<body class="blog" onload="">

Ratiocination

The Maxim and Determinism


I recently gave a paper on the Kantian maxim that 'ought implies can.' Thanks are due, incidentally, to those who attended the talk; I received some helpful feedback and it will inform my future research into the topic.

That said, Brad Rettler comments on my paper in this post, and I'd like to interact with some of his thoughts.

Brad's central worry is that my affirmation of the maxim is incompatible with my views about free will, moral responsibility, and determinism. For the record, I am a semi-compatibilist. I think moral responsibility is compatible with determinism, while free will is not. I grant the point that my affirmation of the maxim stands in tension with my semi-compatibilsm. In this post, I try to state more precisely why this tension might turn out to be something worse, an incompatibility.

Brad's statement of the worry is as follows:

If determinism is true, then every action one performs, one necessarily performs (where necessity is physical necessity). But some of those actions, one ought not do. But if one ought not do some of those actions, then according to the Maxim one can refrain from doing them. But according to determinism, in fact one cannot refrain from doing them.

Thus, at least one of the following must be false: 1) S cannot refrain from performing a at t (given the laws of physics and the past, S must perform a at t). 2) S ought not have performed a at t (I think one cannot deny this without assuming the Maxim). 3) S could have refrained from performing a at t (this follows from (2) and the Maxim).

So, if we assume determinism, then (1) is true, and the Maxim is false.

The worry is real, but there are a few infelicities I comment on before trying to defuse it. First, determinism doesn't imply that anything that happens happens as a matter of physical necessity. It implies rather that given the actual past, anything that happens happens as a matter of physical necessity. We must not here confuse the necessity of consequence with the necessity of the consequent. Second, the formulation of the maxim I favor doesn't say that if someone ought to do something at a time, she can do so at that time. Rather, I have advocated a maxim with a tracing condition: if S ought to perform act a at t, then either S is able perform a at t or there is a time t* and an act b such that S can perform b at t* and were S to perform b at t*, it would be the case that S can perform a at t.

(3) doesnt' follow from (2) and the maxim, then. It does not follow from (2) and the maxim that I have advanced, at least. But something in the neighborhood does. Similarly, it's not enough to claim that determinism rules out someone's ability to do otherwise than she does. It must also be true that determinism rules out the ability to do anything such that doing so enables one to do otherwise than she in fact does. Following is my attempt to state with just a little rigor the argument Brad proposes:

1. If determinism is true, then no one is able to either do otherwise than she actually does, nor is anyone able to do anything such that had she done it, she would have been able to do otherwise than she actually does. (assumed, incompatibilism about free will and determinism)
2. Determinism is true. (assume for reductio)
3. Someone ought to have done otherwise than she actually does. (assumed)
4. If someone ought to have done otherwise than she actually does, then someone is able to either do otherwise than she actually does, or someone is able to do something such that had she done it, someone would have been able to do otherwise than she actually does. (from the Maxim)
5. Therefore, no one is able to either do otherwise than she actually does, nor is anyone able to do anything such that had she done it, someone would have been able to do otherwise than she actually does (from 1 and 2).
6. Therefore, someone is able to either do otherwise than she actually does, or someone is able to do something such that had she done it, someone would have been able to do otherwise than she actually does. (from 3 and 4)
7. Oops. (from 5 and 6)

Suppose I got the worry right (did I, Brad?); in my next related post, I'll try and defuse it.

Is Justification Closed Under Simplification?


In a word, no. Epistemologists typically distinguish between doxastic and propositional justification. The former is belief entailing; the latter is not. In this post, I shall argue that neither condition is closed under simplification.

Let 'J(s, t, p)' abbreviate 'S at t is justified in believing that p.' We may ask whether the following principle is true on either reading of 'is justified:'

Closure: Necessarily, (J(s, t, p&q) only if J(s, t, p) & J(s, t, q))

Closure has some degree of initial plausibility. In most cases, at least, where we believe conjunctions, we believe the conjuncts too. And it seems that whatever justification we might have for the conjunction would apply to the conjuncts too.

But this is a snare and a delusion. The principle as stated is false.

On the doxastic reading of J, J(s, t, p) is true only if S at t believes that p. But it’s possible to believe a conjunction without believing both conjuncts. Take the busy Jones, who simply hasn’t got the time to perform simplification. Jones is justified in believing some conjunction (he came to believe the conjunction on hearing it asserted by some reliable source, say); but Jones has never reflected on his belief in the conjunction and realized that he could infer either conjunct from it. Jones is justified in believing a conjunction at a time then, even though he does not believe one of the conjuncts at that time (and hence does not believe with justification one of the conjuncts at that time).

Consider now the propositional reading of J, according to which a subject may have justification for p even if that subject doesn’t in fact believe that p. On this reading too, Closure is false, and not because of a failure to believe. Sally has recently taken a course in logic, where she was taught that simplification is invalid. Sally believes that p does not follow from statements of the form p&q. Sally is mistaken, but she has very good reasons to believe this (her teacher is quite well-respected in the field, she hears). Suppose now that Sally has a good argument for the thesis that p&q. Sally has no argument for the thesis that p, however, since she doesn’t think simplification is a valid maneuver. In fact, it seems to Sally that to infer p from the premises before her is a decidedly unjustified move. It’s plausible in this case to think two things. First, Sally has propositional justification for p&q. Second, Sally does not have propositional justification for p (were she to form the belief that p, say, she’d be believing contrary to her epistemic duties). If this is right, Closure fails on the propositional reading of J too.

On Lewis


A Facebook group was recently created: the David Lewis Appreciation Society.

I love it; David Lewis was a philosopher's philosopher, and it warms my heart to find others who appreciate him as I do. I've only been doing philosophy for a few years, and only in the last year or so have I really grown to appreciate the depth, wit, and wide-ranging power of David Lewis' writing. These days, I'm just sad that he's not around anymore. I would have loved to meet the man.

David Chalmers, incidentally, has some interesting (and quite moving, I think) thoughts about Lewis and his passing posted here. They are well worth the read for anyone who has appreciated Lewis.

Warrant: A Unique Epistemic Good


In my last post, I introduced the uniqueness problem for warrant. While warrant is supposed to be just one condition that turns true belief into knowledge, it turns out that there are many such (non-equivalent) conditions.

To solve this problem, I shall first do a little housekeeping on the definition of warrant. I shall refine the standard definition without abandoning its most central claim: that warrant is a condition satisfying a particular function or role in the epistemic economy of an agent. Recall that Plantinga and others following his lead define warrant as that condition (whatever it is) that fills the gap between mere true belief and knowledge. This conception of warrant's role is supposed to be non-partisan (not partial to any particular account of what warrant actually is), and I shall preserve this feature.

As pointed out earlier, on this definition, knowledge itself counts as warrant. This odd result can be shaved off with the following refinement:

D1: Warrant is that logically weakest condition (whatever it is) that fills the gap between mere true belief and knowledge.

(Condition C fulfilling role R is a logically weakest condition for R just in case there is no condition C` such that C` fulfills role R and C` entails, but is not entailed by C). On D1, knowledge doesn't count as a warrant condition, since knowledge entails (but is not entailed by) other weaker warrant conditions. D1 shaves off some other infelicitous candidate warrant conditions too. Mike Huemer anticipates this maneuver in a footnote, and claims that there still can be no guarantee that D1 picks out a unique condition; there may be many logically weakest conditions that satisfy the warrant role. He's right. And we can do better. Before we do, however, there is one more problem to address. Call this the epistemic good problem.

Warrant is supposed to be an epistemic good; an agent A having more warrant than B is in a better epistemic situation than B, all other things being equal. So if a definition of warrant's role is to be any good, it's got to pick out only epistemic goods, I'd think. But some conditions will satisfy the role picked out by D1 even if they aren't epistemic goods. Take this example of Huemer's: p only if C (where C is itself a warrant condition). p only if C is one of those conditions picked out by D1, but it hardly seems to be an epistemic good (it's satisfied whenever p is false, for one). Huemer employs this condition to argue that warrant need not be an epistemic good. But Huemer's modus ponens shall be my modus tollens. Whatever warrant is, it seems to me, it's got to be an epistemic good. Thus, warrant can't include in its extension conditions like p only if C. With this insight in hand, then, I regroup as follows.

Let S be the set of all those logically weakest conditions (whatever they are) that are epistemic goods and that fill the gap between mere true belief and knowledge. Note that conditions like p only if C will not be elements of S. Define warrant thusly:

D2: Warrant is the maximal disjunction on S.

(Where x = {1, 2, 3...} y is a maximal disjunction on x just in case y = 1 or 2 or 3 or... for every element of x). My strategy is simple. Suppose there are a plethora of conditions satisfying the warrant role (broadly construed); let's just call 'warrant' the disjunction of all these conditions. I think this move is consistent with the current conception of warrant. It fulfills some desiderata on definitions of warrant that I gestured towards, for one (functional definition, non-partisan, epistemic good). D2 will also prove fruitful when trying to say something interesting about warrant, since there is just one maximal disjunction on S.

Aside: there are some set-theoretical worries in the neighborhood. They are of little concern to me at the moment, but I will address them in a paper-length version of this series.

I have thus far resolved the uniqueness problem and the epistemic good problem. In my next post, I shall apply my results to the truth problem. That is, I shall answer the question, "does warrant entail truth?"

Warrant and Uniqueness


Warrant, we are taught from our mother's knee, is that condition (whatever it is) that fills the gap between mere true belief and knowledge.

Aside. perhaps it's strange think of this notion of warrant as being taught from a mother's knee, given that it was imported into the epistemology literature fairly recently (by Plantinga in 1993). But I'm only 22, and in all the three or four years that I've been doing philosophy, I've always been taught this notion of warrant. End aside.

Some have tried to say interesting things about warrant on this conception, interesting things that do not depend on any particular cashing out of what warrant actually is (undefeated justification, belief formed in a reliable way, etc.). Most prominently, Trenton Merricks has advanced arguments toward the claim that warrant (whatever it is) entails truth.

In the next few posts, I'm going to evaluate this claim.

I begin by noting the uniqueness problem. Why think that there is only one condition which converts true belief into knowledge? Plantinga's definition seems to assume that there is just one such condition. But is this right? Brief reflection will reveal that it can't be:

Take knowledge that p. This is a condition that, when added to the mere true belief that p begets knowledge: a warrant condition. And knowledge that p is distinct from and not equivalent to any proposed warrant condition I know of.

But there are deeper problems still. Let W1 = S is warranted in believing that p. Let W2 = p only if C (where C is some condition that given the input of true belief outputs knowledge). W1 and W2 are both a warrant conditions; when conjoined with p and S believes that p, W1 and W2 both output S knows that p. Note further that W1 and W2 are not equivalent conditions; one can be satisfied without the other. Thus, for any warrant condition, there is some distinct and non-equivalent warrant condition (this argument I owe to Mike Huemer's 2005 paper in Philosophical Studies).

It gets worse. For any warrant condition W that entails truth, there is a warrant condition that doesn't (p only if W). And for any warrant condition W that doesn't entail truth, there is a warrant condition that does (p and W). It seems that not much can be said about warrant after all, so long as we remain neutral between competing accounts of what actually plays the warrant role.

This is the uniqueness problem, and it can be solved. Doing so is the project I'll take up in my next post.

Use and Mention


Merricks on 'Reading Parfit'


One who comes to Reading Parfit confused about the further details of Partit's reductionism... will leave Reading Parfit confused. Perhaps the further details of Parfit's allegedly revisionary metaphysics of persons and personal identity are so inscrutable because--to wax Quinean--there is nothing there to scrute.

-Trenton Merricks, review of Jonathan Dancy (ed.) Reading Parfit in Philosophical Review 108 (1999): 422-425.