Abstract
What is required for an action to promote the satisfaction of a desire? We reject extant answers and propose an alternative. Our account differs from competing answers in two ways: first, it is contrastive, in that actions promote the satisfaction of desires only as contrasted with other possible actions. Second, it employs a notion of expected fit between desire and world, defined as the weighted sum of the fit between the desire and the world in all possible outcomes, where each weight is given by the probability of the agent’s obtaining the relevant outcome. According to our proposal, then, an action promotes a desire when the expected fit for the desire given that the agent performs the action is greater than the expected fit of the desire given that the agent performs the contrasting action. We highlight this account’s attractive features and explain how it improves on its competitors.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Examples abound. See, for just two instances, Korsgaard (1996) and Scanlon (2015). As Snedegar (2014) points out, even authors not explicitly interested in instrumental reasons—or even desires—seem to appeal to the idea of promotion in stating their views. (Compare Snedegar 2014, pp. 46–47). Here we won’t be concerned with the details of the relationship between promotion and instrumental reasons, or promotion and other possible objects of promotion (such as values, or objectives, or ends); instead, with others in the literature (e.g. DiPaolo and Behrends 2015; Lin 2016), we’ll take there to be some plausible relationship between reasons and promotion and we’ll treat the promotion relation as independently interesting. We’ll also focus our attention in what follows on promoting desires (rather than, say, ends or values), but nothing important hangs on this focus. For more on the relationship between reasons and promotion, see the discussion in Kolodny (forthcoming) and Maguire (2016).
Schroeder (2007, p. 113).
C.f. their (2011, 2016).
Behrends and DiPaolo (2011, p. 2). In order for this counterexample to work, the counterfactual statement at the end of the example must be understood as implying that Debbie’s probability of pushing B given that she doesn’t push A is 1.
Let SD be the proposition that Debbie’s desire is satisfied, and let PA and PB be the propositions that Debbie presses A and B respectively. Now, it is theorem of the probability calculus that
-
pr(SD|¬PA) = pr(SD|PB) pr(PB|¬PA) + pr(SD|¬PB) pr(¬PB|¬PA)
In Buttons, it is meant to be stipulated that pr(PB|¬PA) = 1 (see the previous footnote), and thus that pr(¬PB|¬PA) = 0. It follows from the theorem that pr(SD|¬PA) = pr(SD|PB).
-
For an extended discussion of this counterexample to Schroeder’s account, see Behrends and DiPaolo (2011, p. 4) and Sharadin (2015, pp. 1373–1375). For one targeting Coates’s view, a version of which we’ll discuss in more detail below (Sect. 3.1), see Sharadin (2015) and Behrends and DiPaolo (2016).
Some, for instance Lin (2016) and Fischer (2017), think that when one’s desire is certain to be satisfied no matter what one does (e.g., in Debbie’s case), nothing one can do promotes the satisfaction of that desire. Hence these philosophers think that purely probabilistic accounts are not subject to Debbie-style counterexamples. We think this intuition too firmly (indeed, in Fischer’s case, explicitly) ties one’s view about promotion to one’s view about reasons. In any case, below we’ll argue that probabilistic accounts face other difficulties too; so, for the sake of argument, we can admit that these counterexamples are not decisive.
While Lin himself doesn’t include one, we think it’s natural to suppose that Lin’s view should include a restriction on the relevant times prior to (t1) and after (t2) the agent’s action. Such a restriction would, we think, naturally involve the existence of the desire at t1. To see why, suppose we do not require that t1 be a time when the agent’s desire already exists—or, if you think desires can be satisfied despite no longer existing, has already existed. Now suppose I do not now (at t1) desire a fig newton, but that, unless I’m informed of my serious allergy to figs, I will desire a fig newton in 6 days. Finally, suppose that, no matter what happens, I will be given a fig newton in 7 days. Then it turns out that, given Causal Probabilism (without some restriction on t1), actions I might perform in the next several days that make it less likely that I’ll be informed of my serious allergy to figs—for instance, locking myself in a room—promote the satisfaction of my (future) desire for a fig newton. This is because such actions make it more likely that my desire for a fig newton will come into being (on day six), and hence subsequently be satisfied (on day seven). But, intuitively, what locking myself in my room causally affects is something about which desires I’ll have, not which desires of mine are promoted. The counterintuitiveness of this result appears even greater if we invoke the standard connection between reasons and desire-promotion, since that would seem to imply that I have a reason, at t1, to lock myself in my room (since doing so would promote the satisfaction of my future desire). In any case, nothing in what follows depends on this issue. (Thanks to an anonymous referee for urging clarity on this point.)
Lin (2016, p. 4). Here we precisify Lin’s account slightly. Lin’s official version is: S’s doing A promotes p iff it makes p more likely to obtain than it was prior to the occurrence of A. Here we are focused on the case of desire promotion. We also make explicit that Lin’s talk of “making more likely” should be understood in causal terms. In personal correspondence Lin has confirmed that the version of his account we give here is faithful to his intentions with respect to the account.
Here, prt1(-) and prt2(-) are the relevant probability functions at t1 and t2 respectively.
Here, and in what follows, we use ‘dispromote’ as the opposite of ‘promote’.
Lin appears to be aware that his view generates these sorts of results, and is aware that they appear problematic (see especially Lin 2016, pp. 18–19). Although he’s aware of these problems, he doesn’t offer a solution to them, saying instead that accepting Causal Probabilism would amount to “biting a bullet” (p. 19) on these points. Below, we offer a solution on his behalf (see footnote 18).
Snedegar (2014). Here we follow Snedegar’s proposal about how to understand the structure of contrastivism, though note that our interest in contrastivism concerns the promotion relation rather than the reason-for relation, which is that target of Snedegar’s analysis. While we find Snedegar’s proposal regarding contrastivism about reasons compelling, our account of promotion here is strictly speaking neutral on the nature of the reason-for relation. With that said, we do note that our contrastivism about promotion fits nicely into Snedegar’s contrastivism about reasons, given a very modest assumption about how the two should be linked. Snedegar analyses what it takes for there to be a reason for an agent to φ rather than ψ in terms of whether φ-ing or ψ-ing better promotes one of the agent’s desires. But what is it for φ-ing to promote a desire better than ψ-ing does? We submit that this is (at least extensionally equivalent to) its being the case that φ-ing rather than ψ-ing promotes the desire. If so, our account here in effect is an account of the promotion comparisons that Snedegar appeals to in his contrastivism about reasons.
C.f. Snedegar (2014, esp. p. 52 and following).
Causal Probabilism can also be given a contrastivist spin, roughly by contrasting how much two (or more) actions cause the probability of satisfying a desire to increase. Somewhat more precisely:
Contrastive Causal Probabilism: For time t1 prior to an agent A’s φ-ing or ψ-ing, and a later time t2 after A’s φ-ing or ψ-ing, φ-ing rather than ψ-ing promotes the satisfaction of D iff Iφ > Iψ, where Iφ is the increase in probability of satisfying D between t1 and t2 that is (or would be) caused by φ-ing and Iψ is the increase in probability of satisfying D between t1 and t2 that is (or would be) caused by ψ-ing.
However, it is worth noting that this account involves a comparison between different worlds (viz. worlds in which A φs and worlds in which A ψs), and thus arguably betrays Lin’s (2016, pp. 4–5) central idea that promotion should only be based on a comparison of probabilities at different times in the same world. Partly for that reason, we will not be exploring the plausibility of Contrastive Causal Probabilism in what follows (though we invite others to do so).
Or: set of multiple actions. Since this complication won’t be important in the arguments to come, we ignore it in what follows.
For more on contrastive reasons, see Snedegar (2014, esp. p. 53 and following; 2013).
It is perhaps worth noting that Contrastive Pure Probabilism also entails the following about Debbie:
-
(3*)
Debbie’s pressing button A rather than not pressing button A does not promote the satisfaction of her desire.
This might initially seem like a puzzling or even counterintuitive result. However, this reaction is based on a confusion. Recall that it is stipulated in this case that Debbie would press button B if she does not press A. So, in this case, pressing B and not pressing A are just two different descriptions for the very same course of action. It follows, then, that there is no difference in this case between pressing A rather than B, on the one hand, and pressing A rather than not pressing A, on the other hand. Thus (3) and (3*) stand and fall together. Since (3) is clearly true, the same goes for (3*).
-
(3*)
There is a range of interesting features of contrastivism about promotion that we won’t have the space to address here, e.g. how the contrast class is to be specified in a given case. Happily, the use to which we’re putting contrastivism here doesn’t depend on any particular answer to such questions. So, for reasons of space and ease of exposition, we put off discussion of them to another time, and, again, point the interested reader to Snedegar’s (2013, 2014) work on how to deal with issues of this kind.
Interestingly, this helps explain why the non-contrastive probabilistic accounts discussed above, such as Finlay’s, take the baseline to be not-φ-ing. It helps explain, too, why there might be disagreement among probabilistic accounts over that baseline: Schroeder’s account of the baseline (the agent’s not doing anything at all) might be thought to be an alternative, equally good candidate for what is (sometimes, at least) a conversationally salient contrast class.
In what follows, although Sharadin is one of the authors of this present essay, we continue to refer to Sharadin in the third person, in order to avoid confusion between the previous and the present work.
Sharadin (2016, p. 1).
Either EA is satisfied, or EA is not satisfied. If EA is satisfied, then EA is not satisfied; if EA is not satisfied, then EA is not satisfied. So, no matter what, EA is not satisfied. If some desire is not satisfied no matter what, then the probability that it is satisfied is (always) 0.
Note that Sharadin’s worry is not just a worry for Contrastive Pure Probabilism. It’s also a worry for Contrastive Causal Probabilism. Since, by hypothesis, the probability of EA cannot increase, nothing can, by hypothesis, be a cause of that probability increasing.
C.f. Lin (2016, p. 16): “I agree that on the most natural way of filling in the details of the case, there is intuitively a reason for the agent” to take the pill.
Lin (2016, p. 15).
Sharadin (2016, esp. p. 4 and following). As an anonymous referee points out, Sharadin’s reply relies on two related contentious assumptions, viz. that (1) there must be a reason for the agent to have EA* and (2) the reason there is for the agent to have EA* is explained in the same way as the reasons there are for the agent to perform various actions. Again, we lack the space to adequately address these issues here; moreover, as we explain, even if we stipulate that there are no neighboring desires, there are issues with Lin’s account.
Lin (2016, pp. 15–16).
These two replies to Lin rely on competing intuitive judgments about cases. Of course, the usual caveats that apply to such arguments apply to this one: intuitions will vary, and readers are free to disagree with ours. But since Lin makes free use of his own intuitions in the argument, we think it’s fair to appeal to our own—and, importantly, to the intuitions of the proponent of the idea that desires it is impossible to satisfy can be promoted—in reply.
I.e. a contrastivist modification of what Sharadin calls “Fit-Promote” (Sharadin 2015, p. 1381).
This is the most straightforward way of doing it, but it is not entirely unproblematic. For notice that there might not be a single closest possible world in which the agent either φs or ψs. In that case, we would presumably have to compare the average fit of all the closest worlds in which the agent φs and ψs respectively (when there is a single closest world, the average of the fit reduces to the fit, since the average of a single number is that number itself). Since this complication is irrelevant to our objections to the disjunctive alternative that is currently under consideration, and since the issue does not arise within our own account of promotion below, we will ignore it in what follows.
Sharadin (2015, p. 1382). Although recall that Sharadin did not conceive of promotion in a contrastive manner in that paper.
Sharadin (2015, pp. 1382–1383).
Although we present these problems as targeted at Contrastive Fit-Probabilism it is worth noting that they apply with equal force to Sharadin’s non-contrastive Fit-Probability Promote.
But see Schroeder (2007, esp. Ch 7) for disagreement. Schroeder’s rejection of this idea, what he calls “proportionalism” has not been widely adopted. For criticism, see Shackel (2014), Rieder (2016), and Evers (2013). For one recent attempt to articulate a proportionalist Humean view, see Manne (2016). Here, we assume some version of proportionalism is true.
Other ways of spelling this out might exploit the ratio between pr(satisfaction of D|A φs) and pr(satisfaction of D|A ψs), or the log of the ratio between these two probabilities. These ways of measuring the degree of promotion mirror some different Bayesian measures of incremental confirmation (see, e.g., Fitelson 1999). For a somewhat different proposal regarding degree of promotion, see Manne (2016).
In this case, we could obviously have been more generous towards our protagonist, e.g. by giving her 99 chocolate bars rather than merely 49. However, we have chosen this particular number to get symmetry between the current example and the next one. In the next example, stipulating that our protagonist currently only has a 49% chance of winning the chocolate lottery preempts the response that Annie will win the chocolate lottery in the closest possible world in which Annie refuses the trade. (Thanks to an anonymous reviewer for raising this concern with the second example.)
Since an outcome is just a set of possible world with equal fit with respect to some desire, we will henceforth talk about fit between the desire and an outcome as a less cumbersome way of referring to the fit between the desire and each of the worlds in the outcome.
Readers familiar with the distinction between Evidential Decision Theory (EDT) and Causal Decision Theory (CDT) will notice that our definition of expected fit corresponds more closely to the ‘evidential’ definition of expected utility than the competing ‘causal’ definition. However, a definition of expected fit that corresponds to the latter can easily be provided by simply replacing the conditional probability pr(oi|A φs) with the counterfactual probability that if A were to φ, oi would obtain, pr(A φs > oi). Presumably, a choice between these two slightly different ways of defining expected fit will be made on the basis of the same kind of considerations that are appealed to in the choice between EDT and CDT, including notably one’s reaction to the infamous Newcomb’s Problem—‘one boxers’ will use the definition of expected fit given the main text, while ‘two boxers’ will want to modify the definition in the way we have suggested here. Here we do not wish to take an official stand on this issue, but we will for simplicity operate with the ‘evidentialist’ definition given in the main text in our discussion of examples below.
For example, consider an account on which promotion consists in causing an uptick in expected fit—a view analogous to Lin’s Causal Probabilism, with probability of satisfaction replaced by expected fit. Although we did not specify the details about the degrees of fit of different outcomes in the original Bertie case, it is not hard to see that a version of the case can be constructed in which Bertie’s beginning a sporadic training program would decrease the expected fit of his desire. Roughly, we need only assume that Bertie’s expected fit for this desire was very high to begin with, due to his being very likely to begin a regular training program and there being a high degree of fit for the most probable outcomes of beginning such a training program.
For more on this, see the discussion at the end of Sect. 2.
To see this proposal in action, consider a version of Bart’s case where he is offered the chance to to buy 10 chocolate bars at the greatly discounted price of $1. Intuitively, doing this this rather than buying nothing should promote his desire D100-bars to a much greater extent than buying only 1 bar at the original price of $1 rather than buying nothing. This result is easily obtained: Since the fit between Bart’s desire and the outcome is clearly greater when Bart has 10 bars rather than merely 1, we have that
-
F(o10-bars, D100-bars) > F(o1-bar, D100-bars).
As before (see Sect. 5.2), we have that EF(Bart buys 1 bar, D100-bars) = F(o1-bar, D100-bars). By the same reasoning, EF(Bart buys 10 bars, D100-bars) = F(o10-bars, D100-bars). We thus get:
-
EF(Bart buys 10 bars, D100-bars) > EF(Bart buys 1 bar, D100-bars)
We now subtract by EF(Bart buys nothing,D100-bars) on both sides of the inequality to obtain
-
EF(Bart buys 10 bars, D100-bars) − EF(Bart buys nothing, D100-bars) > EF(Bart buys 1 bar, D100-bars) − EF(Bart buys nothing, D100-bars)
as desired.
-
Again, as in the case of probabilistic accounts, there are other possible ways of spelling out the degree of promotion that exploit the ratio between the two expected fits, or perhaps the log of the ratio between these two values. Here we don’t take a stand on whether one of these is preferable to the others, since we don’t yet see any theoretical reason to prefer one to the others. We think investigating the question of whether there are reasons to opt for one or the other measure would be worth pursuing, but it’s not something we have the space to do here.
This is a question faced by any account of promotion that appeals to probabilistic facts. See Behrends and DiPaolo (2016) for a discussion.
In the literature, the terms ‘subjective reason’ and ‘objective reason’ are used to refer to a range of differences among reasons. For instance, some use ‘subjective reason’ to refer to the subset of objective reasons to which an agent has epistemic access, so that something cannot be a subjective reason without also being an objective reason. By contrast, our use of the terms here allows for this possibility. We don’t care to die on this particular terminological hill. Instead, we’re interested in highlighting the parallel between these two senses of a reason—whatever, in the end, it’s correct to call them—and the two possible interpretations of Contrastive Expected Fit.
References
Behrends, J., & DiPaolo, J. (2011). Finlay and Schroeder on promoting a desire. Journal of Ethics and Social Philosophy, 6, 1–6.
Behrends, J., & DiPaolo, J. (2016). Probabilistic promotion revisited. Philosophical Studies, 173, 1735–1754.
Coates, D. J. (2014). An actual sequence theory of promotion. Journal of Ethics and Social Philosophy: 1–7.
DiPaolo, J., & Behrends, J. (2015). Reason to promotion inferences. Journal of Ethics and Social Philosophy, 9(2), 1–9.
Evers, D. (2013). In defence of proportionalism. European Journal of Philosophy, 21(2), 313–320.
Finlay, S. (2006). The reasons that matter. Australasian Journal of Philosophy, 84(1), 1–20.
Finlay, S. (2014). Confusion of Tongues: A theory of normative language. Oxford: Oxford University Press.
Fischer, S. (2017). On promoting the dead certain: A reply to Behrends, DiPaolo and Sharadin. Journal of Ethics and Social Philosophy, 9(2), 1–12.
Fitelson, B. (1999). The plurality of Bayesian measures of confirmation and the problem of measure sensitivity. Philosophy of Science, 66, S362–S378.
Hubin, D. (1999). What’s special about humeanism. Noûs, 33(1), 30–45.
Kolodny, N. (forthcoming). Instrumental reasons. In D. Star (Ed.), The Oxford handbook of reasons and normativity. Oxford: Oxford University Press.
Korsgaard, C. (1996). The sources of normativity. Cambridge: Cambridge University Press.
Lin, E. (2016). Simple probabilistic promotion.” Philosophy and Phenomenological Research, online first.
Maguire, B. (2016). The value-based theory of reasons. Ergo, 3(9), 233–262.
Manne, K. (2016). Democratizing humeanism. In E. Lord & B. Maguire (Eds.), Weighing reasons. Oxford: Oxford University Press.
Rieder, T. N. (2016). Why I’m still a proportionalist. Philosophical Studies, 173(1), 251–270.
Scanlon, T. (2015). Being realistic about reasons. Oxford: Oxford University Press.
Schroeder, M. (2007). Slaves of the passions. Oxford: Oxford University Press.
Shackel, N. (2014). Still waiting for plausible humean theory of reasons. Philosophical Studies, 167, 607–633.
Sharadin, N. (2015). Problems for pure probabilism about promotion (and a disjunctive alternative). Philosophical Studies, 172(5), 1371–1386.
Sharadin, N. (2016). Checking the neighborhood: A reply to DiPaolo and Behrends on promotion. Journal of Ethics and Social Philosophy, 10(1), 1–8.
Snedegar, J. (2013). Reasons claims and contrastivism about reasons. Philosophical Studies, 166(2), 231–242.
Snedegar, J. (2014). Contrastive reasons and promotion. Ethics, 125(1), 39–63.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Sharadin, N., Dellsén, F. Promotion as contrastive increase in expected fit. Philos Stud 176, 1263–1290 (2019). https://doi.org/10.1007/s11098-018-1062-2
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11098-018-1062-2