Two arguments for scientific realism unified
Introduction
Inferences from scientific success to the approximate truth of successful theories remain central to the most influential arguments for scientific realism. Success can be construed in a variety of ways, but verified novel predictions are often considered the most compelling examples for the claim that science couldn’t be that successful if our theories weren’t at least approximately true.1 Consider Fresnel’s theory’s verified prediction that when a narrow beam of light is directed towards an opaque disc or sphere a white spot will appear at the centre of the shadow that’s created. Intuitively it can seem just too incredible that Fresnel’s theory should correctly anticipate the unexpected phenomenon, yet the theory not be at least approximately true. Nevertheless, even theories that enjoy novel success are sometimes replaced, and sometimes replaced by theories with radically distinctive ontologies.
Historical examples of radical shifts in scientific understanding create well known challenges for the realist. Where the incongruence between two successful theories is sufficiently acute, for instance, we can no longer describe each as approximately true without stretching the notion so thin as to render it useless for realist ends. Examples of such incongruence thus appear to undermine the realist’s conviction that success is a reliable indicator of truth.2 Historical considerations seem to recommend we relinquish realist sympathies.
Several recent responses to the historical challenges remind us that theories divided by even the most revolutionary change still share at least some common assumptions concerning unobservable entities and processes. That continuity recommends a more selective style of realism should be pursued, a realism that commits only to what has been systematically retained across theory change, and a realism that is thereby sensitive to the extent and content of the changes with which history presents us. Selective realist theses shift the burden away from establishing the approximate truth of entire theories (much less of all science), but maintain, first, that our best theories furnish us with knowledge that extends beyond directly observable phenomena and, second, that scientific success is a reliable indication that we have achieved such knowledge.
Selective realism is typically advertised principally as a defence of the inference from success (where realists now infer from success to the approximate truth of only certain parts of theories), but a further argument for such theses has emerged from considerations of science’s history. The second argument, which I’ll call the argument from retention, takes evidence that parts of our past theories are, as a matter of historical fact, preserved within current theories as an independent reason to believe that those stable elements are at least approximately true.3
The arguments from success and from retention would ideally complement and reinforce one another: success would be used to infer which parts of past theories we have reason to believe are approximately true; evidence for the retention of those same parts across subsequent theory change would further justify a realist attitude towards them. A case could thereby be made that science is cumulative, preserving the source of previous successes and dispensing only with what was redundant for such achievements. A failure to unify these two arguments leaves realism in a much attenuated condition. First, if we divide past theories into the parts responsible for success and the parts that were inert but are then confronted with historical evidence that even the parts deemed responsible for success are not being retained, then selective realism has demonstrated no greater capacity for accommodating historical change than the original realist position.4 If, alternatively, selective realism is founded solely upon apparent continuities within the history of science, then realists face the sceptical challenge that having been preserved hitherto is a poor reason for believing something will be preserved heretofore. Furthermore, a failure to relate appropriately the successes of past theories to those parts that were actually retained is a failure to rescue the realist’s inference from success, which, as noted, is usually identified as the primary motivation for developing selective realist theses. As I’ll argue in Section 5, however, it’s unclear whether previous selective defences are capable of producing a synthesis of the two arguments, irrespective of the actual historical details.
In this paper I offer a new strategy for unifying the arguments from success and retention. In Section 2 I reconsider the significance of verified novel predictions and argue that, despite their centrality within recent discussions of scientific realism, an important quality of extant cases of novel phenomena has been persistently overlooked. Specifically, verified novel predictions appear notable, at least in part, for their implication that the successful theory improves upon competing theories. A comparative notion of success sends the realist’s inference from success in a different direction from those previously developed. Most importantly, comparative achievements do not imply any particular degree of proximity to some target, only relative proximity; adopting a comparative conception of success thwarts realist ambitions to infer the approximate truth of even the best confirmed of theories. In Section 3 I follow the lead of previous selective realist theses and maintain that scientific realism can survive and thrive without insisting that our theories are approximately true. In Sections 4 Selective realism and scientific progress, 5 How the arguments might be unified I show that focussing on comparative achievements improves on previous selective theses by unifying the arguments from success and retention, contingent on favourable historical evidence. Finally, in Section 6, I argue a further and underappreciated advantage to selective realist theses: properly understood, they need not rely on the oft denounced realist strategy of appealing to realism’s explanatory achievements.
Section snippets
Success, novel phenomena and excess content
Certain empirical results strike us as providing particularly salient support for certain scientific theories. The phases of Venus, Fresnel’s white spot, and Einstein’s light-bending results are familiar examples. Exactly how important any particular empirical result was (or should have been) in convincing rational agents of the merits of a given theory is difficult to evaluate, but cases such as these cannot be sensibly ignored. Of such achievements realists have recognized, and considered it
Scientific realism without approximately true theories
Conceiving of success comparatively entails only that successful theories are better than the theories they succeed, but progress alone cannot justify an inference to the approximate truth of a theory. Relative achievements imply only relative distances to some objective; adopting a comparative notion of success frustrates the realist’s hope of inferring the approximate truth of theories on the basis of success. Advocates of previous selective realist theses, however, have already distanced
Selective realism and scientific progress
When one theory is judged better than another, then, on any conception of progress, it is sensible to suppose that not all parts of the preferred theory will be involved in the perceived advancement. Our judgement that one theory is better than another cannot be grounded in the shared assumptions of the two theories, for example. Rutherford’s model of the atom is better than Thomson’s, but not because Rutherford described atoms as a seat of both positive and negative charges, since the two
How the arguments might be unified
According to selective realists, greater attention to the nature of past successes and their sources justifies a restricted realist thesis. Advocates offer historical analyses, in part to add credibility to such attitudes. An important antirealist response has thus been to challenge the historical reconstructions and provide new counterexamples to the selective theses. Stanford (2006) argues that neither Kitcher’s nor Psillos’s interpretations of the aether theories are convincing. Chang (2003)
Selective realists needn’t rely on realism’s putative explanatory advantages
Throughout the paper I have been careful not to frame the realist’s inference from success as an inference to best explanation. Both the arguments from success and retention might plausibly be interpreted as appealing to the explanatory power of scientific realism: realism provides the best explanations, it is often suggested, for both scientific success and the stability of various constituents of theories across changes in theory. Appeals to realism’s explanatory superiority, however, have
Conclusions
Conspicuous examples of verified novel predictions involve one theory predicting a phenomenon that is anomalous for rival theories, suggesting that such results are valued in part for the implication that progress has been achieved. Realists have often leaned on such results, but drawn no attention to the comparative dimension. There are advantages to addressing the deficit.
Selective theses appear to provide the realist with her best hope of answering antirealist objections based on the history
Acknowledgements
Thanks to Bill Hart, Dave Hilbert, Andy Blom, Rogerio Severo, Dearbhla Kelly, an anonymous referee from this journal, and, in particular, Nick Huggett for helpful comments on earlier drafts.
References (37)
What is structural realism?
Studies in History and Philosophy of Science
(1998)Reconsidering the Fresnel–Maxwell theory shift: How the realist can have her cake and EAT it too
Studies in History and Philosophy of Science
(2005)- et al.
Prediction and the periodic table
Studies in History and Philosophy of Science
(2001) The paradox of predictivism
(2008)Realism: Deconstructing the debate
Ratio
(2002)The rise of the wave theory of light
(1989)Preservative realism and its discontents: Revisiting caloric
Philosophy of Science
(2003)- Fine, A. (1984). The natural ontological attitude. In J. Leplin (Ed.), Scientific realism (pp. 83–107). Berkeley:...
Unnatural attitudes: Realist and instrumentalist attachments to science
Mind
(1986)Likelihoodism, Bayesianism, and relational confirmation
Synthese
(2007)
Remodeling structural realism: Quantum mechanics and the metaphysics of structure
Synthese
On the predilections for predictions
British Journal for the Philosophy of Science
Prediction versus accommodation and the risk of overfitting
British Journal for the Philosophy of Science
The advancement of science
Understanding philosophy of science
Cited by (5)
An ignored argument for scientific realism
2020, Filozofia NaukiResisting Scientific Realism
2018, Resisting Scientific RealismSuccess of science as a motivation for realism
2017, The Routledge Handbook of Scientific RealismA Likely explanation: IBE as a guide to better (but not more probable) hypotheses
2011, South African Journal of Philosophy