Saturday, October 5, 2013

Underdetermination - a relevant concept for climate science?

During a visit to ETHZ, a philosopher of science lead my atttention to a concept named "underdetermination", of which I was unaware. She gave me the paper of 
Martin Carrier: Underdetermination as an epistemological test tube: Expounding hidden values of the scientific community. Synthese (2011) 180:189–204 DOI 10.1007/s11229-009-9597-6, 
The abstract reads
"Duhem–Quine underdetermination plays a constructive role in epistemology by pinpointing the impact of non-empirical virtues or cognitive values on theory choice. Underdetermination thus contributes to illuminating the nature of scientific rationality. Scientists prefer and accept one account among empirical equivalent alternatives. The non-empirical virtues operating in science are laid open in such theory choice decisions. The latter act as an epistemological test tube in making explicit commitments to how scientific knowledge should be like."
and the conclusions are begun with 
"The underdetermination thesis says that any given set of data can always be represented by empirically equivalent, conceptually distinct accounts. The range of relevant options is dependent on how empirical equivalence and conceptual distinctiveness is spelled out but irrespective of such details, the thesis establishes a leeway for scientific theory when faced with the verdict of nature. The reason for the significance of this leeway is that the criteria appealed to in picking an account from the collection of empirically admissible options bear witness to our epistemological intuitions.
I wonder what, if any,  it means for our (climate) practice? As common in such texts, the examples are from quantum mechanis and celestial dynamics, and not from environmental sciences.


Anonymous said...

What it means is that science cannot do without methodology. When undeterminism was discovered and noted by Poincaré and Duhem during the late 19th century, they claimed that science was driven by concentional choices. Hence the name "conventionalism".

Logic alone is not sufficient: For any set of observations thre can in principle exist many different theoretical explanations. Karl Popper didn't like the conventionalistic view and suggests in his Logic Der Forschung (1934) a number of methodological rules aiming at eliminating at least some of these logically possible theories, typically theories which uses ad hoc and other conventionalistic stratagems to avoid falsification.

In the case of climate science it means that logic alone will never be sufficient for determining if, for example, the main AGW-theory is false. What one has to keep an eye on is if proponents of such a theory reformulates the theory in such a way that it becomes non-falsifiable using for example vague claims, ad hoc explanations when things do not fit, etc. Because that would be counter to good scientific methodology.

Ingemar Nordin

Anonymous said...

I found this:

"Quines These der Unterdeterminiertheit empirischer Theorien, der gemäß es in ihrem theoretischen Gehalt verschiedene, aber dennoch empirisch äquivalente Theorien geben kann, ist nach wie vor mit Unklarheiten behaftet. Eine genaue Analyse der entscheidenden Textstellen zeigt, daß die These offensichtiich nur auf nicht-kompakte Theorienformulierungen angewendet werden kann, daß aber das Kriterium der Kompaktheit recht schwer zu verstehen ist. Darüberhinaus erweist sich der Ausdruck „empirischer Gehalt" als ebenso klärungsbedürftig wie „theoretischer Gehalt". Beide Ausdrücke sind jedoch für das Verständnis der These von entscheidender Bedeutung. Es zeigt sich, daß so dem Versuch einer befriedigenden Interpretation trotz der Modifikationen, die Quine seit der Lancierung der These in On Empirically Equivalent Systems of the World vorgenommmen hat, beträchtliche Schwierigkeiten im Wege stehen."



Hans von Storch said...

If someone wants a copy of the Carrier-paper, please let me know -

eduardo said...

One of the criticism of paleoclimate reconstructions based on proxy data is that researchers measure many records, for instance tree-ring chronologies. Among those many records only a few seem to be correlated with measured temperatures, which could be due to shear chance and not represent a true physical relationship.

In the realm of scientific theories - not only climate research - something similar could be happening. Many theories are tested, but only those that seem to fit the observations survive. Could this apparent success be due to shear luck ?

@ReinerGrundmann said...


this is a good question and the answer is probably: we don't know.

The 'underdetermination' thesis (u-t) has different versions, some authors thought it applies only to physics, others thought it applies to all knowledge.

If we apply the thesis to itself, we would need to know what the data is upon which the u-t is founded (all fields of science?). According to the thesis, there should be another theory compatible with the available data. In this sense the u-t would require the emergence of at least one competitor theory or thesis.

It follows that we cannot overcome the circularity in the argument about the 'truth' of scientific theories.

Because of such considerations, some philosophers of science are convinced that epistemology needs to be 'anti-foundational': there are no firm fundamentals upon which to build scientific edifices. The metaphors used are of piles being driven into quicksand, or of a ship which is rebuilt in open sea without having the option of going back to the dock, as Otto Neurath famously put it in 1932.

@ReinerGrundmann said...

I quoted from memory. Popper used the metaphor of driving piles into a swamp (not quicksand). Neurath coined the boat metaphor already in 1913. I have an unpublished manuscript which deals with these questions, just uploaded to my Academia pages.

Comments welcome!

Mathis Hampel said...

A theory is always underdetermined. the point is to figure out which theory is less underdetermined.

At the moment it looks like the climate shifts hypothesis is less underdetermined than AGW.

To decide which one to believe in we need to understand both the logic of scientific discovery (usually ignored by constructivists) and its politics (usually ignored by positivists).

Anonymous said...


Pure luck does not need a new thesis, like the Gödel's incompleteness theorems for instance. If it could tell us something new about the character of physical laws it would be helpful.

Pure luck can easily be excluded if you really want to.



Anonymous said...

Eduardo said:

"Many theories are tested, but only those that seem to fit the observations survive."

I have got the impression that in scientific practice it is too often the other way round:

Many observations are made, but only those that seem to fit the theory survive.

@ReinerGrundmann said...

It is perhaps even worse as : the Economist says

@ReinerGrundmann said...

The missing link from the Economist

Martin Mahony said...

You may also be interested in this paper by Gregor Betz:

'Underdetermination, model-ensembles and surprises: On the epistemology of scenario-analysis in climatology'