The Recognitional View of Instrumental Rationality

The "recognitional view" of practical rationality – as I shall use the term – is the view that all
requirements of practical rationality are justified by the fact that it is reasonable to expect
one’s conforming to these requirements to lead to one’s practical reasoning’s getting things
right
to a higher degree than one’s violating these requirements. If R is a requirement of
practical rationality, then it must be reasonable to expect the results of one’s conforming
to R to be better – specifically, to involve one’s practical reasoning’s "getting things right" to
a higher degree – than the results of one’s violating R.

Can this recognitional view explain what justify the requirements of instrumental rationality
(i.e., roughly, the requirement that we should take appropriate means to our ends)?  As I shall
argue, it can.  Once we have a plausible view of what it is for one to "get things right" in
one’s practical reasoning, this recognitional view will have no difficulty explaining the
requirements of instrumental rationality.

The crucial point is that to "get things right" in one’s practical reasoning, it is not enough for
one to make fine decisions or choices.  Good intentions are not enough.  One’s practical
reasoning must also lead to one’s acting as one should act (all things considered), through
one’s executing the decisions or intentions that one has made.  It is part of the point of practical reasoning, on this
view, to get things done  – specifically, not just to get any old things done, but to get the right sort of things done.

In short, let us say that one has "got things right" in one’s practical reasoning about how to
act in a given situation S if and only if:

  1. one will actually execute all the intentions (if any) that one has about how to act in
    situation S; and
  2. one will thereby act as one should act in S.

So, given this account of what it is to "get things right" in one’s practical reasoning, your
practical reasoning about how to act in the relevant situation has not "got things right" if you
intend A but fail to execute that intention.  (Of course, if intending A is a thoroughly bad
intention, then your practical reasoning has also not got things right if you intend A and
succeed in executing this intention; in this case, the only way for your practical reasoning to "get things
right" in every way is for you not to intend A at all.)

Now let us put this conception of what it is for one’s practical reasoning to "get things right"
together with our "recognitional conception" of practical rationality.  This leads to the
following picture of practical rationality: The (possibly empty) set of intentions that you have about how to act in a given situation S is rational if and only if it is reasonable for
you to expect
that (1) you will execute all of these intentions (if any) in S, and that (2) you will act as you should in S.

According to this approach, then, practical rationality requires that one should not form any
intention if it is not reasonable for one to expect that one will execute that intention. E.g.,
suppose that all my evidence indicates that if I intend A, I will just give in to temptation and
abandon my intention at the last moment.  So if I intend A, it will be unreasonable for me to
expect that I will execute this intention. Thus, it is a requirement of practical
rationality that I should not intend A, when it is unreasonable for me to expect that I will
execute this intention. (It is clear that this approach can also explain why it is irrational to have inconsistent intentions.)

Now, suppose that you intend to achieve an end E, and you also know that if you do not take
means M, you will not achieve end E.  Suppose that you also know that you will not take
means M unless you intend means M.  Then if you do not intend means M, it will be
unreasonable for you to expect that you will execute all the intentions that you have – in
particular, it will be unreasonable for you expect that you will execute your intention to
achieve end E.  So, it seems, you will be violating a requirement of rationality.

In this way, the recognitional view can explain the basic requirement of instrumental
rationality – the wide-scope requirement that you should be such that whenever you intend E, and know that you will not achieve E unless you both have and execute the intention to take the
means M, then you will also intend the means M.

So far, this only explains why rationality requires that you intend the known necessary means
to your ends.  But what if there are no means that you know to be strictly necessary for the
end?  It seems plausible that in this case, instrumental rationality requires that you intend
what you believe to be the best available means to your end (or at least what you
believe to be one of the optimal available means to your end).  The recognitional view can
explain this too – at least given the assumption that it is unreasonable for anyone to believe
that it can count as "acting as one should" to achieve an end through the use of suboptimal
means.

Suppose that you intend end E, but you do not intend to achieve end E through what you
believe to be optimal means: instead, you either have no intention about which means to use;
or else you intend to achieve E by using what you yourself believe to be suboptimal means.
Then it will be unreasonable for you to believe that your practical reasoning will "get things
right": either it will be unreasonable for you to expect that you will execute your intention to
achieve end E, or it will be unreasonable for you to expect that you will act as you should in
that situation – since it is unreasonable for you to believe that achieving end E through the use
of suboptimal means can be an instance of acting as you should in this situation.

In general, this approach can explain why it can sometimes be irrational not to form any intentions at all about how to act in a situation S (say, if one sits around dithering while disaster strikes).  In these cases, although it is reasonable for one to believe that one will execute all the intentions that one has about how to act in that situation (viz., no intentions at all), it is not reasonable for one to believe that one will act as one should in that situation.  In such cases, it is only if one forms some intentions about how to act that it can be reasonable for one to believe that one will act as one should.

At the same time, this approach does not imply that it cannot sometimes be quite rational to form no intentions at all.  Suppose that you reasonably believe that you
will act as you should act in situation S even if you form no intention at all about how to act
in that situation.  Then it is quite reasonable for you to believe that you will execute all the
intentions that you have with respect to that situation (viz., no intentions at all), and that
you will thereby act well.  So in cases of this kind, it can be quite rational not to form any intentions at all.

In this way, I think that this recognitional view of practical rationality looks like a promising
approach to explaining the requirements of instrumental rationality.

Note: The argument that I have given here is an expanded version of the argument that I gave very briefly, towards the end of my paper "Choosing Rationally and Choosing Correctly". I hope to refine and expand this argument in greater detail and precision before long.

19 Replies to “The Recognitional View of Instrumental Rationality

  1. Ralph,
    A very interesting post. I’m a bit puzzled, though, by your suggestion that the recognitional account explains the wide-scope view of instrumental rationality; it seems, instead, that your view conflicts with it. Suppose I recognize that by doing E, I would not be getting things right. Nevertheless, I akratically intend to do it anyway. Suppose, further, that M is a necessary means to E. Should I intend to do M, given that I intend to do E? Not according to your view, if I understand it: by doing M, I would be doing E, and hence not getting things right. But a wide-scope account seems to give a different answer: perhaps I have already violated a norm of practical reason by intending to do E; but if I then fail to intend to do M, I will have compounded my irrationality by violating the wide-scope norm. In short, your view seems to suggest that there is no reason for the akratic to do M, whereas the wide-scope account seems to suggest that there is.
    But maybe I’m reading too much into your view. If, *with respect to instrumental reasoning*, “getting things right” merely amounted to “successfully executing one’s intentions”, then the conflict between your view and wide-scopism would vanish. For in that case, even if doing E wouldn’t be getting things right, doing M would get them right *with respect to instrumental reasoning*. So both this version of the recognitional view and wide-scopism would say that the akratic has a reason to do M. But I’m assuming that this isn’t your version of the recognitional view.

  2. Thanks, Allen!
    I think that you’re misinterpreting the wide-scope view. The wide-scope view does not entail (as you seem to suggest) that in the case that you describe, the akratic person must intend M. It entails only that rationality requires that the akratic person must *either* cease intending E, or else start intending M. I.e., what the wide-scope view forbids is the *combination* of intending E and failing to intend M.
    However, my view *does* imply that if the akratic person both intends E and fails to intend M then he has in one way “compounded [his] irrationality”. Formerly, his intentions (about how to act in situation S) were irrational because they made it unreasonable for him to expect that he would act as he should (in S). Now they are irrational for another reason: it is now unreasonable for him to expect that he will execute all of the intentions that he has about how to act in S. (On the other hand, in another way, he may actually have mitigated his irrationality somewhat, since it now may be more reasonable for him to expect that he will in fact act as he should in S!)
    At all events, I’m not working with the notion of “getting things right *with respect to instrumental reasoning*”. As I’ve defined it, to get things right in one’s practical reasoning (about what to do in S), you must *both* (1) execute all the intentions that you have (about what to do in S), *and* (2) act as you should (in S). Both of these conditions are necessary: neither is sufficient all by itself.

  3. Ralph,
    I agree that the wide-scope view only forbids the combination of intending E without intending M; that, of course, is the point of calling it wide-scope! On that view, the sheer fact that you intend E does not entail that you should intend M.
    Nevertheless, suppose both Al and Bob recognize that E would be wrong, yet they akratically intend to do it anyway. And suppose that M is a means to E, and that Al intends M whereas Bob does not. The wide-scope view does not entail that Al and Bob should do M; on the contrary, both should give up the intention to E. But as far as I can tell, the wide-scope view does entail that Bob is less rational (i.e., violated more norms) than Al, since he violated the wide-scope norm and Al did not. Specifically, each should give up E, and neither did so. Also, on the wide-scope view, each should see to it that: if he intends E, he intends M. Al did this, and Bob did not.
    If your view explains the wide-scope view, then it too should entail that Bob is less rational than Al (in the above sense). Granted, Bob failed to execute his intentions in S, and so violated (1). But perhaps he thereby did what he should do in S (let’s suppose so), and so adhered to (2). Al, on the other hand, failed to do what he should do in S, and so violated (2). But he did manage to execute his intentions, and so adhered to (1). I don’t see how your view entails that Bob is less rational (i.e., violated more norms) than Al. Since your view and the wide-scope view have different entailments here, I don’t see how they’re compatible.
    I certainly don’t mean this to be an objection to your view. I can understand the intuition behind saying that Bob mitigated his irrationality somewhat. But I can also understand the intuition (captured, I think, by the wide-scope view) that Bob is less rational. So I guess my objection – or at least uncertainty – is wide-scope: it’s an uncertainty regarding the combination of your view with the wide-scope view.

  4. Ralph,
    A couple of questions that occur to me.
    1. Suppose that I can reasonably expect that if I engage in instrumental reasoning about optimal means to my ends, the length and difficulty of it will wear me out and, when I get around to acting, I will flub it. Or maybe I will run out of time and not be able to execute all my intentions. So I decide to satisfice: I take the first sufficient means I can think of, and accomplish my end. Have I acted rationally?
    2. Suppose that I can reasonably expect that I do not have enough information to know whether any of the various actions I have available will be sufficient means to my end. (I am trying to talk a bureaucrat into giving me a grant.) I just take an educated guess about means and act. I succeed. Have I acted rationally?
    3. Suppose that I can reasonably expect that the process of instrumental reasoning about an end will undermine my accomplishment of it. (For example, suppose I want to respond to my lover with spontaneous affection; if I think about it too much, I’ll fail. Or I need sleep, but thinking about how to go to sleep keeps me awake.) I keep my end, but avoid reasoning instrumentally about it, deciding that my best bet is to just do what comes naturally as circumstances arise. Have I acted rationally?
    4. Suppose I am trying to wean myself off an addiction to tobacco. I figure I can only do this by intending not to smoke when the desire arises. I can reasonably expect that I will fail in my intentions many times. Nevertheless I form the intention not to smoke each time the desire to smoke arises, and as expected, fail the first fifty times. But I keep trying; and eventually I break my addiction. Have I acted rationally?

  5. Small worry about this wide-scope thing and the following discussion:
    “the basic requirement of instrumental rationality – the wide-scope requirement that you should be such that whenever you intend E, and know that you will not achieve E unless you both have and execute the intention to take the means M, then you will also intend the means M”
    I wonder if there are even cases where one violates instrumental rationality thus formulated. In the end you accept that it can (contrary to a literal reading of the wide-scope principle) be rational not form any intentions about how to act, if you can expect to act as you should. This can be the case before the time of needing to take the means if you intend to start intending to take the means just at the last moment before the action or if you have past evidence that you usually start intending to take the means just at the last moment.
    Now, come the very last moment and you realise that you are still not intending to take the means. If you now know that you won’t be able to use the means and thus believe that you cannot achieve the end, it’s not clear that we would attribute an intention to pursue the end either.
    So, is there a milli-second at some point where you in fact are in the violation of the principle of instrumental rationality? It’s hard to pinpoint down. Also as a bit of advertisement, my friend Jeppe Anderson did a great thesis here in Reading on instrumental rationality which anyone working on the topic should read!

  6. Jussi’s comment brings to the surface what had been implicit in Heath’s examples, namely the thought that any characterization of specific case of instrumental irrationality is inherently contestable.
    What their comments press for is some basis for attributing ‘ends’ to an agent that he does not, in fact, pursue. For with respect to the ends an agent does pursue, he cannot do so but through the means he believes will (best) ensure them. Any judgment on the agent’s part that the means taken were insufficient (or suboptimal or not necessary, or what have you) must necessarily be retrospective, but such a judgment does not support the conviction that he (instrumentally) reasoned (and acted) irrationally, but rather that he falsely believed things about his circumstances for acting.
    It would seem that if the idea of violable requirement of instrumental rationality is to have any bite, it should be possible for an agent to conceive of himself as (that is, to understand himself to be) violating it. I don’t think that is possible. Clearly he cannot violate it intentionally. Can he do so inadvertently?

  7. Ralph,
    I may well have missed it, but I was not sure if you were saying that an act is rational if it is reasonable to expect that it will lead one to act in the way that is in fact the way one has most reason to act or an act is rational if it is reasonable for one to expect that acting that way will lead one to act in ways that it is reasonable for one to think is the way one has most reason to act. Since it seems that there could be reasonable disagreement about what one has most reason to do, the two interpretations come apart.

  8. Ralph,
    I may well have missed it, but I was not sure if you were saying that an act is rational if it is reasonable to expect that it will lead one to act in the way that is in fact the way one has most reason to act or an act is rational if it is reasonable for one to expect that acting that way will lead one to act in ways that it is reasonable for one to think is the way one has most reason to act. Since it seems that there could be reasonable disagreement about what one has most reason to do, the two interpretations come apart.

  9. Ralph,
    I may well have missed it, but I was not sure if you were saying that an act is rational if it is reasonable to expect that it will lead one to act in the way that is in fact the way one has most reason to act or an act is rational if it is reasonable for one to expect that acting that way will lead one to act in ways that it is reasonable for one to think is the way one has most reason to act. Since it seems that there could be reasonable disagreement about what one has most reason to do, the two interpretations come apart.

  10. David,
    I was thinking about the same question. I take it that this is the sentence:
    “The (possibly empty) set of intentions that you have about how to act in a given situation S is rational if and only if it is reasonable for you to expect that (1) you will execute all of these intentions (if any) in S, and that (2) you will act as you should in S.”
    As far as I can see, (2) belongs under the scope of the ‘reasonable for you to expect’ ‘operator’. This would make the sentence something like your second option (even if (2) doesn’t say act as you think you should). I was thinking that if Ralph doesn’t mean this, then there will be the problem of seeminly rational bad people. But, he can probably tell you better what he means. Sorry about intervening…

  11. Thanks for all these comments! In this comment, I’ll give a quick response to Allen, Jussi’s first comment, and Joseph. (Then I’ll post two more comments, with a response to Heath, and finally a response to David and Jussi’s second comment.)
    Allen – I don’t think it’s right to say that the wide-scope view implies that Bob “violates more norms” than Al, or that Bob is “less rational” than Al. Indeed, the wide-scope view seems completely compatible with the view that the whole idea of “counting norms” makes no sense at all – or even with the view that the whole idea of “degrees of rationality” (so that we can say that out of two irrational intentions, one is “less rational” than another) also makes no sense. So, at least all by itself, all that the wide-scope view implies is that Bob does violate a requirement of rationality. It doesn’t entail that Bob’s intentions are less rational than Al’s.
    Jussi – I think Joseph has in effect answered your first comment. Yes, it is hard to imagine someone completely consciously violating the wide-scope requirement. (This would be someone who consciously thinks of his intended end, and yet consciously refuses to take the means, even while consciously judging that the means are necessary for the end; and I agree that that is hard to imagine.) But it’s not at all hard to think of someone who distractedly or absent-mindedly or inadvertently violates it.

  12. Heath – Your four questions are all about whether the agents in these four types of case have “acted rationally”. But actually, I haven’t said anything about rational *action* at all, only about rational *intention*. So let’s consider instead whether the agents’ *intentions* in these four types of case are rational.
    1. I haven’t said anything about satisficing vs. optimizing here. But it is surely compatible with my view that an intention can make it reasonable for me to expect that I will “act as I should”, even though I haven’t considered all the available options, and know full well that I haven’t (indeed, even if I know that I would find a better option if I devoted unlimited resources to deliberating about this particular choice). So I could certainly say that in at least some of the cases that you describe, the relevant intention is rational.
    2. I also haven’t really said anything about uncertainty. In fact, I used the term ‘expect’ (instead of ‘believe’) to hint that I would in the end have to tell a much more detailed story about when the reasonableness of a mere *partial belief* in the chances of one’s “getting things right” is sufficient to make a set of intentions rational. But I obviously haven’t told this story here. (Note that I say that it must be “reasonable” for you to have this expectation – not that you must actually have this expectation.)
    3. I also haven’t said anything about reasoning at all. I’m certainly not claiming that instrumentally rational intentions require any special sort of instrumental reasoning. So I don’t need to worry about cases where this special sort of reasoning would be self-defeating. So, let’s focus on the case of self-defeating intentions instead – i.e., cases where merely *having* a certain intention (regardless of the reasoning that led to the intention) will guarantee that this intention is not realized.
    Now, my view is precisely designed to explain why an intention that the agent knows to be self-defeating is irrational. (If you have such an intention, then it will surely not be reasonable for you to expect that you will execute your intentions!) So, if it is reasonable for you to believe that you will succeed in realizing your end without intending any means, then it may certainly be rational to intend the end but not the means.
    4. There are several ways of filling in the story about the smoker. (a) Perhaps the smoker is trying (but at least initially failing) to be rational. (b) Perhaps the smoker is rationally intending to do something that will result in his having an irrational intention (i.e., an intention that he can’t reasonably expect to execute); it could be rational for him to intend to give himself this irrational intention because of the expected benefits of having this irrational intention. (This would make the case like the case that Parfit called “Schelling’s Answer to Armed Robbery”.) (c) Finally, perhaps the smoker does not really intend to give up smoking now, but only to *try* to give up now; then his intention could certainly be rational.

  13. Ralph,
    Should we cash out the reasonableness of reasonable expectations (that adhering to a practical rule will lead to one getting things right) partly in terms of the requirements of theoretical rationality? If so, it looks like you would explain the normativity of certain rules of practical thinking, in part, by reference to the normativity of the same sort of rules of theoretical thinking. But if all rational requirements are of a piece, there should be a single, unified account for all of them. The lack of such is especially problematic if our reasons for thinking that the recognitional view might be correct apply equally to practical and theoretical rational requirements.

  14. David (and Jussi’s second comment) – I didn’t talk about “what one has most reason to do”. I talked about “acting as one should”. What I meant by this was an “objective” notion – so that whether or not one is “acting as one should” may depend on facts about one’s situation that one does not actually know. I take it that this is what David means by speaking about “what is in fact the way one has most reason to act”.
    Unfortunately, David’s phrasing is syntactically ambiguous, and so I’m not quite sure what he means by his question. At all events, Jussi is absolutely right that I meant the phrase ‘acting as one should’ to appear (de dicto) within the scope of the phrase ‘it is reasonable for you to expect that …’. So, for it to be rational for you to intend to take a certain course of action in situation S, this intention must allow it to be reasonable for you to expect that you will *act well* in S, or in other words that you will *act as you should* in S.
    I.e., it is a normative or evaluative proposition that it must be reasonable for you to expect to be true, if the intention is to be rational — specifically, a proposition involving an “objective” normative or evaluative notion.

  15. Michael,
    You ask, “Should we cash out the reasonableness of reasonable expectations (that adhering to a practical rule will lead to one getting things right) partly in terms of the requirements of theoretical rationality?” — My answer is, “Yes”.
    However, I think that you’re wrong to suggest that this means that my view must preclude a “single unified account” of the requirements of rationality. Still, it may mean that this single unified account will in the end have to invoke some more general notion, which non-accidentally turns out to coincide with “reasonable expectation” in the case of practical rationality, but is itself ultimately to be explained in more fundamental terms.

  16. Sorry my earlier post was poorly worded. I wasn’t thinking about cases where the agent lacks descriptive info. I was thinking about cases where a person reasonably believes a false view of what would count as getting things right. So suppose someone could count as reasonably believing that Williams’ internalism is the true account of what would count as getting things right in the practical realm. In this case it would seem, for example, that rationality would not generate merely wide scope oughts in cases where one reasonably believed one informedly wanted something. More generally, what would be rational for me would be relative to my reasonable conception of getting things right and the same for you.
    I kinda like this feature of the view as I understand it, but I just want to make sure that I am understanding it.

  17. Sorry my earlier post was poorly worded. I wasn’t thinking about cases where the agent lacks descriptive info. I was thinking about cases where a person reasonably believes a false view of what would count as getting things right. So suppose someone could count as reasonably believing that Williams’ internalism is the true account of what would count as getting things right in the practical realm. In this case it would seem, for example, that rationality would not generate merely wide scope oughts in cases where one reasonably believed one informedly wanted something. More generally, what would be rational for me would be relative to my reasonable conception of getting things right and the same for you.
    I kinda like this feature of the view as I understand it, but I just want to make sure that I am understanding it.

  18. Sorry my earlier post was poorly worded. I wasn’t thinking about cases where the agent lacks descriptive info. I was thinking about cases where a person reasonably believes a false view of what would count as getting things right. So suppose someone could count as reasonably believing that Williams’ internalism is the true account of what would count as getting things right in the practical realm. In this case it would seem, for example, that rationality would not generate merely wide scope oughts in cases where one reasonably believed one informedly wanted something. More generally, what would be rational for me would be relative to my reasonable conception of getting things right and the same for you.
    I kinda like this feature of the view as I understand it, but I just want to make sure that I am understanding it.

  19. David —
    Yes, that’s exactly what I had in mind. If I have a reasonable but false view of what “getting things right” and “acting as one should” involve, then this may well make a difference to what intentions it is rational for me to have. In such cases, it can happen that I am rationally required to intend a certain course of action, even if — as a matter of fact — if I take this course of action, I will not be “acting as I should” at all.
    So, I have to recognize at least two kinds of ‘ought’ — the ‘ought’ of rationality and the ‘ought’ of correctness or “getting things right”. These two kinds of ‘ought’ can come apart, especially when the agent reasonably believes something false. (I don’t think it makes any difference whether the false proposition that the agent reasonably believes is a normative proposition or a non-normative (“factual”) proposition.)

Comments are closed.