first book is another - this too is concerned with limits to knowledge.

The new law, as stated by Shewhart, is in two parts. First that a
state of statistical control can exist. This is so familiar to us that
we forget how remarkable it is. Out of the chaos of a complex
system we produce a statistically predictable state. He goes on:

"This state of control appears to be, in general, a kind of limit
 to which we may expect to go economically in finding and
 removing causes of variability without changing a major
 portion of the manufacturing process...."

As a scientific law, this sounds very weak. "In general" and
"a major portion" are left undefined. But it is quite enough
to improve the system. To say "a limit exists" is interesting,
but not necessarily helpful. Nothing can travel faster than light,
but that is no use in designing a plane.

On the other hand,  to show how to know you have *reached*
the limit" is extremely practical. It means that to make progress
we must change what we are doing.

The apparent weakness of the statement just shows the exactness
of Shewhart's thinking. He was exact even to the point of drawing
attention to the points which need further clarification. He said
in the preface:

"This book is the natural outgrowth of an investigation started
 some six years ago to develop a scientific basis....

  .... As such this book constitutes a record of progress and an
 indication of the direction in which future developments may
 be expected to take place."

Six years is no time in the development of a new science.
Proposed laws must be thoroughly checked, not only against
observation, but against other related theory.

I feel sure that if Shewhart had stated his theory in terms of
models or explanation, more people would have understood
what the theory was about. But models are always to some
extent speculative. In pure science speculation is normal,
perhaps necessary. Unfortunately in applied science the danger
of publishing speculation is that people may act on it, not
understanding the difference between speculation and prediction.

Now, over 70 years later, we not only have ample experience
of applying Shewhart's theory, but other laws that connect
with it. Some, at least, of the "gaps" in Shewhart's theory can be
filled in. But there seems no way round the difficulty that the
operational-predictive model of science is necessary if we are
to see what it all means, and especially if we want to expand it.

--
Best wishes

David

dfkerridge@mac.com