One of the big issues in epistemology is the problem of infinite regress. “I believe the sun will rise.” “How do you know that?” “Because it always has.” “How do you know that?” “Because my memory and human records confirm it has.” “How do you know that?” “Because I’ve examined those memories and records.” “How do you know that?” And so on. It looks like this could go on forever. It seems like any answer you give can be doubted. We can always keep asking “How do you know that?” And that isn’t the only line of regress. “I believe the sun will rise.” “How do you know that?” “Because it always has.” “How do you know something that’s always happened will continue to happen?” And so on.

The difference between those two lines of questioning is that the first is about the facts, while the second is about inferences, the question of which rules are valid when interpreting those facts. Every rule is doubtable, because exceptions might always be possible; and every fact is doubtable, because we could always be mistaken. Someone could always have made an error, or lied, or our memories could be inaccurate or false, and so on. Thus, the problem of regress is just this: Where is it reasonable to stop doubting, to stop asking questions? When should we just shut up and believe?

Warrant and Proper Function

Some of you may already recognize much of this article. The original is still on my old blog from twenty years ago. But I am rewriting it here, and substantially updating and improving it. Both because it needs a rewrite and because I am gradually salvaging the articles on my old blog that are most worth preserving, in anticipation of a possible deprecation of Blogger. It’s a Google property and though there is no particular reason to think they’ll abandon it anytime soon, my past experience with old platforms makes me wary. So I’m starting early and going slow. I’ve already rewritten my articles on defining the supernatural and the mathematical universe and host them here (as Defining Naturalism: The Definitive Account and All Godless Universes Are Mathematical), and several more were already rewritten into chapters in Hitler Homer Bible Christ. I have a list of my favorites I might work through (but not all of it).

Today I’m redoing an article I refer to a lot, because it often comes up. I won’t belabor its backstory, but just catch you up with some cliffsnotes. Alvin Plantinga wrote a book Warrant & Proper Function (among others) which tried to argue that Christians don’t need evidence to warrant believing something, while atheists can never have warranted belief. The latter crystalized into his Evolutionary Argument against Naturalism that was widely panned by philosophers as pseudoscientific and irrational. I’ve already refuted that (in Why Plantinga’s Tiger Is Pseudoscience and The Argument from Reason). It trades on the scientifically illiterate mistake of confusing sensory systems with learning and problem solving algorithms and what their evolved functions (and thus limitations and biases) are expected to be, and of confusing evolved with human-made learning and problem solving algorithms.

But that leaves the other point: is it rational to believe things without evidence? The answer is obviously no (see On Andrew Moon’s Defense of Circular Arguments and Which Is ‘Rational’: Theism or Atheism?). But if you reject that, you still have to provide a theory of rational warrant. I did that in Sense and Goodness without God (which is being rewritten and updated and will be published as A Better World without God next year) with my own theory of warrant and properly basic beliefs. This was then challenged by claiming it doesn’t escape the problem of infinite regress because reasoning depends on trusting our memory at every step of any line of reasoning, and there is no non-circular way to do that.

So what is Plantinga’s solution to this? To just assume Christian Theism is true, and that we are fully justified in assuming this without needing any evidence Christian Theism is true. “I don’t need a reason to believe it.” This is irrational. But Christians are not rational. Nevertheless, regress must end somewhere. So where is the rational end game? It is tautological that there must always be some fundamental “givens” that end an infinite regress of reasons to believe something, since that is literally what it means for regress to end: something is just believed for no reason—or the reason to believe it is itself. The mistake made here is to assume a false equivalence between these two options. But to believe something for no reason is to rest all your conclusions on something you have no reason to believe, entailing no reason to believe those conclusions. And you can’t just gainsay your way out of that.

There is clearly only one sound solution to epistemological regress: the end game is always something self-evident. That is, all beliefs rest ultimately on a bunch of things you believe because they are sufficient evidence for themselves and thus no further evidence is required, and thus no further regress. This is called Cartesian Knowledge: raw, uninterpreted, present experiences, which alone have a zero probability of not existing when they exist for an observer because “they exist for an observer” is what a raw, uninterpreted, present experience is. They are thus self-evident. I don’t need any extra other reason to believe I am having the experiences I am having right now. I can doubt past experiences. I can doubt whether my present experiences correspond to anything outside the theatre of my mind. But I can’t doubt the theatre is there and has the contents it does right now.

All other knowledge is built on top of that foundation (see Epistemology without Insurmountable Regress or Fallacious Circularity and Hypothesis: Only Those Who Don’t Really Understand Bayesianism Are Against It). As Keith Parsons pointed out to Plantinga in God and the Burden of Proof, if we get to just assume Christian Theism is true, then we could just as easily assume Great Pumpkinism is true (or Pastafarianism or Simulation Theory or Scientology). The reason we can’t just start with assumptions like those is because they are dangerously and irresponsibly arbitrary. Declaring Cartesian Demons as a basis for a belief is just as foolish as declaring them as a basis for doubt: of all logically possible things, without a good reason, the epistemic likelihood of you picking the correct thing is functionally zero (the complexity of assumptions in allowing there even to be such a demon is so great, yet with no evidence for any of them being true, that their compound logical probability approaches zero: see why We Are Probably Not in a Simulation and The Principle of Indifference).

And adopting a belief whose probability of being correct is effectively zero entails unacceptable risks of bad outcomes (see What’s the Harm? Why Religious Belief Is Always Bad and Christianity Is a Conspiracy Theory and Dear Christian: You Might Be Worshiping the Antichrist and Justin Brierley and the Folly of Christianity and Debunking John Davidson’s “Pagan” America and so on). You need to behave more responsibly than that.

Cartesian Knowledge

The foundation of rational belief can never be arbitrary guesses. Guesses can be hypotheses you test against the evidence, but never as presumptions by which you interpret the evidence. The foundation must be where all reasoning stops: the undeniables of direct present experience. Basic beliefs are things you believe without further justification and on top of which you build and justify all other beliefs. Properly basic beliefs are basic beliefs that you are justified in believing without further justification. Everything else is improper—and thus simply unjustified, and therefore irrational. Properly basic theism is irrational, because no such thing exists. Theism is always a derivative, never actually a basic belief, and even when treated as basic, can never be properly basic.

The fundamentals of experience go beyond just the five senses (there are actually dozens of senses), but include thoughts and feelings, everything you experience in any given moment when you are experiencing it. So, for example, an “interpretation” of an experience is itself a basic undeniable experience—because whether the interpretation is correct can be doubted, but not that you are experiencing it. And that you “experienced” it could be doubted, as your memory could be fake; but that you are experiencing a memory of it now cannot be doubted. And so on. Raw, uninterpreted experience is undeniable because it is logically impossible for it to be false, because for it to be false would mean that it does not exist, so when it self-evidently does exist, it would be a logical contradiction to say that it simultaneously does and does not exist.

Even illusionism accepts this point, whereby experiences are merely the belief that you are experiencing and no more real than that: the data still exist, and cannot not exist at the same time they exist (see What Does It Mean to Call Consciousness an Illusion?). Even an illusion exists—as an illusion. You can say a mirage is an illusion, meaning there is no water at the end of the road, but that you are seeing a reflective surface that looks like water is still happening. You cannot at the same time be experiencing that and not at all experiencing it—experiencing it is reductively what its existence is. Ultimately all logical laws derive from the singular law of identity, that a thing is what it is and not something else, or else it isn’t that thing in the first place (and therefore contradictions cannot exist, nor excluded middles), and that all derives from the undeniables of experience.

Hence all logic simply describes a singular law of physics: that distinctions exist (see The Ontology of Logic). “Experiencing P” and “Not experiencing P” is a distinction. And the existence of P as an experience is the distinction here being made. Therefore no other extra thing has to be the case for it to be true that P exists. Which means it can never be false that P exists when P exists. And hence our Cartesian knowledge is always known to a true with absolute 100% certainty. Everything else is a probability, and as such, needs external information to warrant believing it has that probability and not some other. This does not mean Cartesian knowledge requires no justification to believe it. Experiencing it is sufficient justification to believe it. The only difference is that nothing else is needed to justify that belief but the belief itself.

Properly Basic Belief

That means Cartesian knowledge is the only properly basic belief. To say something is “properly basic” is to declare that it’s something we get to assume without needing a reason to believe it—other than itself. We need another reason, at least some reason, to believe anything else. In fact anything that could be false requires a reason to believe it other than itself. Therefore only things that cannot be false can be properly basic. And that means, quite simply, Cartesian knowledge. It doesn’t even include logically necessary facts, because we can always be mistaken about those. As I explain in Proving History, even if a thousand mathematicians confirm a formal proof is correct, there is still some nonzero probability that they all made a mistake. And if we vet it ourselves, there is some nonzero probability that we will make a mistake. We can always be mistaken. Therefore we always need a reason to believe we are not mistaken. Except for undeniable present experience: because about that we can never be mistaken. Everything else is doubtable.

And so it is an inescapable fact that if there is any possibility a belief could be false, then we need some reason to believe it isn’t false. It needn’t be a weighty or elaborate or air-tight reason. Any genuine reason will do. But if we need even a tiny little reason to believe something before we are warranted in believing it then that belief cannot be called properly basic. And that’s that. Theism, therefore, can never be a properly basic belief.

For example, the fact that our thoughts and “interpretations” exist at the moment we experience them is undeniable, regardless of whether they are true or correct, and therefore our belief in the existence of those thoughts and interpretations is properly basic. Likewise, it can also be undeniable that there exists at this moment an experience of our “interpretations” cohering well—or not cohering well—with everything we are experiencing at the same moment. Obviously, just because we are experiencing an interpretation of the facts that is cohering well with everything else going on doesn’t mean it is cohering well. We could be in error about that. Nor does such coherence mean our interpretation is true. Since there may be countless explanations of the same facts that are equally coherent. But the fact that we are experiencing that coherence is undeniable. Since it cannot be false that we are experiencing it right here and now, it is properly basic. We get to believe we are having that experience without needing any reason to believe that, other than the one reason entirely contained within itself: the fact that it is there, and thus cannot be false for as long as it remains there. But whether it is correct is another matter.

The same holds for memories, which are also properly basic. Whether they are true memories or false, or accurate or inaccurate, or flawed or precise, none of that is properly basic. But the conclusion that they are true or accurate or precise can be (and ought to be) an inference from properly basic experiences, which include thoughts and thus “interpretations,” as well as other memories, “experiences of coherence,” and “experiences of an idea making sense,” and so on (as having an intuition can be undeniable regardless of whether that intuition is true). These basic beliefs can even include undeniable desires, like an experience of having the desire to follow a certain rule in your thinking.

There is literally nothing else left. When you add up all the reasons (all the reasons) you have right now to believe any x, you will always (always) end up with a finite collection of experiences (whether a combination of perceptions, emotions, thoughts, memories, etc.), which are in turn entirely and without remainder reducible to a finite (not infinite) collection of properly basic, in fact literally undeniable, experiences (again, whether these be perceptions, emotions, thoughts, memories, etc.).

Therefore, every rational epistemology avoids infinite regress. Because there is always a point where the justifying evidence simply stops: at the basement of Cartesian knowledge. You couldn’t continue the train of evidence from there even if you wanted to. Therefore, there is no such thing as infinite regress in epistemology (see Epistemology without Insurmountable Regress or Fallacious Circularity). Which means the only difference between my epistemology and Plantinga’s is that he stops with an arbitrarily selected set of deniable assertions, whereas I argue we must keep going until we’ve gotten to the bottom, which is always a finite set of undeniable experiences. This is all it can ever be, because in the end, there isn’t anything else left (see Defending Naturalism as a Worldview).

Memory and the Subconscious

Hence I do not believe we must presume memory is true in order to reason. If that were the case, we could never identify a false or inaccurate memory. Instead, we hypothesize the reliability of our memory, and constantly test that hypothesis against current experience, which includes current experiences of those and other memories, and experiences of coherence among all our presently occurring experiences and memories, and so on.

Of course this goes on unconsciously or nonpropositionally most of the time, but the process is the same then as when it is fully cognitive (there is plenty of evidence to believe that, and no reason to believe otherwise). So just because a process of reasoning is unconscious does not mean we believe its conclusions for no reason. We likewise take for granted that a science textbook tells us the truth about what scientists have observed, but we don’t take this for granted for no reason. We have accumulated a great deal of evidence regarding the reliability not only of science textbooks, but of the processes and events that go into producing their contents. Just because we don’t reason through all this evidence every time we pick up a science book doesn’t mean our trust in that book is properly basic. It is not even basic.

There is a difference between working assumptions (things we don’t believe yet but are testing out), and empirical assumptions (things we assume only because we have some evidence or reason to believe them), and basic assumptions (things we assume but with no evidence or reason to believe them) and properly basic assumptions (things we assume but with no evidence or reason to believe them except themselves—the fact that they literally cannot be false is the evidence, the reason, we believe them).

Hence when we intuit that someone is lying to us, that they are lying is not a properly basic belief. If we can’t work out specific reasons why that intuition is valid, we won’t be justified in trusting it at all (see Correcting 5 Mistakes Atheists Make About Epistemology). But if we can work out such reasons, then we’re not looking at a properly basic belief. Because if there are reasons to believe it, it cannot even be a basic belief—much less properly basic. And if it could possibly be false, it cannot be properly basic anyway. So even when our intuition has become reliable enough that we trust it even without examination, we are still doing that for a reason, such as the fact that we have a large number of experiences of memories of our intuition’s success in relevantly similar cases. That these experiences exist is properly basic, but not what we induce from them, like that our intuition in such cases is reliable.

Thus, we don’t have to work out all the steps of reasoning or all the evidence we actually are working from. But we still need all that evidence and reasoning in order to be warranted in believing something. A belief can only be properly basic if we don’t have all that evidence and reasoning stashed away in our subconscious, and yet still are warranted in maintaining that belief. And only Cartesian knowledge meets that bill. In the end, we operate on the assumption of the reliability of memory in reasoning not as a given but as a justified hypothesis, based on long experience—in other words, a ton of evidence. We have good reason to assume memory is sufficiently reliable for us to successfully reason. The experience of those memories is properly basic; but the content of those memories is not. And it is that content that we use and rely on in reasoning.

Why We Believe in Logic

Our trust in logic is not a basic belief. We have accumulated extensive evidence (and hence reasons) to trust it (see The Argument from Reason and The Ontology of Logic).

Deductive reason appears to involve simultaneous perception of the major and minor premise, while induction is simply deduction with the major premise being some general inductive principle (like that “how things have gone is likely how they’ll go”). Our experience of the overlapping content of those premises, which simply is the conclusion, is a properly basic belief. So we do not require memory to arrive at a conclusion from two premises; we can do that in our immediate apprehension. We only require a memory to have previously stored, and thus recall at the present, the premises we are using in our reasoning at any given time, and thus to keep track of more than two premises and the results of their overlap. Likewise we need memory to keep testing our apprehensions (our conclusions) to ensure we have sufficient reason to believe them. But the reasoning itself, which we continually test with all these memory and sensory experiences, is properly basic.

Obviously, most remembered premises are the conclusions of previous acts of reasoning. But we still don’t require memory to reason, only to recall the results of past acts of reasoning, or to juggle multiple or complex premises. That this all works (and thus our premises and conclusions can be trusted) is not basic: it’s evidence-based. Because our working memory has a very small limit in terms of RAM: only about three bytes, maybe five, in terms of a lexical index—and though these bytes of information can index to far more complex models in our cognition, the limit on their number at any one time remains. So we do need to trust memory to engage in complex reasoning. We just don’t trust it because it is properly basic. We trust it for reasons. “But those reasons circularly include other memories you have to trust” is true, but the end game is the experience of all this, which includes an experience of its coherence: that it is working now is evidence that it has and will.

I go into how indexes and models work, and their importance to understanding all conscious thought, in What About Propositions? But note here that this is often the reason people are gullible or bad reasoners: they cannot keep track of more than two premises, and thus more than one step of reasoning, at any given time. That they fail where we succeed remains continual present evidence that our reliance on memory is working. The takeaway here is that we do not need to pretend even that the belief “our working memory is reliable” is properly basic. Only our experience of it is. That its content is reliable is something we need a lot more reason to believe. And indeed we know our memory’s reliability is not total, and therefore the possibility of its being unreliable is not even theoretical: we have to continually guard and manage against its repeated actuality. Which is why our belief in it cannot be properly basic, or even basic at all.

All Epistemologies Are Fundamentally Normative

This does get us to a realization, though: all epistemologies are fundamentally built on axioms that are, in fact, imperative propositions. In other words, every epistemology is constructed on top of a set of “I ought to believe x when y” propositions, and therefore, if it is true that any epistemology ought to be adopted by everyone, then epistemology as such is a subset of morality—and it would therefore be immoral to knowingly violate the axioms of a true epistemology. There is an ethics of belief.

All epistemological arguments end with a set of imperative propositions, not with a lone set of non-normative facts. For example:

• “I ought not to believe x when I have no reason to believe x“
• “I ought to believe x when x coheres with all other current experiences (including experiences of memories, etc.), with less fudging than any alternative I know”
• “I ought to believe the relative probability of x is y when y is (or is as close as I can know to) the actual frequency of x relative to ¬x in my experience”

And so on. Each of these general principles is irreducible in the sense that they cannot be made the conclusion of any non-question-begging set of premises. In a sense they are true by definition, insofar as I might choose to define such terms as “true,” “plausible,” “probable,” “credible,” etc., in exactly these ways. But in choosing to define these terms in such a way I am making a normative judgment, which rests on some belief regarding what I ought to do, which in turn rests ultimately on what I want. Do I want any of my desires fulfilled or thwarted? Do I want any of my plans to succeed or fail? Do I want any of my expectations to come true or be dashed? The answer to these questions entails a particular course of action, as much in epistemology as in any other matter.

To say that these principles are irreducible doesn’t mean I can’t defend them. It’s just that any such defense will ultimately still rest on some normative-making premise, such as “things will go better for me if I follow these principles,” which is itself, ultimately, a conclusion based on those same principles. Hence we always end up in some circular argument, a fact even Plantinga admits of his own and in fact every conceivable epistemology—even God’s. Yes, God as well. For even He can never be certain he is not the victim of a Cartesian Demon, or that there isn’t some flaw in his knowledge or memory or reasoning that prevents his detecting it. Because even he can only reason his way out of this by some circle or other (see, again, Andrew Moon’s Defense of Circular Arguments).

But this circularity ends infinite regress, which is why no epistemology really suffers from the problem of regress. The only difference between epistemologies is not whether they avoid regress, but whether they work. Which requires us to decide what counts as “working.” To choose “how I feel” instead of “whether it reliably gets me to what is the case” has the catastrophic outcome of failing to “reliably get me to what is the case” and thus failing to “reliably get me to results I really want.” This is a “you can’t serve both God and Mammon” moment (I discuss why in What’s the Harm? Why Religious Belief Is Always Bad). You have to choose whether “reliably gets me to results I really want” is the best metric for your epistemology, or something else. There is no third option. And when you work it out rationally, you’ll find that anything less than “reliably gets me to results I really want” will not be what you really want. So a rational person is always stuck with that metric (see The Objective Value Cascade). Everyone else is behaving irrationally.

How so?

The Foundation of Rational Belief

Well, remember how I started this whole essay by pointing out that there are always two lines of questioning, one about the facts and one about the principles we use to interpret those facts?

Okay.

The first line of questioning ends with a full dead stop at some finite set of undeniable experiences. For once you end up with an answer that cannot be false, you can no longer ask, “How do you know it’s true?” How do I know? Because it can’t be false. It would be nonsensical to then ask, “How do you know that?” For if I said it were false, then I would be saying “this experience exists right now and this experience does not exist right now,” which is a meaningless sentence, because it asserts what it also denies, and therefore it describes nothing. Thus, obviously, I cannot fail to know I am having such-and-such an experience right now. To express doubt about this would then be to question not the facts, but the rules I choose to follow when interpreting those facts—in particular, how I choose to define “true” and “false.”

Hence we end up with the second line of questioning, about our inductive and deductive principles. But all such lines of questioning end with a circular argument: our principles are true because our principles are true. There is therefore nowhere else to go. This is like Hawking’s notion of nutshell cosmology, where there is no first moment of time because as you approach it you eventually just curve back around and end up where you started. In such a case the timeline is not infinite, and yet also has no starting point. Because it’s circular. So, too, in epistemology: once your only justification for adopting a principle is the principle itself, you can no longer ask, “How do you know that principle is true?” How do I know? Because I observe it to be true. How do I know I observe it to be true? That I am observing it to be true is an undeniable experience. Hence we are back to the first line of questioning, about what I am experiencing, where all questioning ends.

Thus, there is no regress. But the underlying normative nature of this end game must not be overlooked. In effect, my entire epistemology rests on a conjunction of just three premises, which I will greatly oversimplify for the point here:

• A: “Following certain principles will probably make things go better for me than not following them will”
• B: “If I want things to go better for me, I ought to follow the principles that will probably make things go better for me than not following them will”
• C: “I want things to go better for me.”

Properly interpreted, C is an undeniable experience of desire and thus properly basic. Of course, what I mean by “I” and what it means for “things to go better for me” is a whole other matter, and not the subject at present, but whatever we settle on, we can certainly assign some meaning to “I” and “things going better for me” that would make C a properly basic belief. Because I don’t need to work out whether my beliefs are true to accept that the principle is true.

For example, I can be wrong or confused about what “I” means and what “better for me” means, but I cannot be confused about the fact that on some construction of those two terms it is always true that “I want things to go better for me.” That realization is properly basic. Because it can’t actually be false. Even if I incidentally, irrationally, want things to go worse for me, I am then simply redefining what is better for me. There is no way around the fact that whatever I really want is what I really want and thus is best for me. What can be false is what “what I really want” happens to be. I can be wrong about that. But there is some truth to find there, and whatever it is, it will tautologically be what will most satisfy me to pursue. So the principle is properly basic.

Meanwhile, in support of A and B, I have the evidence of cause and effect, which consists of a collection of undeniable experiences (including experiences of memories) of these causal hypotheses being fulfilled, combined with the undeniable experience of my lacking any memory or perception of these causal hypotheses being invalidated. In other words, that following certain principles more often than not makes things go better for me—better than not following them does—is not a basic belief, because I justify it by appeal to a ton of evidence making it probably true. But at the bottom of all that evidence is nothing but a ton of properly basic beliefs: all of it raw, uninterpreted experiences—of the now and of the past (in my memories: see Bayesian Analysis of the Barkasi-Sant’Anna Defense of Naive Memory Realism). Both A and B thus reduce to a future subjunctive: “I predict things will go a certain way.” Which means simply “they will” go that way (which is then either true or false) and thus that you “ought” to thus act a certain way means “you will” act that way when you are sufficiently informed and behaving rationally (as I lay out in my chapter on moral reasoning in The End of Christianity).

Thus, the buck stops with the evidence, and the evidence is a finite collection of undeniable experiences. Though these undeniable experiences are also compatible with a contrary hypothesis (¬A or ¬B), the experiences I am having do not entail that contrary hypothesis unless we introduce an intervening premise, such as “a Cartesian Demon has fooled with my memory and is deceiving my senses.” Which we only have reason to believe is always improbable. Thus, as long as we have no reason to believe any such premise, then we have no reason to believe any hypothesis contrary to A and B. And as long as we have some reason to believe A and B instead, then we have reason to believe A and B. In other words, among the total collection of our undeniable experiences at any given moment, all we have is evidence supporting A and B and none supporting ¬A or ¬B, which is to say, none supporting any premise that would have to be true in order for ¬A or ¬B to be true.

So our bottom basement of circularity arrives here, at the point when we decide on the most fundamental principle underlying all of the above, which I will call principle K:

• K: “I ought to believe x when I have (a) evidence supporting x and (b) no evidence supporting what would have to be true for me to have (a) and yet for x to be false.”

Here, too, I am oversimplifying. A complete principle, for instance, would include the relationship between degrees of evidence and degrees of belief. And I am hiding within term (a) all cases when x being undeniable is evidence supporting x. And I am hiding within term (b) the distinct case of having direct evidence against x (being just another way of having “no evidence supporting” that). And I am setting aside complex cases when (a) obtains but not (b), which need more tangled principles to resolve. And so on. But we can expand K to include all these things. It’s just easier to keep track of the point I want to make if we focus on this simple version of it instead. Such is K.

The contrary inductive principle ¬K would then be:

• ¬K: “I ought to doubt x when I have (a) evidence supporting x and (b) no evidence supporting anything else that would have to be true for me to have (a) and yet for x to be false.”

We are thus faced with an ultimate choice: K or ¬K? Which principle do I follow? I can try them both out right now, and immediately see that following K leads to correct predictions and the satisfaction of my desires and the fulfillment of my plans, while following ¬K does much poorly in all three respects. I can repeat this test endlessly. It still remains that a Cartesian Demon could be meddling with my mind so that I keep falsely experiencing and remembering the good performance of K and the poor performance of ¬K, when all the while, unbeknownst to me, ¬K has been performing better than K (or equally as well). But as long as any Cartesian Demon keeps doing this, what’s the difference between that, and K actually performing well and ¬K performing poorly?

Until you allow that there could be some potential difference between the world created for us by a Cartesian Demon and a real world, there is no difference between them that matters. The world created by the Cartesian Demon is a real world: our desires are actually fulfilled, our plans are actually realized, our predictions actually come true, and continue to do so, always and forever (until we’re dead). Only if you allow that it might not be “always and forever” does it become meaningful to talk about a difference, because then, and only then, is there a difference that matters (hence theists want God to be this Cartesian Demon: see The Argument to the Ontological Whatsit). But then we have the possibility of predicting different outcomes for each hypothesis, and so the two hypotheses become testable, which means a Cartesian Demon hypothesis is only ever confirmed when its unique predictions come true. But as long as it remains unconfirmed, we continue to lack any reason to believe that hypothesis—even if (unbeknownst to us) it turns out to be true.

And Thus Regress Ends

This is the epistemological end game. We always end with a conundrum Plantinga faces every bit as much as you or I. So pointing out the inevitable circularity of rejecting all Cartesian Demon hypotheses (by relying on our evidence-based trust in reasoning on a foundation of the undeniables of direct experience) is no objection to my epistemology, since it is equally an objection against all epistemologies. It forces us all against the rocks of the same dilemma: we simply have to choose how to behave. Will it be in accordance with K, or ¬K? We constantly observe, in every waking moment that we bother to test, that the undeniable facts of both our desires and our immediate and present experiences are only satisfied by following K. Therefore it makes no sense to follow ¬K, even if it happened to be the case (unbeknownst to us) that ¬K is true. So that’s where all epistemologies end: “I ought to follow K.”

This is not a properly basic belief, because we admit K could be false, and yet we have a reason to follow K rather than ¬K, namely the undeniable fact that we desire things, combined with the undeniable fact that in any moment we put to the test, we will usually experience the fulfillment of our desires only when following K, but not when we follow ¬K. These two facts do not combine into a deductive proof that K is true, but they do provide a reason to follow K—in fact a quite good one—and as long as we have no reason not to follow K, having a reason to follow K is a sufficient reason to do so—and even more so when having a good reason to. And yet ultimately all of those reasons, which are all that justify my embracing the fundamental principle underlying the whole of my epistemology, end at undeniable facts, which are therefore properly basic.

Regress ends.

⌘

Note to My Readers: Since this is a difficult and convoluted subject I may have made errors in its precise wording in places. Accordingly I will continually revise this article‘s wording whenever I or anyone find problems to correct. Ask about any you find or are unsure about in comments, and I’ll look into it and see if any revision is needed.