Realism Without Interphenomena: Reichenbach's Cube, Sober's Evidential Realism, and Quantum Solipsism

ABSTRACT

In ‘Reichenbach's cubical universe and the problem of the external world’, Elliott Sober attempts a refutation of solipsism à la Reichenbach. I here contrast Sober's line of argument with observations a hypothetical scientist makes on what is known as the ‘Mermin contraption’ and show that Sober's extension of Reichenbach's argument fails. However, there is a further line of defense against solipsism, endorsed by some interpreters of the quantum theory whose position is otherwise rather close to solipsism. As I shall argue, this line of defense remains viable, but we still fall short of a story in terms of what Reichenbach calls ‘interphenomena’ in the case of the Mermin contraption.

Acknowledgments

The author thanks Alex Gebharter for some help with causal graphs, and an anonymous referee for helpful comments.

Notes

1 By ‘logico-analytic independence’, I mean that the co-occurrence of X and Y is neither fixed by a meaning postulate, such as the co-occurrence of being a bachelor and being unmarried, nor by logical implication, such as the co-occurrence of heads with heads-or-tails. Neither of these would be in need of explanation other than by pointing out the logical or conceptual relations between them.

2 I deliberately leave it open whether X and Y are event types or tokens, for reasons raised in Fn. 7.

3 Sober uses ‘correlation’ to exclusively refer to a theoretical property associated with probabilities, and ‘association’ to refer to the observed regular coincidence of two events of distinct types. I find ‘empirical correlation’ the more natural terminology, and I will allow myself to sometimes suppress the ‘empirical’ when the context allows this.

4 I will here not question the notion that likelihood provides a sensible measure for evidence in this way. A detailed defense is offered, for instance, by Forster and Sober (2004).

5 As a referee has rightly pointed out to me, however, this already makes the example considerably less forceful.

6 The debate actually reaches back further, to Yule (1926).

7 It is unclear to me whether this is equally true of the other counterexample discussed by Sober (2001, 335). Sober considers the correlation, over a lifetime, between the heights of specimens from two species that are homoplaseous rather than homologous. That is, the two species have a common trait that ‘evolved independently in the two lineages’ and is not ‘inherited unmodified from a common ancestor’ (Sober 2001, 335). To give Sober's own example: robins and sparrows have inherited their wings from a common ancestor (they are homologous) whereas robins and bats have not (they are homoplaseous). However, finding specimens from two species which have extremely similar height development requires a contrived selection of individuals. If a fair sample is used, I suspect that the empirical correlation is rather nullified. On the other hand, if expressed as a relation on the type level, such as ‘the average relative height of specimen of species x is almost exactly the same as that of specimen of species y after such and such a fraction of their life span’, there is a perfectly acceptable type-level common cause: similar selective environmental pressures. Sober's (2001, 339) claim that ‘the principle of the common cause[…] is about common token causes’ I believe to be incorrect as it stands (see, for instance, the discussion of Reichenbach's theatre troupe in Hitchcock and Rédei 2020).

8 See https://www.ons.gov.uk/economy/economicoutputandproductivity/output/articles/fivefactsabouttheukservicesector/ \Clinicaltrialid {\Clinicaltrialid {2016-09-29}}.

9 See, e.g. https://www.bbc.com/news/science-environment-53921121

10 Wroński (2014, 28) actually identifies very similar causal variables, but never extends his considerations to asking whether melting sub-polar glaciers and deteriorating crop levels are causally related, and so does not offer a common cause explanation.

11 In particular, the last independence implies that $P(S\wedge D\wedge E\wedge W\wedge B|C)=P(S\left|C\right)P(E\wedge D\wedge W\wedge B|C)$. Marginalising both sides for D,E and W, we have the desired result. It also follows directly from the ‘decomposition axiom’ assumed in some causal modelling frameworks (Judea Pearl 2009, 11).

12 I believe this is what Sober is getting at with his homoplasy-example, but my reservations expressed in Fn. 7 remain.

13 I do not count, say, impressions of waves on a muted TV here; these are too clearly distinct from ‘original V s’.

14 I have in mind, of course, especially Christopher Fuchs, N. David Mermin, and Rüdiger Schack.

15 For instance, consider Hilary Putnam's (1981, 63) reading of Kant in this connection, according to which ‘any judgment about external or internal objects (physical things or mental entities) says that the noumenal world as a whole is such that this is the description that a rational being (one with our rational nature) given the information available to a being with our sense organs (a being with our sensible nature) would construct. […] There is not even a one-to-one correspondence between things-for-us and things in themselves’.

16 We can imagine this to be brought about by further devices hooked up to A and B, respectively.

17 Note that this is also true of proposals beloved by many philosophers, such as Bohmian approaches (cf. Wood and Spekkens 2015, 20).

18 Paul M. Näger (2016) is one of the few to distinguish the causal aspects of this type of scenario explicitly from its implications for the relativity theories. Note that by now, these impossibility results based on purely causal considerations have been generalised also to the impossibility of classical causal models of contextuality (Cavalcanti 2018; J. C. Pearl and Cavalcanti 2019).

19 A notable exception is Jennan Ismael (2020). However, she does not really address the problem that any common cause explanation of quantum correlations involves fine-tuning (Wood and Spekkens 2015; Näger 2016).

Additional information

Funding

Some research for this paper was done during my employment with the research unit The Epistemology of the Large Hadron Collider, funded by the German Research Foundation (Deutsche Forschungsgemeinschaft) (DFG; grant FOR 2063).