Tuesday 5 October 2010

Operationalisation of measurement - the keystone of modernity

Modernity is apparently about increasing efficiency, mainly by means of new technologies involving the complex organization of more specialized functions.

(So the classic example of Adam Smith's pin factory involved replacing one pin maker who performed dozens of functions with a complex organization of dozens of pin makers each performing a single specialized function).


But measuring efficiency entails creating an operational definition of output - something which can be measured. So that to measure the output of a pin factory requires that a 'pin' is defined in operational terms, so that pins can be counted (and the input required to generate 'n' pins can be calculated).

Something similar applies in medicine - if we are looking for progress in medical treatment, for example in curing pneumonia, then progress can only be measured by operationally defining 'pneumonia', and also 'cure' - such that improved outcomes could be detected.


Operational definitions of diseases, or cures - or inputs or outcomes in general - can be more or less valid, and more or less precise.

In the Soviet Union it was commonplace to 'improve efficiency' by using invalid operational definitions and adjusting them over time. For example, a metal pipe was defined in terms of a tube with a hole in it and the output was measured according to weight - therefore efficiency was 'improved' by making ever-heavier pipes (until and beyond the point where they were  unusable as pipes).

In medicine, a skin lesion is defined as a 'malignant melanoma' in such a way that more and more benign skin lesions are brought within the category ('just in case'), and the death rate among those with 'malignant melanomas' falls. Or breast cancer is diagnosed earlier and earlier, and with less and less certainty that it is a cancer, and the death rate of people with 'breast cancer' goes down.


But these are specific example of a general phenomenon - all operational definitions are incomplete and biased - and there is always the possibility that apparent improvement is merely a by product of the changing nature of measurement; and that the overall situation may be worsening even as all measured proxies are improving.

This is a flaw right at the heart of modernity.

And the flaw is amplified when, as now, people realize that decline can be concealed by invalid operational proxies, or by changing thresholds of operational definitions, or ultimately by *not* measuring that which declines (or which is suspected of declining).


So, the calls for more 'evidence' or more 'objective' evidence. or more 'systematic' evidence' or more statistically-significant data can be and are used to manipulate the official picture.  

The only safeguard against this is that bottom line evaluations be individual, direct, personal and common-sensical.

Can individual common sense be wrong? Of course it can be and often is!

But the fact is that we have nothing better - and the worst error is to believe that we have.


dearieme said...

As I am wont to say, in Social Science "data" is the plural of "bollock".

xlbrl said...

Truths are revealed by individuals, but individuals find safety in numbers.