symposium.gif - 4.0 K
Michael A. Smith
Australian National University
"Evaluation, Uncertainty, and Motivation"
Ethical Theory and Moral Practice, Vol. 5 2002, pp. 305-320.
Bear4.gif - 2.6 K

James Lenman, University of Glasgow

Posted 01/15/03

Non-Cognitivism and the Dimensions of Evaluative Judgement

Review of: Michael Smith: "Evaluation, Uncertainty and Motivation", Ethical Theory and Moral Practice 5(3), September 2002

James Lenman
University of Glasgow


Smith’s paper opens with the following observation about evaluative judgements. Evaluative judgements can vary in the dimensions of

(1) Certitude: the confidence a subject has in her evaluative judgements;
(2) Robustness: the stability of this confidence in the light of new information; and
(3) Importance: the strength of the relative desirabilities we impute to things.

The strength of the motivation it is rational to have in the light of an evaluative judgement covaries independently with both certitude and importance in ways which, Smith argues, his own cognitivist theory of evaluative judgement is well placed to explain. Not so for noncognitivism which identifies evaluations with desires (very broadly construed). Desires can vary in strength both relative to each other and over time: this does not seem like enough structure to accommodate all three structural features that evaluative judgements have. Suppose more structure is imported by saying that valuing something is a matter of desiring to desire it. We might then identify certitude with the strength of the second order desire and importance with the strength of the desired first-order desire. But this assignment seems arbitrary. Why is it superior to the converse assignment? There seems to be no reason. Moreover this picture contradicts commonsense. For strong second-order desires are apt always to defeat weak second-order desires whatever the relative strength of the desired desires (desired desires as such are just intentional objects and pull no motivational weight). Whereas commonsense informs us that, where our motivation is concerned, sometimes great confidence of minor importance trumps faint confidence of great importance and sometimes faint confidence of great importance trumps great confidence of minor importance. Noncognitivism is thus, Smith concludes, ill-suited to capture both the structure evaluative judgements enjoy and the motivational significance of this structure.

The central challenge is to distinguish certitude and importance. Whatever feature we say certitude is, we can then readily enough characterize robustness as the diachronic stability of that feature. So I will here only address the certitude/importance issue.

In what follows I will seek to respond to Smith’s claim that no account of what evaluative judgement is that is consistent with noncognitivism can meet his challenge. I seek to defeat his claim by counterexample, offering two such accounts that do meet his challenge. I should not be understood as claiming that these are the best accounts a noncognitivist might give or that they are true. I only wish I had enough metaethical certitude to be making claims like that.


A familiar challenge to noncognitivism, let us call it the Demarcation Challenge , is one Smith himself revives a version of in another recent paper, “Which Passions Rule?” ( Philosophy and Phenomenological Research 65(1), 2002.) The central challenge Smith lays down there consists in noting that even if all evaluative judgements are desires not any old desire is an evaluative judgement and questioning the resources of noncognitivism satisfactorily to demarcate the relevant proper subset.

One sort of noncognitivist, call him the impure noncognitivist, might urge in response that it is distinctive of evaluative judgements always to combine or couple both cognitive and noncognitive elements in their meanings. Such a position might remain noncognitivist insofar as it insisted that the descriptive aspect of meaning alone would not suffice to constitute an evaluative judgement while consistently insisting that some distinctive kind of descriptive meaning was always present and that without it no noncognitive thought qualifies as evaluative. A variant would be a form of pure noncognitivism that read evaluative judgements as possessing entirely uncontaminated noncognitive meaning but that regarded them as distinguished from just any old expressions of desire by their distinctive grounds, where those grounds might themselves combine cognitive and noncognitive commitments.

Smith anticipates the thought that noncognitive evaluative judgements might be ‘Janus-faced’ in this noncognitivism-friendly way by restricting his focus to evaluative judgements which are “intrinsic or fundamental or non-derived”, giving as an example the judgement that pleasure is desirable in itself. But this restriction will do no work for him where the sort of noncognitivist we are considering regards all evaluative judgements as implicating, in virtue of their status as such, cognitive judgements in the manner just sketched.


Now there is a familiar enough story that the noncognitivist may tell about evaluative uncertainty. According to this story, it perfectly consistent with noncognitivism that I can e.g. wonder whether abortion is wrong as what I then do is wonder what attitude I would take to abortion if I were improved in various ways. That's consistent with noncognitivism for when I speak of such improvement I thereby express desires about the kind of person I would look favourably on becoming. But it suffices to open a door to uncertainty that the specifics of the ways I would look most favourably on becoming should encompass such properties as better informed or cleverer or differently endowed with any other properties whose likely impact on my desires is not immediately transparent to me and so is something about which my certainty may vary. In debate with Smith, Simon Blackburn has tabled a story of roughly this kind. (See e.g. Ruling Passions (Oxford: Clarendon Press, 1998) chapter 8, section 5 and his response to Smith in the issue of PPR cited above.) The noncognitivist (Blackburn dislikes this word but I’ll stick with it for the sake of consistency with Smith) can harmlessly concede to Smith that what I ought to do is what I would want my (actual) self to do were I wholly rational, but save his noncognitivism by a reading of ‘rational’ that defuses the significance of the concession: my rational self is a self that has been maximally improved where ‘improved’ is understood in the way sketched above. (Indeed the noncognitivist will insist it must be so understood as more robustly realist attempts to analyse evaluative judgements in terms of hypothetical ideally rational motivation are apt ultimately to flounder either on the Scylla of circularity or the Charybdis of an implausible reductionism (See Blackburn: Ruling Passions, pp 116-119. And cf. Allan Gibbard: Wise Choices, Apt Feelings (Oxford: Clarendon Press, 1990), pp. 18-22, 183-188)).

We can imagine this story being told by way of response to the Demarcation Challenge: evaluative judgements are distinguished from the class of desires more generally as judgements expressing noncognitive attitudes of desire to their objects on the grounds that, suitably improved and ideally circumstanced, we would, we think, desire these things. The details do not matter so long as it is not transparent to us that these grounds obtain, thereby making the space we need for certitude to gain a foothold. On this account even an apparently basic evaluation such as the desirability of pleasure is not “intrinsic or fundamental or non-derived” – rather it is understood as a desire for pleasure grounded in a – defeasible - thought that such a desire would be shared by an improved version of myself.


I anticipate here an objection accusing this story of incoherence. I say, in effect, that a desire of mine that P counts as an evaluative judgement if it is based on a judgement that Improved Lenman would desire that P. However we surely want the desires of mine respecting the kind of person I would look favourably on becoming, the desires that such talk of improvement expresses, themselves to count as evaluative judgements - as surely they had better if we are to allow words like ‘rational’ onto the table in characterizing Improved Lenman? So am I not saying that an improved Lenman is a Lenman whom an Improved Lenman would like? The story says that a judgement (e.g. that it is right to keep my promises) is evaluative only if it has a certain kind of grounding but also seems to say the judgements that feature in the grounds in question (e.g. that a possible Lenman who wanted the actual Lenman to keep his promises would be a good and splendid sort of Lenman) are themselves evaluative. A regress threatens.

I don’t think this is an objection that should stop me sleeping. Suppose, in the interests of simplicity, the descriptive specification of what I count as an improvement is rather modest: it just involves full information and being very clever. My concern for what, so improved, I would desire (i.e. the actual desire, expressed in my talk, in this context, of improvement, to act in accordance with that desire of Improved Lenman) counts as an evaluative thought and not just any old desire insofar as I am confident that I would go on being pretty keen on doing what, if fully informed and very clever, I would want myself to do, no matter how well-informed and clever I got. So that, by the time I get to be maximally well-informed and clever, continuing as I do to think these properties desirable ones, I am liable to be pretty wholeheartedly pleased about holding them and want to act in their light. So insofar as I take Improved Lenman to be normatively authoritive for me, Improved Lenman, does not disagree: he is a man wholeheartedly at ease with his own desires, who wants Actual Lenman to conform to the wise and sensible (as he himself regards them) desires his cleverness and informational plenitude have led him to. So the evaluative judgement implicated in this conception of improvement establishes a test for evaluative judgements that it itself passes.

So, on this account, the thought

1. H!(Keeping my promises)

counts as evaluative because it is supported by the belief that

2. A clever and fully informed Lenman would desire that I keep my promises

coupled (and this keeps the account distinctively noncognitivist) with the thought

3. H!(Conforming with the desires of a clever and fully informed Lenman).

And 3 counts as evaluative because it is supported by the belief that

4. A clever and fully informed Lenman would desire that I conform with the desires of a clever and fully informed Lenman

coupled with itself, 3. That is holism, not incoherence: any regress is benign.


On this version of noncognitivism we meet the Demarcation Challenge by insisting that evaluative thoughts count as such by being always grounded in some thought of the form:

3’. H!(Conforming with the desires of a V Lenman). where V is some descriptive specification such that the speaker in question is willing to endorse and be guided in his deliberation by 3’, together with a thought of the form:

2’. A V Lenman would desire that I φ .

We should not however insist on any particular V. For neo-Kantians, V may be a matter of what, following Backburn ( Ruling Passions, pp. 117-119) we could call muckiness (maximal unification and coherence). For others, V may be something very different indeed. So there is a substantive normative disagreement between neo-Kantians and others but we don’t want to disqualify either from counting as making evaluative judgements.

For the response-dependent cognitivist (RDC) variant of this position the domain of evaluative thought is characterized as consisting in or as grounded in some such thoughts as:

5. To φ would be to conform with the desires of a V Lenman.

Where V is some valued attribute. But which one? If the RDC theorist insists on some specific and particular descriptive specification of V his theory starts to look too restrictive (as Ridge in his contribution to the present symposium argues at greater length) at least as an account of what makes evaluative judgements evaluative (as opposed to e.g. correct). If V is left unspecified and the RDC theorist characterizes evaluative judgements in a less restrictive way as grounded in any thought of form 5 where the specifics of V get filled in for a particular speaker by reference to which possible values of V are evaluatively salient for a that speaker, then the account is crucially incomplete until we are told what ‘evaluative salience’ comes to. This is just the gap the noncognitivist account fills so nicely with 3’.


Where does this leave us with respect to Smith’s point about certitude and importance? Well, on this account, our evaluative thoughts can vary in at least these four dimensions:

A1. The strength of my actual world desire to φ .
A2. The strength of my actual world desire to act as my maximally improved (i.e. clever, fully informed and/or whatever) self would desire that I act.
A3. The strength of the desire I believe my maximally improved self (i.e. clever, fully informed and/or whatever) would have that I φ .
A4. My confidence that my maximally improved (i.e. clever, fully informed and/or whatever) self would desire that I φ .

That is a lot of structure. But, it might be objected, it is now too much. It makes a great deal of sense to say that A4 is certitude. Fine. And that A3 is importance. Fine. So where do A1 and A2 fit in? Well, plausibly, together these desires provide the measure of how much I care about acting as my evaluative judgements dictate (both de re (A1) and de dicto (A2)). Noncognitivists are evaluative judgement internalists and consequently they will say that I had better care some about this if I'm to count as making an evaluative judgement at all. But I might not care very much. I might care about doing the things my evaluative judgements recommend less than I care about other things. That might make me imperfect but we all know that some people are imperfect in this way. It is called weakness of the will and noncognitivists are sometimes accused of being unable to make sense of it.


Of course the story I have told here is artificially neat and simple in a number of ways. For starters, I may very well have not just one thought of the form:

3’. H!(Conforming with the desires of a V Lenman).

but many that differ both in the specifics of V and in the confidence with which they are held. That confidence can then partly be explicated, as described in section 4 above, as my confidence that a V Lenman would himself endorse my conforming my actual desires to his advice qua V, but may also be a matter of my confidence that the advice of a V Lenman would cohere with the advice of a W Lenman where a W is some different specification of a counterfactual Lenman whom I am similarly disposed to take to be normatively authoritative. Thus, for a religious believer, V might be “fully informed” and W might be “fully conforming my will to God’s commands”, but different religious believers may both have differing degrees of confidence that a fully informed adviser would share their belief in (or perhaps their admiration for) God and differing degrees of confidence about how bothered God is that their actions be guided by a full grasp of the facts (perhaps there are some facts to which it is better for believers not to be exposed). For some agents whose evaluative judgements have what is surely a very rare degree of structured organization there may be some one master specification X to which they accord an overriding regulative role. In that case, for any V other than X such that such an agent valued the putative advice of a V agent their confidence in doing so would be primarily a matter of their confidence that their X self would advise them to; while their confidence in valuing the advice of their X self would be primarily a matter of their confidence that their X self would himself endorse their doing do. Even that confidence might very well fall short of certainty: perhaps to be X is to come to think that X-ness is not all it was cracked up to be!


As an alternative, possible noncognitivist story, return to the possibility Smith criticizes of understanding evaluative judgements as second order desires. But again let us refine this proposal in a way that is somewhat friendly to him. Let us say that for me to make an evaluative judgement that I ought to φ is for me to desire to desire to φ on the grounds that suitably improved and in ideal circumstances that is what I would want myself to desire. Here again I may be altogether unsure what set of first order desires would maximally so conform. This gives us:

B1. The strength of my actual world desire to desire to φ .
B2. The strength of my actual world desire to desire as my maximally improved (i.e. clever, fully informed and/or whatever) self would desire me to.
B3. The strength of the desire I believe my maximally improved (i.e. clever, fully informed and/or whatever) self would have that I φ .
B4. My confidence that my maximally improved (i.e. clever, fully informed and/or whatever) self would desire that I desire to φ .
B5. The strength of the first order desire to φ I believe my maximally improved (i.e. clever, fully informed and/or whatever) self would desire me to have.

Perhaps now we have too much structure. Surely real-life evaluative thoughts are not that complicated. But a little reflection should satisfy us that they are. I propose the following assignments. B4 is certitude. Plausibly B3 is importance for the same reason A3 was. And plausibly B1 and B2 play just the roles I just characterized A1 and A2 as playing. But what about B5? This, I would suggest, is importance too but it is importance of a rather different sort. Consider an example. Say I think that I ought to desire certain bodily pleasures but I should not desire them too much. I can nonetheless envisage that my maximally improved self would want very badly that my desires take on this moderate form (so high B3-importance). I may say, echoing Aristotle (NE, 119a, 6-12), that bodily pleasures are important in that someone who didn't want them at all would be inhuman and insensible and that would be awful. But that's consistent with thinking them not very important in that my maximally improved self wouldn't want me to want them all that much (so low B5-importance). Importance surely really does have these two quite distinct dimensions and it's no embarrassment to take them on board. That said in many cases they tend naturally to coincide and when they do we can harmlessly run them together without clearly distinguishing them – my maximally improved self might want very badly that I want very badly not to kill people. Here saying simply that not killing people is very important serves our purposes and there is no need to disambiguate these two species of importance. The sort of noncognitivism then gives us all the structure Smith insists on and more besides, where the more besides has real work to do in capturing distinctions that feature in commonsense moral thought.


Commonsense informs us and Smith reminds us that, where our motivation is concerned, sometimes great confidence of minor importance both does and should trump faint confidence of great importance and sometimes faint confidence of great importance both does and should trump great confidence of minor importance. Is this a problem in the context of the noncognitivist stories I have told here? Not at all. On both these stories, when I make evaluative judgements I am keen that my actions conform to a certain condition – one concerning the relevant desires of an improved version of myself. The importance of satisfying that condition is a matter of degree as that improved version of myself may have desires of greater or lesser strength. Of course both Improved Lenman’s desires and the desires Improved Lenman desires me to have are mere hypotheticals that pull no motivational weight in their own right. But insofar as I have desires like A2 and B2 the judgements I make about them are reliably apt to impact on my motivation in just the way needed to meet Smith’s challenge. The thought that Improved Lenman e.g. cares a great deal about my being kind to small children and animals is then apt to result in a commensurate actual concern. My confidence that that my judgement about the specifics of what that condition demands is correct is also a matter of degree. I am quite sure Improved Lenman would like me to quit smoking. Does he think it is high time I was married? Search me. Again insofar as I have desires like A2 and B2, this difference in certitude will readily impact on my actual motivations. When a greater degree of certitude pulls me one way and greater importance pulls me the other, both the stories considered above are entirely consistent with ultimate victory, both de facto and de jure, for either side.

Smith makes some important observations about the structure of evaluative thought and challenges noncognitivists to do them justice, urging that we cannot. I have argued that we can.

(I would like to thank Michael Smith for helpful feedback on earlier critical thoughts about the paper here discussed and James Dreier and Michael Ridge for useful comments on earlier drafts.)

back-symp.gif - 1.6 K