Reliabilism and the New Evil Demon Problem
Ali Sarvi

Suppose that there is a possible world in which all the inhabitants are the Brains In a Vat (BIV) who are the victims of an evil demon; BIV’s experiences and beliefs are the same as our beliefs and experiences in the actual world; we both are thinking in the same way and forming an analogous belief on the basis of an analogous experience.
From this consideration, we can reasonably infer the New Evil Demon thesis (NED) as follows:
(NED): The beliefs of the brains in a vat are as justified as our beliefs.

Now consider the general formulation of reliabilism as follows:
(R): S’s belief in P is justified if and only if S’s belief in P is caused by a reliable process.

(NED) raises a problem for reliabilism (R) which has been called the new evil demon problem. Here is the argument which shows the problem:

NED Argument:
1- Our beliefs in P are justified.
2- The beliefs of the brains in a vat are as justified as our beliefs (NED).
3- If (2) is true, then if (1) is true, then BIV’s belief in P is justified.
4- if (1) is true, then BIV’s belief in P is justified.
5- BIV’s belief in P is justified.
6- S’s belief in P is justified if and only if S’s belief in P is caused by a reliable process (R)
7- If (6) is true, then if (5) is true, then BIV’s belief in P is caused by a reliable process.
8- If (5) is true, then BIV’s belief in P is caused by a reliable process
9- So, BIV’s belief in P is caused by a reliable process

The conclusion of the argument is false, although the argument is valid; so some of the premises might be false. Let us suppose that we are not sceptics and so we will not reject premise 1. And also suppose that our above-mentioned consideration of the possible world of the brains in a vat plausibly implies that (NED) is true and so premise 2 is not rejected; although some may object to premise 2 in some other senses, but for our purpose we do not consider those objections. Hence, the only remained option to reject would be premise 6, i.e., (R).
Juan Comesana (2002)[1] wants to defend reliabilism and his strategy is to indicate that the (NED) argument is fallacious. He emphasizes that we should note that (R) has two different possible readings as follows:

R1: S’s belief in P is justified if and only if S’s belief in P is caused by a process which is reliable in the actual world.
R2: S’s belief in P is justified if and only if S’s belief in P is caused by a process which is reliable in the world where it is used.
Then Comessana argues that if we reallege the above (NED) argument by means of R1 as premise 6, then the conclusion would be:
9*- So, BIV’s belief in P is caused by a process which is reliable in the actual world. Now it is obvious that 9* is true; because it asserts that the processes which BIV use in their world are reliable processes in the actual world, although they are unreliable in their world. For example a BIV uses vision perception to make her belief; and vision perception is a reliable process in the actual world since most of whose output would be true in the actual world. So, if we consider R1 as the appropriate reading of R, then NED argument is sound and the new evil problem for reliabilism would be perished.
On the other hand, if we consider R2 as the appropriate reading of R as premise 6, then the conclusion would be as follows:
9**: So, BIV’s belief in P is caused by a process which is reliable in the BIV’s world.
It is obvious that 9** is false, because it asserts that the processes which BIV use in their world are reliable processes in the BIV’s world, but they are not reliable in that world since most of whose output would be false in the BIV’s world. Then Comesana suggests that since there are two different readings of R in premise 6 in the NED argument, so the argument is a fallacy of equivocation and hence the problem of the new evil demon is refuted.
Here is Comesana argument:

1- (R) is used in two different meanings in premise 6 of the NED argument.
2- If (1), then NED argument is a fallacy of equivocation.
3- Then NED argument is a fallacy of equivocation.

I think there is an objection to premise 1. Comesana declares that when a speaker utters that “process F is reliable”, her utterance has idenxical feature just as when she utters:“it is raining here”. The proposition of sentence “it is raining here” is true, for instance, in Winnipeg, but is false in Tehran (if it is not raining in Tehran). But we should note that although the proposition of sentence “it is raining here” has two different truth-value in two different places (i.e. true in Winnipeg and false in Tehran), it has just one and only one truth value in Winnipeg. Similarly, when S says “process F is reliable”, one utterance of it can be true, for instance, in the actual world where it conveys that “process F is reliable in the actual world”, but another utterance of it could be false in world w where it conveys “process F is reliable in world w”. But again we should notice that although the proposition of sentence “Process F is reliable” has two different truth-value in two different possible worlds (i.e. true in the actual world and false in the BIV’s world), it has just one and only one truth value in the actual world.
Therefore, we should say that (R) has just one and only one truth-value in the actual world, whereas Comesana says that (R) has two different truth value in the actual world.
Now we can allege the following argument against Comesana:

1- If (R) does not have two different readings with two different truth value in the actual world, then the NED argument is not a fallacy of equivocation.
2- (R) does not have two different readings with two different truth value in the actual world.
3- So, the NED argument is not a fallacy of equivocation.

But something is interesting here: If the only utterance of (R) in the actual world is R1 (S’s belief in P is justified if and only if S’s belief in P is caused by a process which is reliable in the actual world), then the conclusion of the NED argument is 9*:
9*: So, BIV’s belief in P is caused by a process which is reliable in the actual world
And, as it was mentioned earlier, 9* is true and so it seems that the NED argument is sound and we can say that the problem of NED is refuted.
But the problem is that the NED argument with R1 is not sound because R1 is simply false; R1 is false because there is some counterexample against it. Laurence BonJour’s example of clairvoyance is a counterexample for R1; because although clairvoyance is a reliable process for the clairvoyant to make her true beliefs, she is not justified in her beliefs.

















[1] “The Diagonal and the Demon”, Philosophical Studies 110 (3): 249-266.