Tuesday 8 June 2010

Science is about coherence, not testing 'predictions'

Until recently I usually described science as being mostly a matter of devising theories which had implications, and these implications should be tested by observation or experiment.

In other words, science was about making and testing predictions.

Of course there is more which needs to be said: the predictions must derive from theory, and the predicted state should be sufficiently complex, so as to be unlikely to happen by chance.

But it is now clear that this sequence doesn’t happen much nowadays, if it ever did. And that there are weaknesses about the conceptualization of science as mostly a matter of testing predictions.

The main problem is that when science becomes big, as now, the social processes of science (i.e. Peer review) come to control all aspects of science, including defining what counts as a test of a prediction.

This is most obvious in medical research involving drugs. A loosely-defined multi-symptom syndrome is created and a drug or other intervention is tested. The prediction is that the drug/ intervention ‘works’ or works better than another rival, and the test of prediction involves multiple measures of symptoms and signs. Within a couple of years the loosely defined syndrome is being diagnosed everywhere.

Yet the problem is not at the level of testing, since really there is nothing to test – most ‘diagnoses’ are such loose bundles that their definition makes no strong predictions. The problem is a the level of coherence.

Science is a daughter of philosophy, and like philosophy, the basic ‘test’ of science is coherence. Statements in science ought to cohere with other statements in science, and this ought to be checked. Testing ‘predictions’ by observation and experiment is actually merely one type of checking for coherence, since ‘predictions’ are (properly) not to do with time but with logic.

Testing in science ought *not* to focus on predictions such as ‘I predict now that x will happen under y circumstances in the future’ – but instead the focus should be – much more simply – on checking that the statements of science cohere in a logical fashion.

It is an axiom that all true scientific statements are consistent will all other true scientific statements. True statements should not contradict one another, they should cohere.

When there is no coherence between two scientific propositions (theories, 'facts' or whatever), and the reasoning is sound, then one or both propositions are wrong.

Scientific progress is the process of making and learning about propositions, A new proposition that is not coherent with a bunch of existing propositions may be true, and all or some of the existing propositions may be false indeed that is the meaning of a paradign shift or evolutionary science: when new incoherent propositions succeed in overturning a bunch of old propositions, and establishing a new network of coherent propositions.

(This is always a work in progress, and at any moment there is considerable incoherence in science which is being sorted-out. The fatal flaw in modern science is that there is no sorting-out. Incoherence is ignored, propositions are merely piled loosely together; or incoherence is avoided rather than sorted-out, and leads to micro-specialization and the creation of isolated little worlds in which there is no incoherence.)


Using this very basic requirement, it is obvious that much of modern science is incoherent, in the sense that there is no coherence between the specialties of science – specialties of science are not checked against each other. Indeed, there is a big literature in the philosophy of science which purports to prove that different types of science are incommensurable, incomparable, and independent.

If this were true, then science as a whole would not add-up – and all the different micro-specialties would not be contributing to anything greater than themselves.

Of course this is correct of modern medical science and biology. For example ‘neuroscience’ does not add up to anything like ‘understanding’ – it is merely a collection of hundreds of autonomous micro-specialties about nervous tissue. This, then that, then something else.

These micro-specialties were not checked for consistency with each other and as a consequence they are not consistent with each other. Neuroscience was not conducted with an aim of creating a coherent body of knowledge, and as a result it is not a coherent body of knowledge.

‘Neuroscience’, as a concept (although it is not even a concept) is merely an excuse for irrelevance.

It is not a matter of whether the micro-specialties in modern science are correct observations (in fact they are nowadays quite likely to be dishonest). But that isolated observations – even if honest - are worthless. Isolated specialties are worthless.

It is only when observations and specialties are linked with others (using theories) that consistency can be checked, that understanding might arise - and then ‘predictions’ can potentially emerge.

Checking science for its coherence includes testing predictions, and maximizes both the usefulness and testability of science; but a science based purely on testing predictions (and ignoring coherence) will become both incoherent and trivial.