Sunday, 20 April 2014

Scientist is what scientist does

Across the blogosphere the vexed question keeps surfacing: what is a scientist?  Do they have privileged access to knowledge denied to the rest of us.  Should their views be given the status of indisputable advice on how to conduct our lives? In medicine, the climate, economics, and in the conduct of our lives who are these modern day priests who whilst not calling us to church on Sunday or to prayers twice a day still demand our respect?  Governments and courts call them to give expert witness and their conclusions they call 'evidence backed'.  The post-modern era with its slogan 'anything goes' has come and gone but we are left with a resurgence of 'speculative science' driven by a multiplicity of agendas.   So, let us be clear: what makes a scientist?

Scientists come in two broad categories: theoretical and experimental.  There is also a third type we will return to shortly.  Theoretical scientists are those who, starting from some well established theoretical or observational background spin out a range of further theories which they claim explain or predict the nature of underlying reality in novel ways.  Some of the greatest theoretical scientists such as Einstein, Hawking, Newton and Schrodinger have created powerful explanatory theories giving us explanations of gravity, space time, black holes and the sub-atomic world and, crucially predictions of observable phenomenon which have been astonishingly successful.  However, their language is constructed of theoretical terms which do not have direct observational reference.  Take a term like 'temperature'.  Temperature is something we talk about all the time but how do we define it?  Very soon we discover that we are using other theoretical terms in our definition which do not have direct observational reference but appeal to further terms: heat, energy and so on which in their turn are not directly observable.  Theoretical scientists use predominantly mathematics to develop their conceptual understanding of the world: tensor geometry, number theory, calculus and so on.  By and large they are platonists, looking for that 'ideal reality' which the real world more or less imperfectly reflects.

Empirical scientists are, in their turn, of a different ilk.   Their interest is in constructing experiments. Their language is observational using terms which, they believe, will lead to measurement of empirical phenomena of various sorts.  Sometimes these experiments will be strictly controlled in a laboratory environment, sometimes the observation is less direct, undertaken in context perhaps through field trials or directly through observation.  Some try to directly engage the subject blurring the object-subject divide in order not to gain knowledge in the Platonic sense of 'justified true belief' but rather understanding of processes physical or social.

The problem for science is how to bridge the gap between the theoretical and the empirical in such a way that the former can be justified by the latter.   The classical way that this has been done is through the creation of 'hypotheses' - compound statements that linked theory with observation.  The problem is that observation is not theoretically neutral.   Observation relies upon measurement and measurement has its own theory of measurement.   The notion that observation is intrinsically 'theory laden' underpinned the challenge of Kuhn, Lakatos and Fayerabend against Popper's simple falsificationism as the defining standard for science.  Science, Popper argued is about putting 'conjectures' to crucial tests with the aim of refuting them by the simple logic If p is true then q is true, not q, therefore p is not true where p is the premise and q the consequent.

The problem, as Lakatos pointed out is that (i) all observation is theory laden and (ii) with any well formed theory there will be so many ceteris paribus clauses that it is impossible to determine uniquely what has been refuted.  Lakatos went on to propose what this blogger believes to be the most cogent exploration of how sciences develop.  He argued that scientists coalesce around a series of 'core terms' - theoretical constructs if you like which they consider irrefutable and which they protect against refutation through the creation of a range of ad-hoc modifications as a way of immunising them against challenge.   Theories are not the way science progresses, Lakatos argued.  They progress by scientists becoming committed to research programmes which either progress when they generate novel predictions or degenerate when the desire to sustain the core of belief against refutation overwhelms the empirical significance of the programme.  

However, since Lakatos untimely death in 1974 a new type of science has arisen.  It is where scientists seek to create computer based models of more and more complex social and physical systems reflecting a series of underlying laws which whilst not conjectural in themselves are spun into more and more abstract sets of interacting relationships.   By priming these computerised models of reality with a small number of empirically determined parameters and seeding them with given initial conditions projections of future outcomes are derived.   These theoretical realities, created in cyberspace, are designed to give an insight into how real systems operate.    In practice they rarely perform well in that what they project rarely maps onto what is observed and measured.   Indeed, what I will refer to as Bob's Law applies:  as models gain complexity they progressively lose any connection with reality.    This is an application of what I regard as an almost universal principal:  the more complex a model the less real are its outcomes.

No comments: