Monday, 4 January 2010

Teaching philosophy of science

I am teaching a philosophy of science module for the first time in a while (by contrast with a philosophy of science for psychiatry module) and it gives me the chance to think what overall view I have of the area (via the question of the overall narrative for the module).

Like both Rachel Cooper (to whom, congratulations on her first child, by the way) and Matthew Ratcliffe, I spent time in the Cambridge HPS department and thus was inducted into a view of the philosophy of science as a special and interesting approach to philosophy rather than just one area among others. That view, however, was (sequentially) held in place by two quite particular views of both science and the role of history and philosophy of science departments (of which there were a flurry in the 1960s and 70s).

On the first, science was held to have a central place in exemplifying rationality and thus inheriting the latter’s key importance for philosophy as a whole. The role of history and philosophy was (widely held to be) as follows: the latter outlined an a priori model of scientific rationality. The former then filled in the explanatory gaps when scientists failed to be governed by that model. (Normally, rationality was its own explanation. Only deviation needed further historical explanation.)

With the rise of modern sociology of science, that approach received a substantial tweak. Science still has a central role in exemplifying rationality but now rationality itself is seen as open ended. The ongoing development of scientific practice constructs the standards of rational thinking. And thus, whatever light philosophy can shed on them, there is a fundamental role for the history or sociology in charting those changing practices.

On both of these pictures, scientific method (and hence methodology) is the main focus. But on the second picture, and partly because of the failures, worked out by both sociologists and philosophers, of the first picture, there seems to be less and less work for a philosophy of science. Methodology has fallen out of favour with philosophers of science. That still leaves, of course, a number of important topics for philosophical exploration (including explanation, laws, causes, the nature of induction, the relation of theory and observation, the under-determination of theory by evidence etc), but the narrative that ties them together has changed. So what should the narrative be now?

My instinct is that that whilst the obvious answer to them has proved more complicated, the underlying questions are still the same: what is special about science as an instance of rationality? what account can and should we give of its status? what insight can we have into its success? Now, I think, the answers have to be more nuanced than I might once have thought.

There are some things important we can say about scientific method (and hence some gestures to why science is distinct from non-science) that seem particularly important given what hangs on science (eg the debates about climate change or MMR). But also some things where it is important also to say what paths are not taken by placing nomological explanation at the heart of our ways of making sense of the world. There are some issues where philosophical insight into science seems central and others where it is best to leave it to a local view (such as local questions of method).

So whilst it seems a mistake to think that science is our exclusive shot at rationality, the complexities make the relation between our ideas about what is rational in general, and this central family of approaches in particular, more rather than less interesting. The fact that scientific method cannot be codified says something good about human scientists - their judgement - rather than bad about science.


1: Induction and its problemsPerhaps science explains events by fitting them into inductive patterns. Patterns that have held in the past will hold in the future. But can induction be justified? What of counter-induction (the principle that patterns that have held in the past will not hold in the future)? And even if some form of induction holds, what properties should we project into the future (is grass green, or is it grue?).

2: FalsificationismPopper argued that the problem of induction can be solved (or perhaps more accurately side-stepped) by saying that science should aim at conjecture (forming a hypothesis or perhaps just a guess) and refutation (trying to prove it wrong) since whilst it seems impossible to prove through enumeration that all swans are white, a single black swan refutes that hypothesis. But how straight-forward is refutation? And what kind of positive explanation does Falsificationism permit?

3: The Theory Dependence of ObservationWhether one aims to confirm or refute theories, observational evidence surely plays a key role in the rise and success of science. That seems to require that theory and observation are distinct so that the latter can be a true test of the former. But there are powerful arguments to think that both observation statements or reports and the very process of observation is theory dependent. So how should we understand the role of observation?

4: Duhem-Quine thesis and Falsificationism #2The thesis that observation is infused with theory suggests a complication to the relation of theory and data. If theories can only be tested against a background of other, assumed, hypotheses, then a theory could be defended against contrary evidence by suitable changes to these auxiliary hypotheses. This suggests that science is a matter of convention rather than truth. Can this be domesticated within a picture which is still broadly falsificationist?

5: The Deductive-Nomological (DN) model of explanation
So far we have assumed that explanation has something to do with fitting particular events to general patterns and then worried about how general theoretical patterns can be established. But what, exactly, is explanation. One influential model says that it is a matter of inferring the event to be explained from general laws in logical arguments.

6: Laws and accidentsIf explanations do depend on laws of nature, how are they characterised? One way to approach that question looks back to Hume’s analysis of causation and asks what real progress is made by moving from singular instances of causation to general patterns, as Hume suggests. What is the difference, at the general level, between accidentally true generalisations and genuine laws?

7: Objections to the DN model and an alternative causal model of explanationSo let’s assume that we can tell the difference between laws and accidents, does that fill out the DN model of explanation? Not yet. First there are some standard objections to the DN model (both apparently good explanations it rules out and crazy ones it lets in) and second there’s a rival causal account. It claims that the difference between genuine and fake explanations is that the latter trades in causal information. But perhaps laws are the way that genuine causal information is presented?

8: The challenge of epistemological anti-realism
So now it seems that causal information is key to science, that laws may play an important role in charting genuine causal relations and that evidence, although ‘infected’ by theory, is still key. But this package of ideas still comes under threat. Since evidence is, by definition, observable but since the observable evidence is consistent with an infinite number of different theories about the unobservables, it is irrational to be other than agnostic about whole tracts of the world.

9: Two responses to anti-realism: causal agents and inference to the best explanation
Two responses are critically examined. First, the view that whilst what are advanced as laws of nature may well be false, that is no reason to doubt explanations which invoke unobservable causal agents. Second, there is more to the relation of theory and evidence than mere consistency (ie that their joint truth is not logically ruled out). We can infer from observable evidence to what best explains it. On this view, explanation is, again, key to science.

10: Kuhn and the later rise of the sociology of science
The difficulty with outlining either a confirmatory or a falsificationist methodology of science together with the various challenges to simple accounts of explanation, observation and evidence suggests the opportunity for a more detailed account of scientific development. In this session we look to Kuhn’s account of revolutionary jumps between incommensurable paradigms.

11: Explanation versus understanding
The module has placed explanation at the heart of scientific thinking (and looked at the connections between it, causation, laws and evidence). But what does this rule out? In this session we look at the idea that understanding and explanation are distinct and thus that there can be no such thing as social science. Can that really be true?

PS (one year later): I taught the module pretty much as described here. Two problems: I took too long to get going (in effect, two introductory sessions rather than one) and the discussion of the metaphysics of laws got in the way of the main storyline. So this is a link to the revised module for 2011.