Abstract
In an essay recently published in this journal (“Is Safety in Danger?”), Fernando Broncano-Berrocal defends the safety condition on knowledge from a counterexample proposed by Tomas Bogardus (Philosophy and Phenomenological Research, 2012). In this paper, we will define the safety condition, briefly explain the proposed counterexample, and outline Broncano-Berrocal’s defense of the safety condition. We will then raise four objections to Broncano-Berrocal’s defense, four implausible implications of his central claim. In the end, we conclude that Broncano-Berrocal’s defense of the safety condition is unsuccessful, and that the safety condition on knowledge should be rejected.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
See, for example, Duncan Pritchard (2005, 163): “If a believer knows that p, then in nearly all, if not all, nearby possible worlds in which the believer forms the belief that p in the same way as she does in the actual world, that belief is true.” And Steven Luper (2006): “at time t, S knows p by arriving at the belief p through some method M only if: M would, at t, indicate that p was true only if p were true.”
Pritchard (2005, 147–52) argues that it can capture the intuitively attractive idea that knowledge is non-lucky true belief, the central dogma of popular anti-luck epistemologies. Sosa (1999) argues that the view that knowledge must be safe gives an excellent account of inductive and anti-skeptical knowledge. And, according to John Hawthorne (2004, 56 n. 17), the view seems poised to explain why the subject in standard Gettier-style cases lacks knowledge: the subject could so easily have been wrong: “Insofar as we withhold knowledge in Gettier cases, it seems likely that ‘ease of mistake’ reasoning is at work, since there is a very natural sense, in such cases, in which the true believer forms a belief in a way that could very easily have delivered error.”
An anonymous referee helpfully encouraged us to consider another natural interpretation of condition (iii) in principle (R4), namely: the circumstances in which the target belief is formed via m 2 are in the set of circumstances that fix the global reliability of m 1 (to whatever degree it is reliable). There are two reasons to think this is the charitable interpretation of Broncano-Berrocal. First, it’s plausibly the more general principle—applying to reliable and unreliable methods alike—that lies behind and explains Broncano-Berrocal’s formulation of (R4), and it could easily be overlooked given the focus on generally reliable methods central to the debate. Second, it allows Broncano-Berrocal to avoid the objection of this section, since the unreliable method we described, when used on the second occasion—m 2 —would indeed be used within that range of circumstances that fix the (very low) global reliability of the method used on the first occasion—m 1 . And in that case our new interpretation of condition (iii) would be satisfied.
However, there is a strong reason not to prefer this alternative interpretation of condition (iii). Namely, it would take out some crucial legs from under Broncano-Berrocal’s response to Atomic Clock. For, in that scenario, the method one would use were the isotope to decay would be used in circumstances that indeed fix the “global” reliability of the method one actually uses. The global reliability of the actual method used in Atomic Clock is determined by its likelihood of error on a range of propositions in a range of circumstances across logical space. If the unreliability of the method in circumstances is no barrier to those circumstances being included among those that fix the method’s global reliability, and if the counterfactual conditions in Atomic Clock are sufficiently mundane (there’s no more soon-to-decay isotope, the clock has either slowed down or stopped as clocks often do, etc.), then it’s hard to see why those nearby counterfactual circumstances in which the isotope has decayed would not be among those that fix the global reliability of m 1 . But then this interpretation of condition (iii) is satisfied in Atomic Clock, and Broncano-Berrocal loses one of his only two arguments for the conclusion that condition (iii) is not satisfied in Atomic Clock. So, this interpretation of condition (iii) may rescue Broncano-Berrocal from one of our four objections, but at the high cost of sacrificing half of his defense of the safety condition.
This objection is due to Marxen.
Perhaps you’re a stickler here, and you wish to point out that there may well be some circumstances, however remote, in which these two methods would not be equally reliable. As a far-out but possible example, consider trying to do some multiplication problems in a room full of aggressive, muscular people who hate lattices, and who will rough you up should you try to draw any lattices. Your performance on those multiplication problems in that circumstance would likely be worse using lattice multiplication than using traditional long multiplication. Doesn’t that show that condition (i) is not met in this example, and so that we don’t have a counterexample to the sufficiency of the conditions of (R4)?
We respond: yes, that is a way to evade this counterexample, though it’s much less clear that it can evade the example of coin and the six-sided die given in the text, or other examples such as tyromancy and tasseography (reading tea leaves). What’s more, this evasion comes with a high cost: it raises a new threat to (R4). Suppose we individuate methods so finely that a method m 1 can be identical to a method m 2 only if their respective sets of ordered triples of the form <proposition, circumstance of evaluation, reliability>—for every proposition, every possible circumstance, and relativized to subjects—are exactly the same. Then we’d get bad results: methods that intuitively ought to come out as identical wouldn’t, according to (R4). For example, suppose you use lattice multiplication in ideal conditions on Monday, and call that method m 1 . Now suppose that, in virtue of that practice with lattice multiplication, you get slightly better at using it under those conditions and subsequently use it again on Tuesday, calling that method m 2 . The set of ordered triples—relativized to you—associated with m 1 on Monday is therefore different from the set of ordered triples associated with m 2 . With respect to some propositions, in those circumstances, the reliability of this method (for you) is higher on Tuesday than it was on Monday. And so, on this fine-grained principle of individuation (R4), m 1 ≠ m 2 . But clearly the method has remained the same: you’ve used lattice multiplication on both occasions. So (R4) is false on this strict construal of condition (i). Yet if we loosen up condition (i) so that methods can tolerate minor changes in their global reliability profiles, then the problems of this section will plague (R4). Therefore, the defender of (R4) has here a dispiriting dilemma.
We are grateful to an anonymous referee for encouraging us to interact directly with this quotation.
Just what does it take for circumstances to be among those in which a method is globally reliable? Broncano-Berrocal gives us a hint in Fake Barn Country: he says, of Henry’s method, that for circumstances “to belong to that set, the light conditions of the circumstances must be good as well, the distance must be appropriate, and so on.” And we are grateful to an anonymous referee for helpfully suggesting that, for Broncano-Berrocal, all and only a method’s “normal” circumstances of use will fix the method’s degree of global reliability.
As we’ve pointed out in this section, on this understanding of “global reliability,” (R4) entails that no method can fail in abnormal circumstances. But it gets worse: (R4) will also entail that no method can even be used in abnormal circumstances, for any such use will violate condition (iii). We’ll be forced to say, counterintuitively, that those brave astronauts who gazed back at Earth from the lunar surface used a method distinct from our vision. They didn’t see anything up there, in fact. And if no method can be used in abnormal circumstances, just what are we to make of this distinction between “normal” and “abnormal” circumstances in which a method is used? On this interpretation, the latter category is necessarily empty, for every method: belief forming methods can only be used in normal circumstances, circumstances in which they are reliable. That’s powerfully counterintuitive, and that’s a serious strike against (R4).
References
Bogardus, T. (2012). Knowledge under threat. Philosophy and Phenomenological Research. doi:10.1111/j.1933-1592.2011.00564.x.
Broncano-Berrocal, F. (forthcoming). Is safety in danger? Philosophia.
Hawthorne, J. (2004). Knowledge and lotteries. Oxford: Oxford University Press.
Luper, S. (2006). Restorative rigging and the safe indication account. Synthese, 153, 161–170.
Pritchard, D. (2005). Epistemic luck. Oxford: Oxford University Press.
Sainsbury, R. M. (1997). Easy possibilities. Philosophy and Phenomenological Research, 57, 907–919.
Sosa, E. (1999). How to defeat opposition to Moore. In J. Tomberlin (Ed.), Philosophical perspectives 13: Epistemology. Blackwell, 141–154.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Bogardus, T., Marxen, C. Yes, Safety is in Danger. Philosophia 42, 321–334 (2014). https://doi.org/10.1007/s11406-013-9508-4
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11406-013-9508-4