Tuesday, 2 January 2007

Un-discriminating Reliabilism (Part 1)

J-reliabilism is perhaps the most well-known, if not most widely held, response to Gettier’s criticism of the JTB account of knowledge. However, in this post (as well as in the one to follow), I argue that J-reliabilism is insufficient for responding to Gettier. Furthermore, I argue that what is really doing the work in classical formulations of reliabilism, such as that found in [Goldman 1976], is the discrimination requirement, or (DR):
(DR) For any subject S, if S knows that p then S can distinguish the actual situation in which p from all relevant or nearby alternative situations in which ~p.
Recall, J-reliabilism amounts to following claim:
(J-Rel) For any subject S, S’s belief that p is justified IFF it was formed via a reliable process (i.e., a process that tends to produce true beliefs).
According to (J-Rel), reliability is what makes a belief justified. (see [Goldman 1979]). By defining justification in terms of reliability, J-reliabilists hope to eliminate that element of luck that allows our beliefs, under the traditional JTB account, to be Gettiered. However, it is not clear that (J-Rel) is sufficient for resisting Gettier type cases. For example, consider the following perceptual Gettier case:

Suppose S has strong perceptual evidence for, and comes to believe, the proposition:
(a) There is a red cube in the box on the table.
Now, it so happens that there is in fact a red cube in the box on the table, though the cube is being obscured from S’s visual field by some sort of barrier. Furthermore, the box is rigged up to a computer which projects a visual hologram of a red cube in the box. However, the computer is programmed to only project the hologram of the red cube in the box when there is a real red cube in the box. Moreover, S lacks any of this background information, and forms her belief that (a) purely on the basis of the hologram of the red cube. All of the following seem true in the above case:
(i) (a) is true
(ii) S believes (a) is true
(iii) S’s belief that (a) is formed via a reliable process
Ex hypothesi, (iii) is true since the computer is programmed to only project the hologram of a red cube when there is an actual red cube present (one may build in whatever stipulations one likes, such as that the computer is eternal and infallible in its operation etc.). Thus, S’s belief that there is a red cube in the box is reliable since the process by which the belief was formed would, given the computer’s programming, tend to produce true beliefs. However, I believe this represents a bona fide Gettier case since, though S has a justified (i.e., reliably formed) true belief, we wouldn’t say that she has knowledge.

I take the above Gettier case to show that mere reliability is insufficient for eliminating the element of luck from S’s belief that (a). To wit, (J-Rel) fails to eliminate the element of luck associated with S’s belief that (a) since the fact that her belief is reliable is (from the subject’s perspective) itself merely a matter of luck. Thus, we may say that it is lucky that S’s belief was formed via a reliable process. Significantly, the particular species of epistemic luck here described is easily eliminated by making the reliability of the belief forming process internally available to S. Thus, if S knew that what she was seeing was merely a hologram, but that the hologram was itself reliably linked to the presence of an actual red cube in the box, then our intuitions would grant that S does in fact know that there is a red cube in the box. Thus, I see the above Gettier case as corroborating the internalist intuition that that which justifies S’s beliefs should be internally available to her.

Notice that I haven’t made any references to S being able to discriminate between actual red cubes and holograms of red cubes, nor to definitions of reliability that appeal to relevant alternatives (See [Goldman 1986]). This failure to invoke (DR) is in keeping with my ultimate goal of showing that what is doing the real work in classic reliabilist accounts is not reliability qua reliability, but rather an implicit commitment to (DR). Above, I have argued that reliability alone [i.e., (J-Rel)] is insufficient for responding to Gettier. In my next post, I argue that the mistaken assumption that (J-Rel) and (DR) must go together has misled reliability advocates to impute to J-reliabilism an explanatory power that it actually lacks.


Goldman, A. (1967), 'A Causal Theory of Knowing', Journal of Philosophy 64: 355-372.

Goldman, A. (1976), 'Discrimination and Perceptual Knowledge', Journal of Philosophy 73: 771-791.

Goldman, A. (1979), 'What is Justified Belief?' Justification and Knowledge, Dordrecht: Reidel.


ABD said...

Nice blog; lots of interesting stuff here.

I don't share your intuition that this isn't a case of knowledge (and hence, provided KP --> JBP) one of justified belief (I would say that she does have knowledge of (a), though maybe not of various closely related propositions involving perceptual demonstratives). But maybe you're right -- cases involving veridical illusions can elicit conflicting intuitions, let alone cases involving reliably produced veridical illusions. Still, supposing you're right, I don't see how this:

"the fact that her belief is reliable is (from the subject’s perspective) itself merely a matter of luck"

explains what's wrong. First, how is it that "from the subject's perspective" the fact that her belief is reliable is just luck? I thought she didn't have access to any of the relevant background information. Second, it seems dubious to me in general to say that if it's just luck that a belief is reliably produced then it's not knowledge. Suppose I had extremely reliable vision, but that this fact about me was a fluke -- I come from a planet with a zillion other creatures with poorly designed unreliable visual systems; somehow I came out working okay. I'm lucky, and it's a matter of luck that my visually-produced beliefs are reliable. But (it seems to me) I can still come to know things just fine by using my visual system. Or maybe this isn't the kind of luck you have in mind.


Let me begin by officially congratulating you on being the first person to post a comment on my blog. (I wish I had a ribbon or something to give you.) And what a comment it is! You have a wealth of great insights here. Thanks.

I can't impugn your contrary intuitions, though I probably should mention that I tested out this example on three people (all non-philosophers) and they all agreed that the subject in my example does not know that (a). My extensive sample size notwithstanding, it may very well be (as you point out) that our intuitions tend to differ in these types of cases. However, one can't help but wonder just how much our conflicting intuitions as philosophers regarding such cases issues from our own prior theoretical commitments. For example, I'll be willing to bet money (well, maybe not) that there would be greater consensus on this matter among individuals who have not been exposed to formal theories like reliabilism etc. Nevertheless, I would genuinely be interested to hear what other reader's intuitions are, whether philosophers or non-philosophers. What say ye?

Intuitions aside, I grant that the locution 'from the subject's perspective' is a bit ambiguous. It could be interpreted as suggesting that the subject views herself as being lucky that her belief happens to be true. However, it could also be read as saying that given all that the subject has subjectively available to her, we (the all-seeing observers) can say that the fact that her belief is true, is really a matter of luck. I had something along the lines of the second interpretation in mind. But what I find most telling about the particular set-up (if it works) is that since the belief was produced by a reliable process, the luck has nothing to do with external factors regarding how the belief was formed. Rather, the fault lies somewhere 'within' the subject, or at least that's the intuition I'm attempting to get at. For example, a virtue reliabilist would say that the problem is that the reliability fails to arise out of the cognitive character of the subject. (I am not advocating such a position, I'm merely presenting one example of a way this worry may be articulated).

Now, with regards to your example of a reliable subject surrounded by unreliable subjects: Given that it is merely a 'fluke' that your visual mechanisms are reliable, the possibility that they are in fact unreliable remains a relevant alternative. Now there seems to be two possibilities on this score: either you're aware that the other inhabitants of your planet are unreliable or you're not. On the first option, I would suggest that you don't have knowledge unless you are also aware (i.e., have introspective access to the fact) that your visual mechanisms, unlike those of your neighbours, are in fact reliable. Otherwise, your knowledge of the general unreliability of the other inhabitants would 'undermine' (a la Goldman) the de facto reliability of your own visual mechanisms. On the second option, in which you are unaware that your neighbours are unreliable, I believe our intuitions would be analogous to those of Henry in fake barn country looking at a real barn. Just as it is a fluke that Henry happens to be looking at a real barn, it is a fluke that your visual mechanisms happen to be reliable. In both cases, the respective subjects lack knowledge. I'm deliberately being laconic here in order to avoid making this comment much longer than it already is. But I'll be happy to flesh out this final point more (think local vs. global reliability) if you like. Just ask. (And that, boys and girls, is what we call ensuring blog security.)

Ang said...


I think S knows the cube is in the box. This is because I think S sees the cube.

Your hologram case sounds very similar to cases of prosthetic vision. Consider the following. Suppose S has lost his eyes, but a video camera is mounted on his head and hooked up to his optic nerve. This camera, however, picks up radiation that is unaffected by the box (the barrier between S and the cube). Does S see the cube? It seems he does. And it seems he knows the cube is in the box too.

I think the key concern is how the hologram is being generated. If the hologram that matches the cube is generated *because* of the cube that is in the box (e.g., special video cameras pick up radiation from the cube through the box), it seems you have described a case of a sort of "prosthetic" vision.

I analyze a number of cases using holograms in my paper on perception, available at my website.

Take a look, in particular, at section 2, pages 12-16.



Thanks for responding to my invitation to provide feedback on my proposed Gettier case.

You appear to be making two separate claims here: an epistemic claim (i.e., S knows there is a cube in the box) and a perceptual claim (i.e., S sees the cube).

I will begin with the perceptual claim: I am willing to grant that the subject in the case of prosthetic vision does in fact see the cube. However, I believe there are a number of striking dis-analogies between your example and my proposed Gettier case. Moreover, I believe the differences are such to render the claim “S sees the red cube” tendentious in the latter case though not in the former. For starters, the subject in the prosthetic case can refer to the cube demonstratively while in my proposed Gettier case, she obviously cannot. (For those uncomfortable with talk of “demonstratives”, just substitute, “spatially pick out”.) In fact, if the Gettier subject were to attempt to refer to the cube demonstratively, she would most assuredly get things wrong since she would point to the hologram and not to the actual red cube (which ex hypothesi is hidden from view). Furthermore, it seems tendentious to describe someone as seeing an object who is unable to refer to that object demonstratively. This, of course is not a problem for your prosthetic vision case.

But that’s just the beginning. Notice that in the prosthetic case, whether or not the subject is aware that her eyes were replaced by a camera makes no difference to whether or not we regard her as “seeing” the red cube. Now imagine that the subject in my proposed Gettier case came to know all the background details of the set up. Thus, the subject learns that what she is looking at is actually a hologram, but also learns that the hologram is causally linked to a concealed red cube. It may be agreed on all sides that the subject in my re-described case knows there is a red cube in the box. But the question is, does she see the red cube? If asked, she would most assuredly say that what she sees is not the red cube, but rather a hologram. Moreover, this also seems to be what we (as observers) would say about the subject, if we knew that she were aware that she was actually looking at a hologram. Admittedly, since she knows that the hologram is causally linked to a real actual cube, she is now in a position to infer that there is a red cube somewhere in the box. But neither she nor a third person observer would say she is “seeing” the red cube. Assuming that all I’ve said so far is right, it seems highly implausible to suggest that whether a subject counts as “seeing” can be undermined simply by her becoming aware of the details of her perceptual situation. Thus, if she fails to see the cube when she knows she is looking at a hologram, then she also fails to see the cube when she does not know that she is looking at a hologram. Notice that this is only a problem for the Gettier subject and not for the prosthetic subject.

Finally, I should point out a faulty assumption in your argument – namely, that causal dependence (your *because*) is sufficient for vision. However, there are any number of deviant causal chains which do not live up to our putative standards of what constitutes a bona fide case of vision. (See my post Looking for a Causal Criterion for Vision) I believe my Gettier case is a paradigm example of such a deviant causal chain. Your prosthetic case, by contrast, hardly seems to qualify as a deviant causal chain since it simply involves substituting elements (i.e., “cameras” for “eyes” etc.) along what is pretty much the standard causal pathway for vision. This, then, amounts to a third dis-analogy between the two cases.

Given these striking dis-analogies between your prosthetic vision case and my proposed Gettier case, the fact that the subject counts as “seeing” in the former seems insufficient for motivating the claim that she also counts as “seeing” in the latter.

Now, I turn to your epistemic claim. As with the perceptual claim, I am willing to grant that in the case of "prosthetic vision" you describe, the subject does have knowledge. However, I believe there is an (epistemic) difference between the prosthetic and Gettier cases, and it is a difference that gets to very heart of the point I'm driving at. In short, the Gettier subject only has local reliability [i.e., (J-Rel)], while the prosthetic subject has global reliability [i.e., (J-Rel) + (DR) inter alia). That is, given your description of prosthetic vision, we have no reason to believe that the subject goes around mistaking holograms for real cubes. In fact, far from suffering from any such deficiency, your subject has the added ability to see through black boxes!

The Gettier subject, however, clearly lacks the ability to discriminate between actual red cubes and holograms of red cubes. Admittedly, in the present (i.e., local) case, the subject's belief that (a) happens to be true, and it could not have easily been false. Thus, the subject possesses local reliability. However, her local reliability does not translate into the type of global reliability that would prevent her from being deceived by holograms of red cubes in general. (Again, we may say that it is merely a matter of luck that her belief that (a) is reliably formed, read locally reliable.) I believe it is this fact that lies behind the "common sense" intuition (or at least what I claim to be such) that my Gettier subject lacks knowledge.

In sum, if your example is to effectively mirror my own, then the subject in question must possess visual abilities that are only locally reliable. Unfortunately, the case of prosthetic vision does not seem to fit the bill.

Ang said...

You make a number of points in your reply. To keep this comment short, I will respond to the points about perception. I will respond to your points about knowledge later.

I) I do not think that a mere causal relation is sufficient for vision. Although I see no reason that any of our concepts (including perception) should decompose neatly into necessary and sufficient conditions, I propose (in my paper on perception, p. 9) that in core cases where S perceives O, 1) O causes S’s sensory experience, 2) some counterfactual dependence with O must obtain along both the factual and
perspectival dimensions of S’s sensory experience, and 3) this counterfactual dependence must, at minimum, be robust enough to enable S to experience perceptual constancy along both dimensions. I get the distinction between "factual" and "perspectival" from Alva Noe.

II) It's not clear to me why you think the subject can't refer to the cube demonstratively (or "spatially pick it out") in your Gettier case. I would like to understand this better: A.D. Smith, for instance, thinks he can refute indirect realism in less than four pages by appealing to some points about demonstrative reference. I must admit, I found his argument wholly unconvincing. (See his The Problem of Perception, pp. 13-17.)

Consider this situation. Blindfold everyone in the room. So everything is "hidden" from everyone. But these are special blindfolds. On the outside of the blindfold is a video camera. On the inside is a display. The display reproduces what the video camera picks up. Turn on the apparatus: voila! everyone can see. Now everyone starts pointing to stuff, "demonstratively referring" to this chair, that table, etc.

It seems to me that this scenario poses no problem at all for demonstrative reference.

III) Now ask these people another question: are they looking at objects in the room, or are they really looking at the blindfold?

OK, I suppose if you press them, they'd have to confess they're looking at the blindfold. But there is a clearly unproblematic sense in which they are looking at objects in the room, right?

Ang said...

IV) How about this variation that tries to address your global/local distinction.

Suppose someone glues five paper-thin devices to the sides and the top of your box. The devices fit the box exactly. On the inside surface of each device is an array of cameras which pick up radiation from the red cube on the inside of the box. The outside surface *displays* what the cameras pick up.

Hopefully, all I have done here is to give a technical explanation of *how* the hologram of the red cube is being generated. So this is also a case of local (not global) reliability.

With the thought experiment so described, it seems to me that our subject sees the red cube. In fact, the devices seem straightforwardly to allow the subject to "see through" the otherwise opaque box surface.


My reply to both your perceptual and epistemic claims are closely related, so I'll respond to both in this comment:

The reason the subject in my Gettier case cannot refer to the cube demonstratively is because if she attempted to do so, she would point to the hologram and not the cube (which, ex hypothesi, is hidden from view). To make this explicit, let me exaggerate the scenario a bit: Imagine that the black box is the size of a room. In the middle of the room, the computer projects a hologram of a cube. The hologram is causally linked, via the computer, to a real cube which is hidden from view in the far right-hand corner of the room. Now, according to my Gettier case, the subject forms the belief, “there is a cube in the room”, based solely on the hologram located in the centre of room. Thus, if asked to point to the red cube, she would point to hologram. In short, she mistakes the hologram for the genuine article. Now, this does not undermine the truth of her existential claim, “there is a red cube in the room”, but it does seem to undermine the claim that she sees the red cube. If she does in fact “see” the red cube, then why doesn’t she point to the red cube? Why does she point instead to the hologram?

I think the mistake you’re making is that you’re trying to force the particular set-up of my Gettier case to fit your prosthetic vision model, when it does not. I’m not denying that you could modify my Gettier case sufficiently so that it becomes one of your prosthetic cases. But at that point, it would cease to be my Gettier case.

Your new case seems to suffer the same basic problem as the others. But before I respond I want to be clear that I find your prosthetic vision cases very interesting, particularly when put to the use you originally designed them for. But I’m afraid they simply don’t mirror my Gettier case enough to be of any real value here.

Consider your example with the box with the paper thin censors all around it, certainly ingenious in its own right. But notice that the end result of your set up is that the subject is allowed to straightforwardly “see through”, as you say, the otherwise opaque box surface. Presumably, this means that the subject in your example can, in at least some meaningful sense, refer to the cube demonstratively. However, as I argue above, my Gettier subject cannot. Given this crucially important dis-analogy between the two cases, the fact that the subject has knowledge in the one case provides little motivation for the claim that she also has knowledge in the other.