Default Privilege and Bad Lots: Underconsideration and Explanatory Inference

Abstract

The underconsideration argument against inference to the best explanation and scientific realism holds that scientists are not warranted in inferring that the best theory is true, because scientists only ever conceive of a small handful of theories at one time, and as a result, they may not have considered a true theory. However, antirealists have not developed a detailed alternative account of why explanatory inference nevertheless appears so central to scientific practice. In this paper, I provide new defences against some recent objections to the underconsideration argument, while also developing an account of explanatory inference that both survives these criticisms and does not entail realism.

Acknowledgements

I would like to thank the two anonymous referees of this journal, Brad Cokelet, and students in my 2009 ‘Concepts of Explanation’ seminar at Middlebury College for their insightful feedback on earlier drafts of this paper.

Notes

[1] As Stanford (2006) argues, if underconsideration arguments are not to trade in merely idle sceptical possibilities (on a par with worrying that scientists are brains in vats), then these arguments should be historically grounded.

[2] Wray (2008, 322) cites Carrier (1991), Laudan (1984), and Stanford (2006) as providing particular difficulties for Lipton’s argument.

[3] Indeed, by way of reductio, with further argumentation, this could refute the idea that one must adopt a realist attitude towards background theories. This would have ramifications for Psillos’s more global defence of the approximate truth of background theories, which rests on an abductive defence of IBE. Since this paper is only concerned with underconsideration, the details of this issue would lead us too far afield. Iranzo (2008) critiques these further aspects of Psillos’s argument.

[4] Another option (Kuipers 2004) is to retain a broadly reliabilist bent, while also avoiding commitment to Privilege and weakening the conclusions of explanatory inferences so that the best available hypothesis is inferred only as the available hypothesis that is closest to the truth. This still capitulates a good deal to the underconsideration argument, since it is consistent with the best available hypothesis being very remote from the truth, i.e. the best of a bad lot. As a result, I suspect that the ‘constructive realism’ it licences is more constructive than realist, though I certainly cannot argue for that here without going too far afield.

[5] The majority position treats responsible epistemic behaviour as only worth its salt if it is also reliable; though Brandom (2000, ch. 3) and Williams (2008) subvert reliability to responsibility.

[6] Others have identified these features of explanatory justification, e.g., ‘very few beliefs can be warranted if the notion of warrant involves elimination of the possibility that the belief be false’ (Psillos 1999, 216); ‘believing that there may be a better explanation than the best available one does not undermine justification’ (Iranzo 2001, 83). However, they do not develop this point.

[7] The moniker is based on a similar model of justification endorsed by Brandom (1994) and Williams (2001, 2008). Williams cites Carneades, Austin, Wittgenstein, and Harman as proleptic proponents of the default and challenge model. Rolin (2006; 2008) has also applied this model to issues in the social epistemology of science.

[8] Rolin (2006) provides additional examples of default entitlements, including middle range hypotheses and social and political values. Williams (2001) also notes that in many contexts, we have a default entitlement to deny sceptical hypotheses, e.g. I am not a brain in a vat.

[9] Cf. Sellars (1963), Rosenberg (1980), and Brandom (1994) for more on the semantics informing this idea. Sellars and Rosenberg also invoke explanatory considerations in their epistemology, but do not invoke default and challenge; Brandom, the opposite.