«˜ JUAN COMESANA UNSAFE KNOWLEDGE ABSTRACT. Ernest Sosa has argued that if someone knows that p, then his belief that p is “safe”, and Timothy ...»
ABSTRACT. Ernest Sosa has argued that if someone knows that p, then his belief that
p is “safe”, and Timothy Williamson has agreed. In this paper I argue that safety, as
deﬁned by Sosa, is not a necessary condition on knowledge – that we can have unsafe
knowledge. I present Sosa’s deﬁnition of safety and a counterexample to it as a necessary condition on knowledge. I also argue that Sosa’s most recent reﬁnements to the notion of safety don’t help him to avoid the counterexample. I consider three replies on behalf of the defender of safety, and ﬁnd them all wanting. Finally, I offer a tentative diagnosis of my counterexample.
1. INTRODUCTION Ernest Sosa has argued that if someone knows that p, then his belief that p is “safe”, and Timothy Williamson has agreed.1 In this paper I argue that safety, as deﬁned by Sosa, is not a necessary condition on knowledge – that we can have unsafe knowledge. In the next section I present Sosa’s deﬁnition of safety, and in Section 3 I present a counterexample to it as a necessary condition on knowledge. In that same section I also argue that Sosa’s most recent reﬁnements to the notion of safety don’t help him to avoid the counterexample. In Section 4 I consider three replies on behalf of the defender of safety, and ﬁnd them all wanting. Finally, in Section 5 I offer a tentative diagnosis of my counterexample.
2. SENSITIVITY AND SAFETY What else, besides truth and belief, is required for knowledge? Robert Nozick famously proposed as a further requirement that the belief in question be sensitive to the truth of the matter, where a belief that p by a subject S is sensitive if and only if S would not believe that p if p were false.2 In the terminology of possible worlds, a belief is sensitive for a subject just in case the subject does not believe it in any of the close possible worlds where it is false.
Soon after Nozick proposed it, the sensitivity requirement was subject
to counterexamples like the following:
Synthese (2005) 146: 395–404 © Springer 2005 DOI 10.1007/s11229-004-6213-7 396 ˜
JUAN COMESANAGARBAGE CHUTE: I throw a trash bag down the garbage chute of my condo. Some moments later I believe, and know, that the trash bag is in the basement. However, the closest possible world where my belief is false is plausibly one where, unbeknownst to me, the bag is stuck somewhere in the chute, and I still believe that it is in the basement.3 In GARBAGE CHUTE, my belief that the trash bag is in the building’s basement is not sensitive, yet it counts as knowledge.
The sensitivity requirement has also strong skeptical consequences. If our beliefs have to be safe to count as knowledge, then our beliefs that skeptical scenarios do not obtain (that we are not brains in a vat, that we are not dreaming right now, that we are not in The Matrix) do not amount to knowledge. If it were false that I am not a brain in a vat, then I would be a brain in a vat who believes that he is not a brain in a vat.
Nozick thought that this consequence of the requirement of sensitivity was a virtue of the notion, but many epistemologists think that we do know that skeptical scenarios do not obtain, and so they have accordingly taken this consequence as a further counterexample to sensitivity.
The sensitivity requirement is a counterfactual and, as such, is not equivalent to its contrapositive. Having noticed this, Ernest Sosa proposed that we replace the requirement of sensitivity with its contrapositive, which he calls safety. Sosa has several formulations of the safety requirement.
Here is his ﬁrst approximation:
Call a belief by S that p “safe” iff: S would not believe that p without it being so that p.
(Alternatively, a belief by S that p is “safe” iff: as a matter of fact, though perhaps not as a matter of strict necessity, S would not believe that p without it being so that p.) (Sosa 1999, 378) In terms of possible worlds, a belief that p by S is safe if and only if there is no close possible world where S believes that p and p is false.
That characterization of safety is not entirely satisfactory, though. One problem with it is that a good basis for my belief can preempt a bad one.
SICK PATIENT: I am seriously ill and I ask my doctor whether I will live one more week.
The doctor performs a test on me, and answers afﬁrmatively. I now base my belief that I will live one more week on the doctor’s testimony. But suppose that my condition is caused by a rapidly spreading virus, whose rate of infection is somewhat erratic, and which could have acted in my body more quickly than it actually did. Had it acted more quickly, the test would have indicated this and the doctor would have told me that I wasn’t going to live one more week. But suppose also that, had that happened, I would still have believed, out of wishful thinking, that I was going to live one more week.
In SICK PATIENT, I still know that I will live one more week (in the actual scenario, my belief is based on the reliable testimony of the doctor, and wishful thinking doesn’t play a part), even though there are close possible worlds where I have that same belief and it is false.
UNSAFE KNOWLEDGEIf the problem is that a belief can have different bases in different (but
close) possible worlds, then the solution is to require same-basis safety:
A belief that p by S is safe if and only if S would not believe that p on the same basis without it being so that p.4 My belief in SICK PATIENT does satisfy this revised safety requirement, and the requirement itself is initially intuitively plausible. Moreover, unlike sensitivity, it does not entail that I do not know that the trash bag is in the basement in GARBAGE CHUTE, and it also does not entail that we do not know that skeptical scenarios do not obtain.5 Despite it having these virtues, I will argue that the safety requirement is incorrect. I will present a case where a subject knows that p and yet the subject would easily have believed that p on the same basis on which he actually believes it without it being so that p.
3. A COUNTEREXAMPLE TO SAFETY
The case is the following:
HALLOWEEN PARTY: There is a Halloween party at Andy’s house, and I am invited.
Andy’s house is very difﬁcult to ﬁnd, so he hires Judy to stand at a crossroads and direct people towards the house (Judy’s job is to tell people that the party is at the house down the left road). Unbeknownst to me, Andy doesn’t want Michael to go to the party, so he also tells Judy that if she sees Michael she should tell him the same thing she tells everybody else (that the party is at the house down the left road), but she should immediately phone Andy so that the party can be moved to Adam’s house, which is down the right road. I seriously consider disguising myself as Michael, but at the last moment I don’t. When I get to the crossroads, I ask Judy where the party is, and she tells me that it is down the left road.
In this case, after I talk to Judy I know that the party is at the house down the left road, and yet it could very easily have happened that I had the same belief on the same basis (Judy’s testimony) without it being so that the belief was true. That is, in this case I know that p but my belief that p is not safe – I have unsafe knowledge.
Sosa (2002) presents a modiﬁcation of the safety condition that might be thought to help with HALLOWEEN PARTY. Sosa says that a basis can be safely related to a certain fact p not directly but dependently on a certain condition. In Sosa’s terminology, belief sources issue “indications” that certain facts obtain. An indication that p, I(p), “indicates the truth outright” if and only if I(p) would be so only if p were so; and I(p) indicates the truth dependently on a condition C if and only if (i) I(p) doesn’t indicate the truth outright, (ii) C obtains, and (iii) C and I(p) would jointly be so only if p 398 ˜
JUAN COMESANAwere so. The safety-related necessary condition for knowledge is, then, the
S knows that p on the basis of an indication I(p) only if either (a) I(p) indicates the truth outright and S accepts that indication as such outright, or (b) for some condition C I(p) indicates the truth dependently on C and S accepts that indication as such not outright but guided by C (so that S accepts the indication as such on the basis of C).6 In HALLOWEEN PARTY, the indication is the fact that Judy tells me that the party is at the house down the left road. Does that testimony indicate the truth either outright or dependently on some condition that guides my belief? It is clear that Judy’s testimony doesn’t indicate the truth outright.
Again, it could easily have happened that Judy said that the party is at the house down the left road without it being so that the party was at the house down the left road. But what about dependent indication? Isn’t there a condition C such that Judy’s testimony indicates the truth dependently on C? There obviously is: that condition is the fact that the subject that Judy is talking to doesn’t look like Michael to her. It could not easily have happened that Judy said that the party is at the house down the left road to someone that doesn’t look like Michael to her without it being so that the party is at the house down the left road. Judy’s testimony, then, indicates the truth dependently on the condition that the subject she is talking to doesn’t look like Michael to her.
But this fact doesn’t save the safety condition from being refuted by HALLOWEEN PARTY, for clause (b) of the necessary condition is not just that there be a condition C such that the basis of the belief indicates the truth dependently on C,7 but in addition my belief has to be guided by the presence of C. And in HALLOWEEN PARTY I am unaware of
the relevance of the respective condition to the truth of Judy’s testimony:
I would have believed that p whether or not I looked like Michael to Judy. Therefore, HALLOWEEN PARTY is a counterexample to the safety condition even taking into account dependent indication.
4. OBJECTIONS AND REPLIES
In this section I will consider three ways in which a defender of safety can reply to the challenge posed by HALLOWEEN PARTY. First, she may say that it is not a counterexample to safety because, contrary to what I said, my belief that the party is at the house down the left road is safe. Second, she may say that it is not a counterexample to safety because, contrary to what I said, I do not know that the party is at the house down the left road.
And third, she may say that it is not a counterexample to safety because,
UNSAFE KNOWLEDGEalthough I do have knowledge, I don’t have the kind of knowledge that requires my belief to be safe. Let’s take those possible objetions in order.
First, then, a defender of safety could say that my belief that the party is at the house down the left road is safe after all. This reply seems to me the most plausible of the three that I will consider, but I think that it is ultimately unsatisfactory.
It is clear, to begin with, that in HALLOWEEN PARTY my belief does not satisfy Sosa’s deﬁnition of safety: it could easily have happened that I had the same belief on the same basis and yet the belief was false. And, given that safety is a technical notion introduced by Sosa as a necessary condition on knowledge, we cannot retreat to a pre-theoretic notion of “safety” and claim that the example doesn’t touch it. It is of course true that my belief has something epistemically good going for it – it is, after all, a piece of knowledge. We can, if we want, call that something epistemically good that it has going for it “safety”. But what we do not have if we do this is a theory of what safety amounts to. Sosa tries to provide such a theory, and claims that for a belief that p to be safe is for it to have a speciﬁc modal relation to the fact that p. Maybe someone else will provide a different theory of what safety amounts to.8 HALLOWEEN PARTY is advanced as a counterexample to Sosa’s proposed deﬁnition, of course, not to the platitude that a proposition that amounts to knowledge has something good going for it, nor to an alternative deﬁnition that doesn’t yet exist.
A more speciﬁc defense of the safety condition appeals to the fact that safety is time-sensitive. Some years ago, it was a close possibility that Brian would become a lawyer; now, however, it no longer is a close possibility: Brian is safely not a lawyer. Similarly, the defender of the safety condition for knowledge might say, before I decided not to dress up as Michael, it was a close possibility that I would falsely believe that the party is at the house down the left road; now, however, it no longer is a close possibility: I safely believe the truth of the matter.9 Again, it is clear that my belief in HALLOWEEN PARTY does not satisfy Sosa’s deﬁnition of “safety”, and so, if it is indeed the case that my belief in that case is safe according to a time-sensitive notion of safety, then Sosa’s notion of safety is not time-sensitive. But even leaving that aside, it seems to me simply false that, in HALLOWEEN PARTY, after I decide not to dress up as Michael it is no longer a close possibility that I have a false belief. When considering whether the proposition that p obtains safely at t in the actual world, we consider whether it obtains in possible worlds that differ from the actual world just slightly right before t.10 And, in HALLOWEEN PARTY, I seriously consider dressing myself up as Michael just before driving to the intersection where Judy is standing.