Cohen on facts and principles

by Chris Bertram on July 31, 2003

I’ve spent this morning puzzling through Jerry Cohen’s “Facts and Principles” from _Philosophy and Public Affairs_ (31:3 Summer 2003). It is, as I and others have intimated already, an important article and I can’t be confident that I’ve “got” it yet. I do think, though, that I can say that his thesis is not quite the threat to naturalism that I took it to be, unless it is coupled with some further commitments (although, as it happens, those dangerous further commitments are ones I accept). The basic argument Cohen puts forward is a really simple one, claiming that where people seek to ground their moral commitments on principles, some of those principles must hold independently of the way the world happens to be (“the facts”).

Cohen argues for three premises:

P1 “…whenever a fact _F_ confers support on a principle _P_, there is an explanation _why F_ supports _P_, that is, an explanation of how _F_ represents a reason to endorse _P_” (p. 217)

P2 “…the explanation whose existence is affirmed by the first premise invokes or implies a more ultimate principle, commitment to which would survive denial of _F_, a more ultimate principle that explains _why F_ supports _P_….” (217-8)

P3 Where the grounding principle explaining why _F_ supports _P_ is itself fact-dependent, and further interrogation reveals yet deeper grounding principles, that iterative sequence of asking for supporting principles will not go on forever but will (pretty soon) come to rest on a grounding principle that is not itself fact-dependent.

Cohen claims that “it follows from the stated premises that every fact-sensitive principle reflects a fact insensitive principle: that is true _both_ within the structure of the principled beliefs of a given person …. _and_ …. within the structure of the objective truth about principles.” (218)

Much of Cohen’s paper consists of clarification of his view and a careful attempt to set out what he is not saying. So, for example, he says his thesis is neutral with respect to the main metaethical disputes (among realists, quasi-realists, emotivists etc etc etc), that it is not the same as Achilles and the tortoise, that it is not a view about how people actually come to acquire the principles that they hold (a process that will require engagement with the facts) and so on. Cohen’s view is, most basically, about the logical structure of people’s moral beliefs. It also presupposes that moral reasoning and judgement consists (at least in part) of the application of general principles to circumstances in the light of the facts, so Cohen’s view is of limited interest to moral particularists who believe that moral judgement isn’t about the application of such principles.

The principal target of Cohen’s article is Rawlsian constructivism. This is because Rawls believes that the way the world is (the facts) enter into the construction of the fundamental principles of justice (via, for instance, the general facts made available to the parties in Rawls’s original position). Cohen believes that Rawls is not altogether consistent here, in any case, since the design of Rawls’s constructivist procedure rests on general claims (that persons are to be considered as free and equal) that are either themselves fact-independent or rest on further principles that are. So, for Cohen at least, Rawls’s putatively fundamental principles of justice aren’t fundamental at all, but merely derivative or regulatory principles that actually derive from deeper fact-independent principles.

Is Cohen’s argument damaging to ethical naturalism? That of course is going to depend on what we mean by ethical naturalism and, as we’ve seen over the past few days, there’s plenty of room for debate and/or confusion about that. Cohen’s argument has nothing to say about what ethical principles amount to, and if all such principles amount to is the expression of attitudes then there’s no going to be not problem for most naturalisic views. On the other hand, if we have a commitment to moral objectivity, then it looks like we are committed to there being objective truths in ethics that are (logically) completely independent of the way the world happens to be (microphysically or otherwise). Whether _that_ is a threat or not probably depends on the sort of objectivity we sign up for. If moral principles are _a priori_ (and so on a par with, say truths of logic) they may not be. But, to be honest, I’m not confident in what I think about this.

In “his post yesterday”:, Brian suggested that Cohen’s view might be correct but trivial. I think that it is probably a mistake to express the point thus. If we are, as Cohen thinks, committed to some ground-level, fact independent moral principles then those principles are likely to be quite substantive (e.g, of the order of the fact-independent principle “all beings with characteristics X have the right to equal concern and respect”).

Is Cohen’s argument damaging to Rawlsian constructivism? If Cohen is correct, Rawlsians might reasonably, though concessively, reply. They might argue that it is true that if we look at what the logical structure of people’s ethical beliefs ought to be, then fact-independent principles are at the bottom. It isn’t the case then, that what justifies and constitutes our most fundamental commitments is that they derive from a constructivist procedure. But (1), epistemically, such a procedure is the best method for getting at what those commitments are and (2) given “the facts”, the regulatory principles which we are practically most interested in are best seen as the product of a constructivist apparatus. Too concessive? I think most Rawlsians could live with it.



Loren 07.31.03 at 9:00 pm

Brian, Chris: interesting posts. Offhand, here’s a possible way in which Gerald Cohen might be forced to take sides in at least some of the metaethical disputes he is at pains to avoid (which in turn might make is “facts and principles” paper more directly relevant to debates about naturalism and ethics).. Imagine an exuberant evolutionary naturalist. Her exuberance is evidenced by the zeal with which she wields the weak anthropic principle. She says to Cohen:

“I take no sides on a range of debates about truth, knowledge, normativity, the nature of reality, etc. I merely make the simple point that such debates could only take place in a universe in which agents of a certain sort have evolved. This ontological dependence constitutes an ultimate factual grounding relation. To be clear, I am not explaining or deducing anything about the sources of our moral intuitions or deepest convictions. I merely make a claim about how normative principles are ultimately grounded in facts about the universe, namely that it is the sort of universe in which conscious beings could evolve in such ways that normative principles are required. Nor need I provide a comprehensive demonstration of my anthropic argument. The burden is on my opponents in this debate to show an example where my argument does not hold — that is, a case of a normative principle that is affirmed independent of facts, including the facts cited by my anthropic argument, i.e. the fact that the universe is such that beings of a certain sort have evolved and can affirm said normative principle.”

Now Cohen might see this sort of claim as analogous to his example of the holist who argues that our moral beliefs are ultimately related to the facts of some far distant galaxy. But I don’t think the evolutionary naturalist can be dismissed so easily, because her arsenal of facts, and the evolutionary logic they impose on forms of life, may well be brought to bear on moral claims in ways that Cohen has to answer explicitly (“no, that sort of grounding relation is not relevant here, because …”)

It occurs to me that Cohen can only exempt such substantively interesting grounding relations (“interesting” and substantive compared to, say, the claim about distant galaxies as grounding facts) by making some explicit claims about the nature of truth, the substance of acceptable grounding relations, the sources of normative judgements, and probably some other points that seem, well, unabashedly metaethical.

And when Cohen takes some meaty metaethical positions to answer folks like the exuberant evolutionary naturalist, I suspect that it might become clear(er?) that his real beef with Rawls is that justice is, for Cohen, ultimately metaphysical, not political.


Brad DeLong 07.31.03 at 9:06 pm

There is a paper by Charles L. Dodgson, “Achilles and the Tortoise,” which points out you only have to acknowledge the force of:

[(P->Q) & (P)] -> Q

if you believe that:

[[(P->Q) & (P)] -> Q]

So the argument is only watertight if you formulate it as:

[[[(P->Q) & (P)] -> Q] & (P->Q) & (P)] -> Q

but then you only have to acknowledge the force of this statement above if you believe that:

[[[[(P->Q) & (P)] -> Q] & (P->Q) & (P)] -> Q]

Is Cohen’s argument any different, at bottom?


Brian Weatherson 07.31.03 at 9:24 pm

Two points on Achilles and Cohen.

First the geek point – the paper was published under the name ‘Lewis Carroll’. It’s not entirely clear why this is, since Dodgson published other things in logic under his own name. (Geek trivia question: Apart from the joke issues, have there been any other papers published in _Mind_ under a pseudonym? For that matter, what was the last (serious) philosophy paper published under a pseudonym?)

Serious point: it isn’t clear that Cohen’s making the same point as Carroll. Here’s the kind of thing you’d have to believe that their _conclusions_ are at base the same:

(a) That the ultimate, fact-independent, principles that Cohen is reaching for are conceptual truths; and
(b) The role of conceptual truths in arguments is more akin to inference rules than to premises.

(Or you could replace ‘conceptual truths’ in both (a) and (b) for same effect, but I think you’d be making both less plausible.)

Anyway, I believe both (a) and (b), so I’m with Brad, and D^2 in a previous thread, in thinking these really are similar points. But both of those are very contentious claims, to say the least.

In any case, I think they are making different arguments since Carroll was talking about arguments and Cohen is talking about explanations. I still think the conclusions are similar, but they are different routes.


Brian Weatherson 07.31.03 at 9:29 pm

That probably wasn’t as clear as I intended. I (following a ton of people) take the point of Carroll’s paper to be that we have to distinguish premises from inference rules, and we have to believe in inference rules.

Are Cohen’s fact-dependent principles and fact-independent principles analogous to premises and inference rules? I think so, but it’s a very hard question. (By the way, I hope my verbosity doesn’t cut off a conversation about Loren’s very interesting post.)


Chris Bertram 07.31.03 at 10:38 pm

On Achilles and the tortoise, best to let Cohen speak for himself:

bq. “…that [first] premise does not say anything like what Lewis Carroll’s tortoise said to Achilles. That misguided tortoise said that an inference is valid only if the principle that validates it is stated as a further premise of the inference. An unmanageable infinite regress ensues, and the take-home lesson is that principles of inferential validity do not function as premises in the arguments that they validate….[My] first premise concerns not inference and what makes them valid but justifying grounds and what makes them justify….When someone claims that a fact grounds a principle, she affirms a _grounding_ relation, not one of deductive inference. And I do not say, no, that fact doesn’t ground that principle unless we add…; I simply ask, non-rhetorically, _why_ the fact supports the principle, and I _claim_ that a satisfactory answer will always feature a further principle, _P1_: that is, precisely, a (correct!) claim, mnot a move demanded by logic. And unlike the sequence generated by the tortoise, the sequence that my claim generates is finite: it comes to an end with the statement of a principle that is fact-insensitive and, therefore, one to which my sequence-generating question (‘Why does this fact support this principle?’) does not apply.” (Facts and Principles, p. 220)


Chris Bertram 07.31.03 at 10:52 pm

On Loren’s post:

I’m not sure I understand exactly what you’re driving at, or, at least, how it is germane to Cohen’s argument, but two possible replies spring to mind.

1) Would be to point to Cohen’s remarks concerning ought implies can at pp. 230-31 and to include (perhaps tacitly) in the proper formulation of all ultimate fact-insensitive principles a conditional clause to the effect that one should phi if one is in a world where phi-ing is the kind of thing that is possible.

2) Would be to worry about what is driving your “a universe in which agents of a certain sort have evolved” formulation. I take it that if we turned out to be caused by a demiurge we’d face the same practical problems as if we have (as I believe we have) evolved from the slime. So our particular causal history can’t be all that important to the issues of ethical principle. But maybe I’m missing the point here.


Brad DeLong 08.01.03 at 12:58 am

>>we have to distinguish premises from inference rules, and we have to believe in inference rules.<< Now are the claims "we have to distinguish premises from inference rules" and "we have to believe in inference rules" themselves premises? Or are they inference rules?


Jon 08.01.03 at 2:27 am

As I’ll explain in a post I’ll put up a little later, I don’t think Rawls (or Rawlsians) need to make any concessions to Cohen – Rawls just doesn’t hold the thesis that Cohen attributes to him.

I think Cohen’s point is, in fact, very close to the Achilles/tortoise argument. Cohen certainly overstates his case when he says that his first premise “does not say anything like what Lewis Carroll’s tortoise said to Achilles.” One lesson from that argument is that it is possible to codify inferences with inference rules. But if these inference rules do nothing but make the inferences explicit, by themselves, they do not add any justificatory weight.

Cohen seems to assume that justification can only procede by appealling to more abstract principles – in particular, the ones that codify an inference from a fact to a principle. But if this more abstract principle is only codifying the inferences we already make, by itself, it adds no justificatory weight.

Of course, appeal to more abstract principles is one way that justification can go (when those principles can be defended), but the idea of reflective equilibrium suggests that justification can also go in the opposite direction, as well.


Brian Weatherson 08.01.03 at 2:57 am

_Now are the claims “we have to distinguish premises from inference rules” and “we have to believe in inference rules” themselves premises? Or are they inference rules?_

Neither – they’re conclusions :)


Shai 08.01.03 at 3:45 am

Apparently I’m not sophisticated enough to understand the argument in Cohen’s article yet. What Chris quotes looks very much like Amartya Sen on “basicness”, and in fact Cohen does cite him (for a simple summary of Sen on this — perhaps too simple — see this specifically, pages 11-14). Can someone describe to me the difference between the two positions? I thought I understood Sen’s argument, but I wonder now that I’m having trouble with Cohen’s (I’m a second year compsci undergrad if that helps — but I do have a 4.0 gpa so it doesn’t have to be dumbed down too much, I hope).


Shai 08.01.03 at 4:30 am

Actually, I suppose he illustrates the difference in his fourth note. He disagrees with the position Sen takes in a 1967 paper; I’ll have to look through some of Sen’s newer books to see if he’s said anything else on this. Does anyone know if Cohen expands on this in his book “If You’re an Egalitarian, How Come You’re So Rich?”.

I’m still wondering… he disagrees with Sen on the stringency of his criteria for basicness and yet seems to take the position to its logical conclusion? Anyone want to comment?


Shai 08.01.03 at 4:50 am

Nevermind, I think I understand at least the basic difference now. Sorry for throwing up on this comments thread.


dsquared 08.01.03 at 6:43 am

I’ve got a feeling that the discussion of the status of functional explanations in KM’sToH might illuminate this article, or be illuminated by it.


Chris Bertram 08.01.03 at 7:37 am

Jon rightly points out that justification can proceed in both directions, as it were. But I don’t think that Cohen need deny this, and indeed he may affirm it, if by “justification” is meant the real-world process by which we acquire and clarify our moral beliefs. Cohen’s focus, if I understand him aright, is on the post-justificatory (in the sense I’ve just used it) structure than our beliefs must have.


Jon 08.01.03 at 1:40 pm

Yes, in the real world, we go back and forth between levels of abstraction. If successful in reaching (or moving toward) reflective equilibrium, we wind up with a complicated structure of more general and more specific commitments.

If Cohen’s point concerns the logical structure of deduction within that structure, we have *exactly* the Achilles/tortoise case. If his point is about what justifies what, we need to determine where our most secure commitments lie. And these need not be the most abstract commitments. (For those keeping score, note that I’m here moving beyond the limits of a political conception of justice and relying on my own (partially) comprehensive doctrine about the structure of practical reason.)

Comments on this entry are closed.