Thursday 27 May 2010

Micro-specialization and the infinite perpetuation of error

Scientific specialization is generally supposed to benefit the precision and validity of knowledge within specializations, but at the cost of these specializations becoming more narrow, and loss of integration between specializations.

In other words, as specialization proceeds, people supposedly know more and more about less and less - the benefit being presumed to be more knowledge in each domain, the cost that nobody has a general understanding.

However, I think the supposed benefit is actually not true. People do not really know more – often they know nothing at all or everything they know is wrong because undercut by fundamental errors.

Probably the benefits of specialization really do apply to the early stages of gross specialization such as the increase of scientific career differentiation in the early 20th century - the era when there was a division of university science degrees into Physics, Chemistry and Biology - then later a further modest subdivision of each of these into two or three.

But since the 1960s scientific specialization has now gone far beyond this point, and the process is now almost wholly disadvantageous. We are now in an era of micro-specialization, with dozens of subdivisions within sciences.

Part of this is simply the low average and peak level of ability, motivation and honesty in most branches of modern science. The number of scientists has increased by more than an order of magnitude – clearly this has an effect. Scientific training and conditions have become prolonged and dull and collectivist – deterring creative and self-motivated people. And these have happened in an era when the smartest kids tended not to gravitate to science, as they did in the early 20th century, but instead to professions such as medicine and law.

However there is a more basic and insoluble problem about micro-specialization. This is that micro-specialization is about micro-validation – which can neither detect nor correct gross errors in its basic suppositions.

In my experience, this is the case for many scientific specialties:

1. Epidemiologists are fixated on statistical issues and cannot detect major errors in their presuppositions because they do not regard individual patient data as valid nor do they regard sciences such as physiology and pharmacology as relevant. Hence they do not understand why statistical knowledge cannot replace biological and medical knowledge, nor why the average of 20 000 crudely measured randomized trial patients is not a substitute for the knowledgeable and careful study of individual patients. Since epidemiology emerged as a separate specialty, it has made no significant contribution to medicine but has led to many errors and false emphases. (All this is compounded by the dominant left-wing political agenda of almost all epidemiologists.)

2. Climate change scientists are fixated on fitting computer models to retrospective data sets, and cannot recognize that retrofitted models have zero intrinsic predictive validity. The validity of a model comes from the prediction of future events, from consistency with other sciences relevant to the components of the model, and from consistency with independent data not included in the retrofitting. Mainstream climate change scientists fail to notice that complex computer modelling has been of very little predictive or analytic value in other areas of science (macroeconomics, for instance). They don't even have a coherent understanding of the key concept of global temperature – if they did have a coherent concept of global temperature, they would realize that it is a _straightforward_ matter to detect changes in global temperature – since with proper controls every point on the globe would experience such changes. If the proper controls are not known, however, then global temperature simply cannot be measured; in which case climate scientists should either work out the necessary controls, or else shut-up.

3. Functional brain imaging involves the truly bizarre practice of averaging of synaptic events: with a temporal resolution of functional imaging methods typically averaging tens to hundreds of action potentials and a spatial resolution averaging tens to hundreds of millions of synapses. There may also be multiple averaging and subtraction of repeated tasks. What this all means at the end of some billions of averaged instances is anybody's guess - almost certainly it is un-interpretable (just consider what it would mean to average _any_ biological activity in this kind of fashion!). Yet this stuff is the basis for the major branch of neuroscience which for three decades has been the major non-genetic branch of biological/ medical science - at the cost of who knows how many billions of pounds and man-hours. And at the end of the day, the contribution of functional brain imaging to biological science and medicine has been - roughly - none-at-all.

In other words, in the world of micro-specialization the each specialist’s attention is focused on technical minutiae and the application of conventional proxy measures and operational definitions. These agreed-practices are used in micro-specialities for no better reason than 'everybody else' does the same and (lacking any real validity to their activities) there must be some kind of arbitrary ‘standard’ against which people are judged. ('Everybody else' here means the dominant Big Science researchers who dominate peer review (appointments, promotions, grants, publications etc.) in that micro-speciality.)

Micro-specialists cannot even understand what has happened when there are fatal objections and comprehensive refutations of their standard paradigms which originate from adjacent areas of science.

In a nutshell, micros-specialization allows a situation to develop where the whole of a vast area of science is bogus; and for this reality to be intrinsically and permanently invisible and incomprehensible to the participants in that science.

If we then combine this fact with the notion that only micro-specialists are competent to evaluate the domain of their micro-speciality - then we have a situation of intractable error.

Which situation is precisely what we do have. Vast scientific enterprises have consumed vast resources without yielding any substantive progress, and the phenomenon continues for time-spans of several human generations, and there is no end in sight (short of the collapse of science-as-a-whole).

According to the analysts of classical science, science was supposed to be uniquely self-correcting - in practice, now, thanks in part o micros-specialization, it is not self-correcting at all. Either what we call science nowadays is not 'real science' or else real science has mutated into something which is a mechanism for infinite perpetuation of error.