Tuesday, 21 September 2010

Positivist negatives

I mentioned in a previous post that I didn't think Comte's name choice was very good in describing the three main stages of development within a science.  Well, I've been reading J.S. Mill's "August Comte and Positivism"  and he also thinks the name choices are a bad idea.  He makes some revealing alternative suggestions, namely: the personal, the ontological and the phenomenological (rather than theological, metaphysical and positive) - this captures a lot more of the influence of Kant on all this (read noumenal for ontological).  And it shows how this Kantian distinction makes its way through Comte into philosophies of science as diverse as Husserl's and those of the Vienna Circle. In the personal approach to explaining nature, everything happens not through invariable law, but because some being (spiritual or otherwise) makes it happen like that.  The approach tries to tease out the workings of these various beings.

The next development in scientific methodology involves the realisation that there just doesn't need to be a being effecting all of these relationships; instead one talks of an occult power or a force which achieves the same result.  The approach to science has been de-animated and the objects of scientific enquiry are Aristotelian final causes, or plastic forces, or vital principles.  Nonetheless this is still weird, and said forces reside in the noumenal.

Finally, we leave behind any talk of the noumenal (a move familiar to many Logical Positivists to come) and concentrate in establishing laws between observed phenomena.  Science becomes preoccupied with a method which allows relationships to be posited betweem observed phenomena and no commitment to any noumenal object is needed. To explain the distinction between the positive approach and the metaphysical approach (which I think as a distinction we can credit Francis Bacon and Thomas Hobbes) using modern scientific language, think of how scientists develop a model - perhaps mathematical, perhaps not - and then set about critiquing it through well designed experimentation.  There is no longer any moment where the scientist 'discovers' that the model corresponds to a final ontological reality - instead there's just the possibility of falsification of that model.  The realm of the model is the metaphysical realm, in the sense that we create a set of relationships which are more or less supported by some scientific evidence.  We don't necessarily get the model from phenomena (though models can be refined by experimental results) - we can postulate a model based on an inspiration, a guess, etc.  But the Comtian 'metaphysical' approach to science is tempted to infer the existence of real objects (Platonic Ideals).  I am reminded of Friedman's polemical position paper in economics which follows in this positivistic tradition.  But I am also reminded that economics is performative.  Options prices changed following the Black-Scholes solution was published in the early 70s and as traders started to use the result.  This performativity, if you think about it, can change reality so that it conforms to the model.  This is something totally unknown to Comte.  I'm particularly interested in this point, since Friedman's approach to economic modelling (the model can be simplistic, the assumptions unrealistic, but we prize it if it gets experimental support and proves useful) provided some support to the increasingly held 'rational expectations' view of people's economic behaviour as defined by a homogeneous agent with a simplistic institutional/communicative context (perfectly functioning, perfectly free markets).

Comte claims that each and every science independently goes through these three stages.  This seems plain wrong to me.  Sciences which have been around for quite a while might be exposed to these 2 earlier stages, but if it is a young science I would find it hard to believe that it is born in the modern world and still would have to makes its own way through the stages to reach the positive stage.  I think it'll just short-cut to the more modern conception of science.  Comte I believe didn't cover this possibility, though he did talk about a classification of the sciences.  Mathematics is at the root, physics next, etc.  Each downstream science is a proper superset of the sciences which came before it - and this relation between sciences (a measure, in fact) - corresponds to the experienced evolution of these sciences.  This doesn't seem right at all either.  As he was writing his theory, new branches of mathematics and science were being formed in peoples' heads.  If mathematics itself is changing all the time, how can it be a proper foundation of all the other sciences.  Take also for example the birth of genetics.  As Comte was thinking about and publishing his ground-breaking philosophy of science, Gregor Mendel was inventing the science of genetics.  Here's an example of a bit of so-called metaphysical thinking which was later supported by micro-biological evidence.  The ontological object known as the gene turned out to have a reality based in DNA located in every cell (excluding red blood cells and some skin cells) of an organism.

I think Comte’s stages apply to a couple of eras of science, essentially pre-Greek, Greek, post-Enlightenment.  All new-born sciences will immediately implement post-Enlightenment approach.

I also realise that this philosophy of science is only somewhat Hegelian, since the progress stops once the positive stage is reached.  This is more in line with Bohme than Hegel, who also could see an end point rather than a never-ending process. 

And I doubt that science is always and everywhere progressive and positive in Comte's sense.  I think there is an institutional (politico-cultural) which drives this, and our perception of it.

The phenomenological inviolate laws which comte sees as positive are no more grounded than the metaphysical stage ‘vital forces’ – the same gradation he applies to theological – from sprites to a single god, could equally well be applied to the metaphysical level, where the ‘final’ level is really the mature metaphysical approach – one real vital force which keeps all the laws in check.

If you apply an experimental approach to the various metaphysical  levels, perhaps you can do good science – maybe you’d just be making a lot more ontological (provisional) statements, then retracting them when a better ontological model presents itself.  This is not significantly different from the positive approach.

And the converse is probably true.  Sticking to identifying correlations between phenomena can lead to pseudo-science.  One example which immediately springs to mind is Elliot wave theory.

So what am I taking from Comte?  A certain post-Enlightenment hopefulness is all that's left for me.  Comte was, I think, too optimistic and too simplistic on science arriving at a final destination which would necessarily result in social goods.  He missed the performativity of models.  This might result in a kind of evolutionary inertia around models, as they 'bed in' by making reality conform to them.  This might not be too bad a thing if Comte's other view - that a positive science would necessarily lead to social good.  But I don't think this is the case, which is another way of saying that I think we should be careful about models with demonstrated performativity - we need to challenge them much more than Friedman et. al. would have you believe.   I plan to talk about Friedman's positivism in his approach to economics in a separate posting, so I'll develop that point later.

A question which interests me is how such a philosophy, which was quite socialistic, both in Comte and in J.S. Mill, was transformed into the much more individualistic classical economics approach of Ricardo, Walras, Jevons.  Comte talks a lot about how women and the working classes were the most likely to be first convinced by positivism, and he sets himself up in opposition to Marx and communism.

Performativity potentially changes everything – if we are to become what our models predict, then in the absence of certainty that a developed model is necessarily socially good, then model choice becomes somewhat of a politically charged, aesthetically charged, or morally charged activity.