Abstract
We often act in ways that create duties for ourselves: we adopt a child and become obligated to raise and educate her. We also sometimes act in ways that eliminate duties: we get divorced, and no longer have a duty to support our now ex-spouse. When is it morally permissible to create or to eliminate a duty? These questions have almost wholly evaded philosophical attention. In this paper we develop answers to these questions by arguing in favor of the asymmetric approach to deontic value. This approach holds that we must assign zero deontic value (a measure of weight, or stringency) to fulfilling a duty, while assigning negative deontic value to violating that duty. Taking the opposing more natural symmetric approach, which holds that fulfilling duties has positive deontic value, leads to perverse recommendations about when to create or eliminate duties. A formal proof supporting the asymmetric approach is offered. We further show that moral theories require a consequentialist component to explain why we sometimes have duties to create or maintain duties.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Examples of theorists who clearly assume that fulfilling a duty has positive value while violating a duty has negative value include Ross (2006), Portmore (2011, Section 1.4), and Lockhart (2000, Chapter 4). Other authors seem to have implicitly assumed this in their discussions of the manner in which some moral considerations can “outweigh” weaker, conflicting moral considerations bearing on the all-things-considered moral status of an action. However, alternative interpretations of these latter discussions are possible.
For purposes of this paper we shall use “obligation” and “duty” interchangeably.
Thomas Hurka has pointed out (personal communication) that A.C. Ewing also claimed that promises should be treated asymmetrically, with the agent receiving no positive credit for keeping a promise, but negative credit for violating one (Ewing 1939, pp. 9–10). James Woodward has likewise asserted that there is no reason to make a promise aside from whatever value its consequences may have if fulfilled (Woodward 1986, p. 826). More recently Johann Frick endorses the first author’s view (in Smith 1997) that fulfilling a promise does not make the world go better than never making the promise in the first place, and explores parallels between this claim and claims about the asymmetries involved in whether creating a new life makes the world go better or worse (Frick 2017, pp. 351–2).
By “professional duty” we mean a moral duty arising from her professional responsibilities.
Of course, if Stacey could either promise History to install the computers, or promise the Dean to set up the projector, then things would turn out differently: Stacey should make whichever promise at 9:00 would lead to the best-in-itself act at 10:00, and the additional value accrued from the promise would not distort the prescription for what she ought to do at 10:00. However, we are not always in a position to make a promise to perform every option available to us. Here we assume Stacey has no way to contact the dean to make such a promise.
This has been advocated to us in correspondence by Thomas Hurka.
Sam Carter, personal communication.
If, however, the individual would violate the created duty, he typically has a derivative duty not to create it in the first place. If the hunter would fail to lock up his new gun, he has a derivative prima facie duty not to buy it.
Note the implications of this position for many Good Samaritan Paradox cases.
One could cogently argue that some of these cases are ones in which the duty-bearing agent actually created his own duty. The parents could have taught the child not to throw rocks at neighbors’ houses, and Stacey could have avoided her duties by calling in sick that day. However, for expository purposes we will continue to view these as duties the agent has not created for himself or herself.
Alternative value assignments would also make the point. Suppose Stacey’s fulfilling 2/5 of her duty to History has 2/5 of the deontic value of completely fulfilling her duty to History. In that case, not explaining the situation to the dean, setting up two of the computers, and installing the projector has a deontic value of + 10 (+ 8 + 2 = + 10). This exceeds the deontic value of any of her alternatives, so this alternative assignment still delivers the wrong prescription.
There is still another possible valuation scheme which assigns 0 deontic value to all violations, and retains the total “spread” between the original positive and negative evaluations by assigning appropriate positive value to the fulfillment. For example, on this scheme, when Stacey doesn’t explain the situation to the dean, her setting up the projector would receive a value of + 16, while her failing to set up the projector would receive a value of 0. The reader can ascertain that on this scheme, too, the incorrect prescription is derived.
A referee for this journal suggests that some obligations are conditional while others are unconditional, and queries how our argument applies to the conditional obligations.
Suppose there conditional obligations, and that these conditional obligations can’t be eliminated, although individuals can work to ensure that their conditions are not fulfilled. In the referee’s example, it might be claimed that there is no unconditional obligation to care for one’s child, but rather there is a conditional obligation to care for one’s child unless one gives the child up for adoption.
How does the Symmetry theorist deal with such conditional obligations? The most natural suggestion would be that fulfilling a conditional obligation is deontically more valuable than ensuring that the antecedent of the conditional is not satisfied. In the example, caring for one’s child would be deontically more valuable than giving one’s child up for adoption and thus ensuring one is not under an obligation to care for it. However, our argument in favor of Asymmetry still works against this view, as can be seen by a variant of Tech support II.
In this variant, Tech support III, Stacey has a conditional obligation to set up the projector if requested to do so by the dean. However, she has the opportunity to contact the dean before he submits his request to set up the projector. She can explain that she expects to receive a request from the History Department shortly and cannot handle both requests. She knows that, if she explains matters to the dean in advance, he will not ask her to set up the projector, thereby ensuring that the condition for an obligation to set up the projector is not satisfied. At 9:00 Stacey has the options of contacting or not contacting the dean, and at 10:00 the options of setting up or not setting up the projector, installing or not installing the computers, or splitting her time between the two tasks. We hold that in Tech support III Stacey should contact the dean, avoid any obligation to set up the projector, and completely fulfill the History Department’s request. As the reader can verify, the numbers used in Table 3 for Tech support II also adequately model this variant case. Thus Symmetry incorrectly recommends that she should split her time rather than contact the dean. By contrast, the Asymmetry theorist assigns 0 value to Stacey’s setting up the projector whether she comes under an obligation to do so or ensures she has no such obligation. This generates the correct prescription to contact the dean (as modelled in Table 4).
Indeed, we doubt that any appeal to conditional obligations can help the Symmetry theorist here. The Symmetry theorist claims that fulfilling and violating are symmetric “around” some 0 point. Our argument can be adapted to show that it will sometimes be worthwhile to violate some other obligation, in order to reap the positive benefit of moving from that 0 point to fulfillment of the base obligation. And this is implausible.
This definition was introduced in Smith (2014, p. 20). Douglas Portmore (Portmore 2011, p. 19 n. 36) defines an act’s “deontic moral value” as “a measure of how much objective moral reason there is to perform it.” For representative references to other authors who use something similar to our notion of deontic value, see Smith (2014, p. 21, n. 20).
See Sect. 3.3. It seems likely that violating a given prima facie duty sometimes has moderate disvalue, and in other circumstances has greater disvalue (for example, violating a minor promise versus violating a serious promise). We shall not try to incorporate this complexity in the present discussion.
In evaluating cases, we will speak as if Actualism is true. However, the reader can verify that in all cases we consider, the action that the agent will take at the later time is the best course of action available at that time. In our cases, Possibilism and Actualism give the same verdicts, so the assumption is harmless.
For the first author’s earlier discussions of this see Goldman (1976, 1978).
There are other approaches to ascertaining the wisest act to perform when the agent has only probabilistic information about her actions’ upshots, but these too require weighting the value of the upshots by their probabilities. See, for example, Buchak (2013).
The most straightforward, although not the only, way to represent the strengths of these deontic values is to represent them with real numbers. We shall use this technique. On the suitability of this technique, see Ross (2006, pp. 754–55).
Throughout the paper we assume, purely for simplicity, that the functions in question are additive. The appropriateness of additivity has been cogently questioned, originally by Kagan (1988).
Many of the issues in this section were previously discussed by the first author in Smith (2014). In Black (forthcoming) the second author shows how expected value theory can be used even by hardcore non-consequentialists.
A number of theorists have argued that there are various ways of responding appropriately to values—one way is to promote them, but other ways involve adopting appropriate attitudes (for example, honor or respect) towards them or towards the entities that embody these values. One could view adopting an attitude as itself a kind of mental act. If so, one could have an obligation to perform such an act. Or one could say that although there is no direct obligation to adopt an attitude (because this is beyond our control), there is an obligation to bring it about indirectly that one has the desired attitude. These obligations would be subject to the same kinds of dynamics as the obligations to promote values through standard bodily actions. How important it is to respond appropriately in one of these ways to a value is probably tied to the deontic values of fulfilling or violating a duty to promote the value. For representative theorists who advocate this pluralistic treatment of appropriate responses to values, see Anderson (1993), Scanlon (1998), Swanton (2003) and Parfit (2011).
One might introduce the term “partial deontic value” to refer to the type of deontic value an act has in virtue of more than one, but less than all, of its morally relevant features. Our discussion will not require use of this term.
Thanks to the referee for pointing out the importance of providing such an explanation.
Note that in our earlier technique of assigning a deontic value of 0 to the fulfillment of a prima facie duty, and shifting all the difference between the fulfilment of that duty and its violation to the deontic value of violating it, we essentially valued fulfilling and violating prima facie duties by their negative regret indices.
There is a second rationale for using the negative regret index to measure ATC wrongness or rightness that holds for symmetric views as well as for the asymmetric view we are defending. First, it is important to be able to use the deontic value of an act to measure how ATC right or wrong an action is. But the measure should meet an important constraint. Whether or not an action is right or wrong must be relativized to the options the agent faces. For example, an option that might be right in one case might be very wrong in another, even though it has the same underlying value, because the agent has a much better option. But if we use the underlying overall deontic value as a measure of the act’s ATC value, and interpret negative underlying values as signaling wrongness, the action will erroneously qualify as wrong in cases that it is better than its alternatives and so ATC right. Hence we need a measure of rightness that cannot have a negative value. Both symmetric and asymmetric theorists should agree to this constraint, which supports using the negative regret index as the measure of ATC moral status.
These remarks apply to constrained forms of consequentialist principles, as well as broad ones. Our discussion here is limited to maximizing versions of consequentialism. Satisficing versions would need to be handled differently, but we will leave exploration of that issue to a future occasion. See also note 28 on supererogation.
As does the journal referee.
This is a variant on a case introduced by Swenson (2016).
It might also involve a supererogatory element, as Tony’s compensatory act would if it involved paying the neighbor in silver dollars, which the neighbor collects. (This is a variant of a case introduced by Goldman (1970, p. 4.) However, we believe that supererogation must be treated differently from standard fulfillment of duty, so will not try to incorporate a discussion of that here.
Temptation to believe that right acts can differ in their degrees of positive deontic value may arise from the fact that there are duties for which the range of acts that can fulfill the duty is vague, so that a number of different acts count as falling within the range. If I have a duty to provide a college education for my child, I can fulfill it by sending her to an inexpensive state university or sending her to an expensive private college. Either act counts as providing her with a college education. Sending her to the private college may be morally better, perhaps because it gives her a better start in life, but it isn’t better because it satisfies to a higher degree the duty to provide her with an education. (However, if my duty is to provide my daughter the best college education I can afford, vagueness of the duty is reduced, and the inexpensive state university may not be adequate.)
Lockhart (2000, Chapter 4) offers an extended defense of the concept of degrees of rightness. We find no compelling reason in this argument to accept the concept, particularly in light of our proof showing the perverse recommendations that inevitably emerge from use of the concept in assigning deontic value to duty fulfillments.
As pointed out to us by Thomas Hurka (private communication). Recall there may be special circumstances in which there would be a deontological duty to create a deontological duty. For example, one might be born with a deontological duty to create certain duties, or one might mistakenly create Duty1 to create Duty2; one would then have a duty to create Duty2.
This case was suggested by Ryan Lindsey.
Although simplicity of exposition leads us to phrase this in terms of maximizing the deontic value of the course of action, other functions may be more appropriate. We intend to leave that open.
In case you believe that fulfillment can come in degrees, \(f\) can be any particular degree of fulfillment. We would rerun the argument over and over again for each degree of fulfillment, showing that none can have more deontic value than not having the duty in the first place. We take it this casts doubt on the “degrees of fulfillment” idea, as discussed in the main text.
If other values for \(B\) are possible, such as \(v\) or degrees of violation, we explicitly disallow that \(o_{i} ,o_{j} , \ldots\) take on those values. We want to allow that \({ \succcurlyeq }_{2}\) may not be well-defined for this range of values. Writing in explicit exceptions for these values would make \(o_{i} ,o_{j} , \ldots\) useless for purposes of abbreviation. However, as remarked above, the values for violations of duty don’t play a role in our argument.
As the reader can verify, the proof would work just as well if there were a set of dimensions \(\{ A_{1} , \ldots \}\), which were together separable from \(B\), and whose defined order were linear and uncountable. Thanks to Zoe Johnson King for discussion.
The proof itself allows that fulfilling a duty is, all else equal, worse than not having it bear on you. But we consider that to be implausible.
References
Anderson, E. (1993). Value in ethics and economics. Cambridge, MA: Harvard University Press.
Anderson, E. (1996). Reasons, attitudes, and values: Replies to sturgeon and piper. Ethics, 106(3), 538–554.
Black, D. (Forthcoming). Absolute prohibitions under risk. Philosophers’ Imprint.
Buchak, L. (2013). Risk and rationality. Oxford: Oxford University Press.
Ewing, E. C. (1939). A suggested non-naturalistic analysis of good. Mind, New Series, 48(189), 1–22.
Frick, J. (2017). On the survival of humanity. Canadian Journal of Philosophy, 47(2–3), 344–367.
Goldman, A. I. (1970). A theory of human action. Englewood Cliffs, NJ: Prentice-Hall Inc.
Goldman, H. S. (1976). Dated rightness and moral imperfection. The Philosophical Review, 85(4), 449–487.
Goldman, H. S. (1978). Doing the best one can. In A. Goldman & J. Kim (Eds.), Values and morals (pp. 186–214). Dordrecht, Holland: Reidel.
Hurka, T. (Forthcoming). More seriously wrong, more importantly right. Journal of the American Philosophical Association.
Kagan, S. (1988). The additive fallacy. Ethics, 99(1), 5–31.
Lockhart, T. (2000). Moral uncertainty and its consequences. Oxford: Oxford University Press.
Parfit, D. (2011). On what matters (Vol. 1). Oxford: Oxford University Press.
Portmore, D. (2011). Commonsense consequentialism. Oxford: Oxford University Press.
Ross, J. (2006). Rejecting ethical deflationism. Ethics, 116(4), 742–764.
Scanlon, T. (1998). What we owe to each other. Cambridge, MA: The Belknap Press of Harvard University Press.
Smith, H. S. (1997). A paradox of promising. The Philosophical Review, 106(2), 153–196.
Smith, H. S. (2014). The subjective moral duty to inform oneself before acting. Ethics, 125(1), 11–38.
Swanton, C. (2003). Virtue ethics: A pluralistic view. Oxford: Oxford University Press.
Swenson, P. (2016). Subjective deontology and the duty to gather information. Ethics, 127(1), 257–271.
Woodward, J. (1986). The non-identity problem. Ethics, 96(4), 804–831.
Acknowledgements
We appreciate the following individuals for their helpful comments on earlier versions of this material: Jeffrey Blustein, Johann Frick, Jimmy Goodrich, Zoe Johnson King, James Mahon, Sam Rickless, Alec Walen, and especially Philip Swenson and the anonymous referee for this journal. Interchanges with Tom Hurka over the years on these topics have been particularly fruitful. We are also grateful to members of the audiences at the Icahn School of Medicine at Mount Sinai, the 2015 San Diego Conference on the Ethics and Law of Omissions, the Rutgers-Lund Conference of 2016, the Rutgers Break-it-Down session in 2017, the University of Michigan First Annual Alumni Conference, the Berkeley Philosophy Work-in-Progress workshop 2017, and the Humboldt University Normativity Conference 2018.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix A
See Table 9.
Appendix B
We offer a formal argument for the asymmetry of the value of fulfilling and violating duties. We do so by showing that symmetry is inconsistent with other, more plausible principles.
We will model actions as vectors of prima facie duties, \((c_{1} ,c_{2} , \ldots )\). These vectors are ordered n-tuples of prima facie duties, where n is the number of prima facie duties that the theory under consideration recognizes. The value that the vector takes on in a particular dimension, say the fifth, reflects the status of that action with respect to the fifth prima facie duty.
For convenience, we will refer to the first dimension as \(A\), the second as \(B\), and so on. Capital letters denote a dimension of the vector; lower case letters are variables, denoting the value that a particular vector takes on in that dimension; and numbers are used to differentiate those values. So \(A\), the first dimension, can take on values like \(a_{1} ,a_{2} ,a_{3}\), and so on. The fifth dimension is denoted \(E\), and it can take on values \(e_{1} ,e_{2} , \ldots\).
When \((a_{1} ,b_{1} , \ldots ){ \succcurlyeq }(a_{2} ,b_{2} , \ldots )\), the action represented by \((a_{1} ,b_{1} , \ldots )\) has at least as much deontic value as that represented by \((a_{2} ,b_{2} , \ldots )\). We require that \({ \succcurlyeq }\) is reflexive: each action has at least as much deontic value as itself.
The first dimension \(A\) represents any prima facie duty, as long as it comes in degrees. The specific kind of degree scale we need will become clear as we go forward.
The second dimension \(B\) represents a purportedly-symmetric duty. There is nothing in the formalism to distinguish created from uncreated duties, and the proof is unaffected by whether the duty is created or not. We allow \(B\) to take on the values \(f\) (for “fulfillment”) and \(n\) (for “no duty”). In a full theory, we will also want to include the value \(v\) (for “violated”), or perhaps even more values, representing how egregiously the duty is violated. For our purposes though, we only need to care about whether the prima facie duty is fulfilledFootnote 33 or whether there is no duty.
We only need to focus on the interaction between the first two prima facie duties. For that reason, it will be helpful to fix all the other prima facie duties at a single value, throughout. We denote that fixed level for the other prima facie duties as \(\vec{c}\).
These are the constraints we impose on \({ \succcurlyeq }\):
-
1.
\(\vec{c}\)-adjusted separability of \(A\) from \(B\): \((a_{1} ,f,\vec{c}){ \succcurlyeq }(a_{2} ,f,\vec{c})\) iff \((a_{1} ,n,\vec{c}){ \succcurlyeq }(a_{2} ,n,\vec{c})\) (Intuitively: all else equal, a given change in degree to the first duty always has the same valence.)
-
2.
\(\vec{c}\)-adjusted separability of \(B\) from \(A\): \((a_{1} ,f,\vec{c}){ \succcurlyeq }(a_{1} ,n,\vec{c})\) iff \((a_{2} ,f,\vec{c}){ \succcurlyeq }(a_{2} ,n,\vec{c})\)
(Intuitively: all else equal, a given change to whether you have and fulfill the second duty always has the same valence.)
For brevity’s sake, it will be helpful to introduce variables \(o_{i} ,o_{j} , \ldots\) that can take on the values \(f\) or \(n\).Footnote 34 These two constraints allow us to define new relations \({ \succcurlyeq }_{A}\) and \({ \succcurlyeq }_{B}\). We say that \(a_{1} { \succcurlyeq }_{A} a_{2}\) if \((a_{1} ,o,\vec{c}){ \succcurlyeq }(a_{2} ,o,\vec{c})\) for all values of \(o\). We say that \(f{ \succcurlyeq }_{B} n\) if \((a,f,\vec{c}){ \succcurlyeq }(a,n,\vec{c})\) for all values of \(a\). “\(a_{1} { \succcurlyeq }_{A} a_{2}\)” can be read “\(a_{1}\) is at least as good with respect to the first duty as \(a_{2}\) is”. Similarly, “\(f{ \succcurlyeq }_{B} n\)” is “fulfilling duty \(B\) is at least as good with respect to the second duty as not having it”.
-
3.
\({ \succcurlyeq }_{A}\) has degrees: \((A,{ \succcurlyeq }_{A} )\) is linear and uncountable.Footnote 35
(Intuitively: the first duty is ordered in a way that is sufficiently similar to the real numbers. In particular, there are uncountably many ways of violating this duty, ordered linearly.)
-
4.
Weak representability of \({ \succcurlyeq }\): there is a deontic value function \({\mathcal{D}}:A \times B \times \{ \vec{c}\} \to {\mathbb{R}}\) such that if \((a_{1} ,o_{i} ,\vec{c}) \succ (a_{2} ,o_{j} ,\vec{c})\), then \({\mathcal{D}}(a_{1} ,o_{i} ,\vec{c}) > {\mathcal{D}}(a_{2} ,o_{j} ,\vec{c})\)
(Intuitively: if an action has more deontic value than another, it can be assigned a higher number. We don’t require the converse of weak representability: we allow for genuine incomparability.)
-
5.
No perverse recommendations: if \(a_{1} \succ_{A} a_{2}\), then \((a_{1} ,n,\vec{c}) \succ (a_{2} ,f,\vec{c})\)
(Intuitively: suppose \(a_{1}\) is better with respect to the first duty than \(a_{2}\) is. Then it is more deontically valuable for the agent to perform an action \(x\), which is just as good with respect to the first duty as \(a_{1}\) is and has no second duty, than it is for the agent to perform an action \(y\), which is just as good with respect to the first duty as \(a_{2}\) is and fulfills the second duty. Upshot: It is never more deontically valuable to violate the first duty to a greater degree just in order to make and then fulfill a duty that wouldn’t otherwise exist.)
With that done, we can precisely formulate the symmetry and the asymmetry theses. According to the symmetry thesis, \(f \succ_{B} n\). All else equal, it is more deontically valuable to fulfill a duty than to not have that duty bear on you. Our view, the asymmetry thesis, holds that \(f \approx_{B} n\). All else equal, it is neither better nor worse to fulfill a duty than to not have that duty bear on you. The two actions are just as deontically valuable.Footnote 36
Theorem
The Symmetry Thesis is inconsistent with the above constraints.
Proof
Suppose that \({ \succcurlyeq }\) obeys the Symmetry Thesis. Fix \(a_{1}\). By symmetry, \((a_{1} ,f,\vec{c}) \succ (a_{1} ,n,\vec{c})\). By weak representability, \({\mathcal{D}}(a_{1} ,f,\vec{c}) \succ {\mathcal{D}}(a_{1} ,n,\vec{c})\). So there is some rational number strictly between them. Choose one such \(q_{1}\). This gives a function \(q:A \to {\mathbb{Q}}\).
\(q\) is injective: suppose \(a_{1} \ne a_{2}\). Since \({ \succcurlyeq }_{A}\) is linear, we either have \(a_{1} \succ_{A} a_{2}\) or \(a_{2} \succ_{A} a_{1}\). Suppose without loss of generality that \(a_{1} \succ_{A} a_{2}\). By No Perverse Recommendations, \((a_{1} ,n,\vec{c}) \succ (a_{2} ,f,\vec{c})\). By Weak Representability, \({\mathcal{D}}(a_{1} ,n,\vec{c}) > {\mathcal{D}}(a_{2} ,f,\vec{c})\).
But \({\mathcal{D}}(a_{1} ,f,\vec{c}) > q_{1} > {\mathcal{D}}(a_{1} ,n,\vec{c})\) and \({\mathcal{D}}(a_{2} ,f,\vec{c}) > q_{2} > {\mathcal{D}}(a_{2} ,n,\vec{c}).\) In all, we have
This means that \(q_{1} > q_{2}\), and so \(q_{1} \ne q_{2}\). So \(q\) is injective.
But there can be no injection from \(A\), an uncountable set to \({\mathbb{Q}}\), a countable set. Contradiction.
Rights and permissions
About this article
Cite this article
Smith, H.M., Black, D.E. The morality of creating and eliminating duties. Philos Stud 176, 3211–3240 (2019). https://doi.org/10.1007/s11098-018-1171-y
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11098-018-1171-y