I wrote the below back when the knowledge topic dealt with the regress argument for justification (the very first semester I took on teaching GET1029, by the way). It also touches on issues relating to how we respond to skepticism. Since a couple of you asked, I’ll make it available here for those interested.

The time stamps refer to the recording for the lecture (“L09”) and I’ve made the relevant part available here. Note that the first 40:15 (which is similar to what you saw for the first part of W09) was removed, so you need to made adjustments to the timestamps.

* * * * *

As I emphasized near the beginning of the lecture (around 7:10-8:20 in the L09 recording), there’s really a lot of stuff in the topic, but my aims are more modest than they might appear. The main thing to understand is how the Regress Argument for Skepticism works. This includes having a basic understanding of such items as knowledge, justification (including the parallels with ethics) so as to appreciate what the argument is even about, how the trilemma is generated, and why each of the three horns of the trilemma is at least seemingly unpalatable. Finally, how the rejection of each horn is associated with a position (or two) on justification (i.e., Foundationalism, Contextualism, Coherentism, Infinitism). But as I explained in lecture, each of them is really a whole topic of its own. For the purposes of the class, I am only holding you responsible for a basic grasp of what each position involves and what might be some pros and cons for each.

In this post, I want to supplement some of the things I discussed in the lecture, and in the process, address a few questions that were raised by students after the class. Specifically, since the topic centers on the Regress Argument for Skepticism, it would be a good idea for me to say something about what the argument is supposed to achieve. At this point, the attentive reader should wonder: “What the argument is supposed to achieve? Isn’t it already stated–it’s an argument for skepticism, specifically, skepticism regarding justification. Is there an issue?” Actually, yes. It turns out that arguments for skepticism are not always what they seem. There are two aspects to this. First, what’s the skeptic really doing when proposing such an argument? Second, what’s the point of the rest of us (who are probably non-skeptics) considering such an argument? Let’s consider each in turn.

* * * * *

Suppose you are a skeptic and you proposed the Regress Argument for Skepticism. It would seem like a most obvious thing to do–if the argument is sound, you would have secured a good reason for your skepticism. And if others can be persuaded as well, you would have managed to give them a good reason to be skeptics as well. But the picture is problematic because precisely if the argument is sound, none of this should be happening. If justification is impossible, then ex hypothesis, it is not possible to justify skepticism as well–whether to self or others. Seen in this light, the Regress Argument is self-refuting: it is wrong exactly when it is right! That is, to the extent that it is successful as an argument for skepticism, it also successfully shows that justification is possible, and hence, it’s own conclusion is false! (More than one student raised this point after the class–good catch!)

There is, of course, an obvious way out for the skeptic proposer of the Regress Argument–qualify the conclusion to say that justification is impossible except for the case of the skeptical conclusion itself. Make an exception. This is not an incoherent move. When the Sergeant shouts “Everyone keep quiet!” there is no self-refutation worth pointing out even if you notice that he obviously wasn’t keeping quiet. He meant that everyone else should keep quiet. So the skeptic could plausibly argue that his conclusion is not meant to be reflexively applied to, well, his own argument. With the exception of the skeptical argument, justification is impossible. In the ancient world of Sextus and his friends, there’s even a name for this position: “Dogmatic Skepticism” (dogma here just means a teaching that can be known).

Now, not every skeptical philosopher like Dogmatic Skepticism. Sextus and friends, in particular, thought that it was a cop-out. After all, it’s not clear what’s so special about the skeptical argument as an argument. If this one is ok, in principle, why can’t others of like nature be accepted too? In contrast, imagine a different way to understand how the Regress Argument works. Rather than take it as establishing a conclusion from premises, see it as functioning in a more indirect, dialectical way. It’s as if the skeptic is in a conversation with a non-skeptic–pressing him to explain why he is so confident that justification and knowledge is possible, and using what he said against him. So you think justification is possible? No? So, if a belief is justified, what are our options? Can the chain of reasons go on forever? Can it go in a circle? Can it just terminate? (Wouldn’t that be arbitrary?) Basically giving the non-skeptic rope to hang himself. But if the non-skeptic turns around and ask–so what do you think? The skeptic can reply: unlike you, sir, I have absolutely no idea–I take suspension of judgement seriously, you know…

And these aren’t the only possibilities as well. But suffice it to say that not every interpretation of the Regress Argument for Skepticism will make it vulnerable to a charge of self-refutation.

* * * * *

So much for the skeptic. But what what’s the point of the rest of us (who are probably non-skeptics) considering such a skeptical argument? As already hinted at when I introduced Foundationalism, the Regress Argument wasn’t just deployed for a skeptical conclusion. You can find essentially the same argument employed by Aristotle (Posterior Analytics I.3) to show that since chains of reasons can’t go in a loop, nor can they be infinite in length, there must be some beliefs that are justified without being justified by a further justified belief. And if at least some of our beliefs are justified in this way, and since we can’t reject Premises 3 (i.e., Coherentism is not going to work) and 4 (Infinitism is nutty), we have to reject Premise 2 (therefore, Foundationalism). Or at the very least, the conditional claim that if at least some of our beliefs are justified, then Foundationalism is true.

In fact, we can generalize that conclusion even further. Granted that the Regress Argument is valid, then one of the following three conclusions can be drawn from it.

If at least some of our beliefs are justified, then if neither Coherentism nor Infinitism are acceptable, then Foundationalism is true.

If at least some of our beliefs are justified, then if neither Foundationalism nor Infinitism are acceptable, then Coherentism is true.

If at least some of our beliefs are justified, then if neither Coherentism nor Foundationalism are acceptable, then Infinitism is true.

Or, if you want to be even more general:

If at least some of our beliefs are justified, then if the chain of reason supporting a justified belief can’t go in a loop nor go on without end, then some beliefs are justified without further justification.

If at least some of our beliefs are justified, then if no beliefs are justified without further justification and the chain of reason supporting a justified belief can’t go on without end, then some beliefs are justified even though the chains of reason supporting them go in a loop.

If at least some of our beliefs are justified, then if the chain of reason supporting a justified belief can’t go in a loop and no beliefs are justified without further justification, then some beliefs are justified even though the chains of reason supporting them go on without end.

The more general formulation allows for multiple ways to understand what it means for there to be some beliefs that are justified without being justified by a further justified belief (e.g., “Foundationalism” vs. “Contextualism”, and other possibilities as well). In other words, what the Regress Argument does is to–powerfully and compellingly–show us what our options are, as far as an acceptable account of justification goes. And this is potentially a very useful insight even for the non-skeptic. Maybe even especially for the non-skeptic philosopher concerned to understand the nature of justification and knowledge.

* * * * *

This doesn’t mean that the ‘Foundationalist solution’ is costless though. So let’s say our Foundationalist steps forward and proposes that the Regress Argument can be answered. Some beliefs are regress stopping. They are justified without being justified by a further justified belief. More generally, the doctrine says:

(F1) Every justified belief is either basic (justified without appeal to other beliefs), or non-basic (justified by appeal to other beliefs, ultimately, basic beliefs); and at least some beliefs are basic.

But of course, that’s not nearly enough–some account of what sorts of beliefs are basic would be needed. Whatever he says, he will be committed to the following:

(F2) A justified belief is basic if and only if it possesses the feature X.

Where F is whatever favored feature that makes a belief basic (Note: X could be a disjunctive set of features; i.e., “either X1 or X2 or X3”.) Let’s label the conjunction of (F1) and (F2) as (FF). But now, the critic is going to point out, it seems that we are justified to lay down a certain belief–let’s say p–as basic only if we have accepted an argument like the following:

(1) Belief p has feature F.

(2) All beliefs that have feature F are justified without further justification. (From (F2))

Therefore: (3) Belief p is justified without further justification.

It doesn’t look as if the purportedly basic belief is just basic. It is acceptable as basic only if we have accepted the meta-justification which shows that it has the appropriate feature marking it out as basic. If this is so, then it looks as if it is basic only if its basicality can be inferred!

There’s more though. What about (2) in the argument above? It is a derivation of (F2), which is a part of the overall package (FF). But what reason do we have to accept either (F2) or the whole package (FF)? It doesn’t seem as if we have stopped the regress completely by adopting Foundationalism.  We might have just pushed the problem back. After all, (FF) and its components don’t look like one of the candidates usually proposed for basic beliefs in (FF)–e.g., self-evident truths of reason, truths evident to the senses, memory, etc. So maybe it’s not basic. But if (FF) is not basic, then, even according to (FF) itself, we need further justification for it. In other words, if (FF), then (FF) is justified by inference from other beliefs (among which (FF) is not included). And if this is so, we are now faced with the trilemma again–and the regress has been restarted.

(The above needn’t spell the end of Foundationalism though; but rescuing it will take a longer story.)

* * * * *

A final bit about Coherentism. Consider the line in L08 Slide #27: “The larger the body of beliefs, the less narrow coherence seems feasible as a gauge of justification.” To explain this, I need to tell you about this thing call the Preface Paradox. Now imagine that you are writing a book. Your ambition is to have it become the new standard for your field of research. And it’s a very long book (the editors are already saying that it’s best pitched as a book series, rather than a single volume). As you complete your (now multi-volume) draft, you set about checking and re-checking every assertion you make in the draft.

For each of the claims you make, you form the justified belief that it is true. That is, you have the justified belief that Claim 1 is true, the justified belief that Claim 2 is true, and so on… and the justified belief that Claim n is true. Normally, you would have thought that if you are justified to believe p, and you are the justified to believe q, then you are justified to believe (p & q). (Philosophers call this the Principle of Agglomeration.) So:

(1) You are justified to believe (Claim 1 is true and Claim 2 is true and  … Claim n is true).

But at the same time, you know full well that you are human, all too human. In other words, you are fully aware that human beings make mistakes, and therefore, you are very likely to have made at least one mistake. So you are justified to believe that there is at least one mistaken claim in your book. In other words:

(2) You are justified to believe (Claim 1 is false, or Claim 2 is false, or … Claim n is false), which is just:

(2′) You are justified to believe it is false that (Claim 1 is true and Claim 2 is true and  … Claim n is true).

With (1) and (2)/(2′), and Agglomeration again, you are now justified to believe a inconsistent pair of propositions! And more to the point for us, the moral of the story is that it seems entirely possible for us to be justified to believe something, even while being entirely aware that it belongs in an inconsistent set of propositions. Which doesn’t speak well for Coherentism as a doctrine about justification since it says that it is membership in a coherent body of beliefs that makes a belief justified.

(The Preface Paradox is so called by allusion to the what an author often says in the Preface of his book, especially an academic book, i.e., “any errors that remain are my sole responsibility” or some such. Though I have seen one in which the author says that any remaining errors are solely due to his friends to whom he circulated the draft beforehand…)

At this point, the alert reader may wonder: if the above reasoning is correct, why should we care about inconsistency? The answer may be that scale matters. When an inconsistency is spread over a very large body of propositions, it seems to post less of a worry, than when it is concentrated in a small number of propositions (think LPOE as Mackie intended it)–and in the limiting case, in a direct contradiction. But this raises an implication for Coherentism as well. Normally, you might have thought that if a belief is justified if it belongs in a coherent body of beliefs, then the strength of the justification would be increased precisely if we are talking about a larger and more varied body of beliefs. But with the Preface Paradox, it may well push us in a different direction–because the larger the body of beliefs, the less likely consistency can be perfectly maintained. So in one sense, the larger the body of beliefs, the less likely that it will be fully coherent.

Do the above considerations refute Coherentism? It’s really unclear. But they certainly suggests that Coherentists have lots of work to do to explain what the “coherence” of a body of beliefs amounts to.