Sunday, April 26, 2009

Corrupt Mind or Honest Mistake?

I think there's an important difference between factual and moral mistakes. Intuitively, the person who mistakenly puts arsenic in their friend's coffee (believing it to be sugar), has made an "honest mistake" that doesn't reflect on their reliability or quality as an agent. Their internal deliberations proceeded appropriately enough; it's just the external inputs that were faulty. Compare this to the person who mistakenly believes that poisoning people is morally okay. We would not be inclined to call such moral ignorance - however sincere - an "honest mistake", for it reveals a kind of corruption in the agent. They are revealed to be vicious, lacking concerns that they ought to have, though they may not themselves realize it.

Gideon Yaffe, in his excellent paper 'Excusing Mistakes of Law' (Philosophers' Imprint, 2009), captures this intuitive idea in the following principle:
Uncorrupted Deliberation Principle: If D falsely believes that p and such a false belief indicates that his deliberations pertaining to A-ing were acceptable, then D is excused for A-ing.

Intuitively: the unwitting (factually ignorant) poisoner deliberates in an acceptable manner, whereas the morally ignorant poisoner does not. In the former case, the flaw is in the inputs to deliberation; in the latter case, the flaw lies in the agent's deliberation itself.

This assumes that normative beliefs play a different role in deliberation from factual beliefs. In particular, whereas factual beliefs are seen as "mere inputs" to deliberation, normative beliefs are seen as partly constitutive of the deliberative process. But what, exactly, does the difference consist in? One possibility is to understand factual beliefs as providing premises of reasoning, whereas moral principles serve not as premises but as rules of inference. [Feel free to skip the following section if you're already familiar with this distinction.]

Compare the following two arguments:
1. The Sun has risen every day thus far.
So: The Sun will (probably) rise tomorrow.

1. The Sun has risen every day thus far.
2. The future will (probably) resemble the past. [In particular, if the Sun has risen before, it (likely) will again.]
So: The Sun will (probably) rise tomorrow.

The first is an inductive argument, which moves directly from a claim about the observed to a conclusion about the unobserved. The second is a deductive argument, which takes the "Inductive Principle" [2] as a premise. There's an obvious connection between the two arguments, since our grounds for believing the inductive inference to be cogent are ipso facto grounds for thinking that [2] is true. Nonetheless, there's an important formal difference between a rule of inference and its associated proposition, since rules of inference license certain moves in a way that inert premises by themselves cannot. (This is the crucial lesson of Lewis Carroll's tortoise.)

In the same way, we may think there's a formal difference between the following two chains of reasoning:
1. I could poison Bob by putting this arsenic in his tea.
Therefore, I have sufficient reason to put the arsenic in his tea.

1. I could poison Bob by putting this arsenic in his tea.
2. I have sufficient reason to poison Bob.
Therefore, I have sufficient reason to put the arsenic in his tea.

The first treats normative principles as rules of inference, internal to the deliberation, licensing moves from factual premises directly to practical conclusions. The latter, by contrast, makes the normative assumptions explicit, treating them as inputs (on a par with any other premise), from which we derive the conclusion using more general rules of reasoning.

Here's my worry: do we have any grounds for favouring the former interpretation? For if one preferred the latter instead, then (at least from a formal perspective) it begins to look as though moral beliefs are mere "inputs" to deliberation in just the same way that factual beliefs are. And so the morally ignorant poisoner could object, "My deliberation was in no way corrupt; I reasoned as well as anybody could ask, given my premises. The only problem was that the premises turned out to be false! Since I had no way of knowing this, given the peculiarity of my circumstances, this in no way reflects more generally on my internal quality as an agent, so I ought to be excused from blame for this inadvertent wrongdoing."

What do you think?


  1. Great post. As for the final dilemma, I'm not sure it's all that big a bullet to bite. I don't have very strong intuitions on the question.
    As a solution, we can distinguish between agents being factually mistaken and normatively mistaken. Obviously (strong intuitions?) it reflects more negatively on the agent in a pragmatic way if she is normatively mistaken than if she is factually mistaken. I'm not sure if that's entirely acceptable, but I'm also not convinced that the presented dilemma is a big deal.

  2. I agree that when we conceive of normative principles as explicit premises, it's easy to see normative ignorance as excusing wrong actions in much the same way that factual ignorance does.

    However, I'm not sure that things change much if we conceive of normative principles as rules of inference.

    For instance, apparently people are inclined to accept that the gambler's fallacy is in fact valid if someone they think is a scientist has "explained" why it's a good form of reasoning. So take somebody who's just heard this misleading testimony from a "scientist", and who commits the gambler's fallacy. We might conceive of him as reasoning in the following way:

    1. Double sixes haven't come up in a really long time, so:
    2. The next roll is likely to come up double sixes.

    Alternatively, we might conceive of him as reasoning like this:

    1. Double sixes haven't come up in a really long time.
    2. The probability of double sixes increases each time they fail to come up. So:
    3. The next roll is likely to come up double sixes

    The former treats the person's acceptance of the gambler's fallacy as the acceptance of a rule of inference. The latter treats it as the acceptance of some new explicit premises. Neither way of representing the person's reasoning makes him seem more or less excusable to me, and I'm inclined to say the same thing about the moral case.

    If I had to defend that gut reaction, I'd say that just as one can be excused for accepting a false premise, one can be excused for using a defective rule of inference. Furthermore, it's not obvious to me how one would motivate the idea that false (esp necessarily false) premises can be excusably accepted, but defective rules never can be excusably used.

    Incidentally, as I'm sure you're aware, Gideon Rosen has written some great stuff on moral ignorance and excuses.

  3. I totally wasn't aware of anything by Gideon Rosen, nor did I even know the name. That's why I subscribe to this blog, cuz it rocks and because I get to learn new things and new names. =P

    Upon reflection, I agree with you on this question. In fact, I would go so far as to say that there are no salient differences between the two presented ways of explaining the gambler's reasoning. The first is pretty much an enthymeme in which the omitted premise is the normative "premise" that you talk about. The "accepted rule of inference" is just the acceptance of the truth of the second premise.

  4. It makes a difference whether we interpret the first argument as enthymematic (with a suppressed premise 2) or as using a genuinely different rule of inference. For I take it that there's at least a possible scenario in which the second premise is true (suppose the dice are fixed like that), whereas the gambler's fallacy would never be a rationally legitimate rule of inference (i.e. to directly employ. The inference must instead be mediated, at least implicitly, by the contingent premise about the quirky dice). Or am I drawing an artificial distinction here?

  5. I think the distinction Richard's drawing is a useful one in general, but it won't help here. If the idea is that the pattern of reasoning that uses the false premise is only contingently bad, while the one with the fallacious rule of inference is necessarily bad, that may apply to the case as I described it, but I don't think it's general enough to explain why, in the moral case, one way of understanding the reasoning has it come out as excusable but the other doesn't.

    Presumably, normative claims about what natural facts give rise to what normative reasons will themselves be necessary. So we can flesh out our description of the moral case so that the person is using a false normative premise (rather than a bad rule of inference), but in which there is no possible situation in which that normative premise is true.

    The distinction between patterns of reasoning that are contingently bad and ones that are necessarily bad is useful, but it doesn't map onto the distinction between patterns with false premises and patterns with defective rules of inference, or onto the distinction between excusable and inexcusable patterns of reasoning.

    Mister Integer: I didn't mean to imply that you should know about Rosen's work. I meant to be referring to Richard, though I should have said so. I said I was sure he was aware because he's a graduate student at Princeton, and if he's interested in these issues, he'd likely know about the work of Gideon Rosen, who's also at Princeton.

  6. * Yeah, I had Gideon (Rosen) in mind as someone with the opposing view.

    * I didn't mean to suggest that contingency was in play in the moral case. Indeed, I've previously suggested that the non-contingency of (basic) normative claims explains why we might treat them differently from factual claims when it comes to assessing "evidence-relative" rational oughts.

    * I should clarify that in the moral case, it doesn't seem like there are two distinct, genuinely possible forms of reasoning here, one excusable and one not. (What would the psychological difference between the two people be?) Rather, the suggestion is that of the two candidate interpretations, we may think that only the first is the right way to conceptualize (any possible) moral reasoning.

    Why think this? It does seem a little ad hoc (unless we're willing to say that non-contingent propositions can never serve as premises). But recall the intuitive datum that moral ignorance, unlike factual ignorance, reveals a fault internal to the agent's deliberative process. How else might we go about explicating this idea? (Suggestions welcome!)

  7. Whether or not the error is in a rule of reasoning or a premise of reasoning, there are surely grounds to criticize the character of someone who is so morally ignorant to think that poisoning people is ok. Namely, any one without a wholly vicious character would think poisoning people is bad. I don't see why this is undermined if the person is ignorant of rule of reasoning or a premise- either way he's vicious, and criticizably so.


Visitors: check my comments policy first.
Non-Blogger users: If the comment form isn't working for you, email me your comment and I can post it on your behalf. (If your comment is too long, first try breaking it into two parts.)