Parfit’s Mixed Maxim Objection against the Formula of Universal Law Reconsidered

Parfit rephrases Kant’s FUL as:

FUL: It is wrong to act on maxims that we could not will to be universal laws. [15: 285]

Parfit’s understanding of FUL differs from Kant’s in that Parfit is sceptical of the idea that it is possible to identify morally bad actions via a conceivability test. He only considers what can be rationally willed as a universal law,³ and does not distinguish between maxims that fail the FUL test because they cannot be rationally willed as universal, and those that fail because they cannot even be conceived of as universal. Hence, we do not find in Parfit an equivalent to Kant’s distinction between perfect and imperfect duties (IV: 424). In what follows, we will base our discussion on Parfit’s understanding of FUL.

Parfit considers FUL to be the best candidate of those offered by Kant for a supreme principle of morality. However, Parfit claims that we have to revise the formula to avoid objections. Ultimately, we arrive at a version of the formula that is supposedly compatible with Contractualism:

“An act is wrong unless such acts are permitted by some principle whose universal acceptance everyone could rationally will”. [15: 341]

One important step on the path from FUL to Parfit’s revised version of it is to abandon a certain understanding of maxims: namely, as policies. According to Parfit, there is no problem with a conception of maxims whereby maxims describe one specific action, performed in a specific situation. These maxims would be equivalent to the term “such acts” in the revised formula just cited. According to such a conception, a maxim would list all the morally relevant properties of a specific action and of the specific situation in which the action is performed.

However, Kant himself considers maxims to be policies that can be very general and can abstract from many specific properties of concrete situations. Examples of such maxims include “increase my wealth by every safe means” (V:27) and “let no insult pass unavenged” (V:19). Such maxims are not descriptions of a specific action, since there are many different types of actions in many, very different situations that could be performed under these maxims. The distinction between two understandings of maxims, as action descriptions and as policies, will be crucial for the purposes of our paper. Parfit presents a number of arguments against the notion that maxims, understood as policies, are the proper objects of moral evaluation.⁴ Of these arguments, we will focus on the Mixed Maxims Objection, and discuss other arguments only insofar as they are relevant for a correct understanding of the nature and force of the Mixed Maxims Objection.

Parfit alleges that Kant “overlooks” [15: 293] the fact that maxims can be mixed. “Mixed” here indicates that maxims can be instantiated in both permitted and prohibited actions. When someone follows the egoistic maxim of doing whatever serves their own interests, they might pay their debts, help others, etc. Alternatively, they might steal, cheat, take more than their fair share, etc. When someone follows the maxim of improving their standing in the eyes of others, they might rescue a drowning child, donate to charity, etc. Alternatively, they might lie about their accomplishments, slander others, etc.

Parfit himself contemplates a number of different ways in which the existence of mixed maxims may constitute objections to Kant’s notion of maxims. Before we can have an informed debate about responses to the Mixed Maxims Objection, however, we must clarify exactly what is the most pressing challenge that mixed maxims pose. To this end, we will distinguish in the current section three versions of the Mixed Maxims Objection. According to the first version, FUL fallaciously condemns all actions that are performed on an impermissible maxim (a). The second version concedes that FUL does not fallaciously condemn all actions (in the sense of ‘what the agent is doing’) performed on an impermissible maxim, but in some cases  when an agent acts on an impermissible maxim, FUL fallaciously condemns the agent’s ‘doing of it’ (b). According to the third version, the problem of FUL is that it does not allow us to derive any conclusions about the permissibility of actions at all (c). We will briefly show how Kantians can avail themselves of responses to (a) and (b) respectively. Moreover, closer scrutiny of (b) will reveal why it is desirable to hold on to Kant’s notion of maxims. By contrast, (c) poses a fundamental problem for Kant that has not yet been sufficiently addressed by Kantians. We will focus on this version of the Mixed Maxims Objection for the remainder of the paper.

(a) The version of the Mixed Maxims Objection that Parfit himself most frequently employs is based on the assumption that FUL is a principle for evaluating the morality of actions. According to this assumption, FUL implies that any action performed on a maxim that does not pass the universalization test is wrong. The Mixed Maxims Objection is then to be understood as:

“When applied to mixed maxims, Kant’s formulas fail, since these formulas condemn some acts that are clearly permissible or morally required”. [15: 293; see also 16: 296]⁵

In cases like the Egoist who pays their debt on a maxim of egoism, Kant would have to conclude that paying the debt is a wrong action. Such a conclusion, however, is absurd.

This version of the Mixed Maxims Objection draws on the idea of false negatives: Kant’s formulas detect a supposedly immoral action when, according to steadfastly held intuitions, the action is not immoral. This line of arguing is akin to other well-known objections from false positives/negatives against Kant’s formulas.⁶ However, there is an important difference between the Mixed Maxims Objection and standard false positives/negatives objections. The latter usually seek to put pressure on FUL by showing how this formula rules out some intuitively permissible maxims and fails to rule out some intuitively impermissible maxims.⁷ The Mixed Maxims Objection, by contrast, focuses on how intuitively permitted actions turn out to be prohibited according to FUL, if they are performed on maxims that FUL prohibits. This difference is potentially significant, as it implies that, even if standard false positives/negatives objections could be overcome and FUL ruled out all and only intuitively forbidden maxims, it could still be the case that some actions performed on impermissible maxims strike us as morally permissible.

As already mentioned, Parfit’s central assumption is that FUL is a principle for evaluating the morality of actions. A number of philosophers agree with this assumption and also concede that, according to Kant, an action would turn out to be wrong if performed on an impermissible maxim.⁸ However, many Kantians, such as Thomas Pogge [17] and Sven Nyholm [10], believe that FUL is not supposed to say anything directly about the morality of actions. Instead, as Kant himself presents FUL (“act only according to that maxim …” IV:421) and frequently applies it to examples (e.g. IV:421-3), FUL is concerned with principles of actions or maxims and does not pertain directly to the moral status of individual actions. According to this reading, although FUL states that it is wrong to act on impermissible maxims, it does not imply that every action performed on an impermissible maxim is a wrong action, as FUL is not concerned with actions.

Parfit is right to argue that FUL would lead to implausible results if it is understood as a principle for the evaluation of actions. However, since there is an alternative reading of FUL that avoids these implausible results, and since this reading is also the one that Kant himself seems to have in mind, we should follow that alternative reading. Moreover, it is worth highlighting the fact that Parfit’s wording of FUL does not support the version of the Mixed Maxims Objection that Parfit himself most frequently employs, as he maintains that “[i]t is wrong to act on maxims that we could not will to be universal laws” [15: 285]. However, the wrongness of acting on certain maxims does not imply the wrongness of all actions performed on these maxims.

Thus, the first and perhaps most straightforward understanding of the Mixed Maxims Objection is not the strongest version of it, as it presupposes an unconvincing interpretation of the role of FUL. This has led a number of philosophers to conclude that the Mixed Maxims Objection as a whole is simply based on a misunderstanding [10: 286, 17]. As we will see, however, this is not the case.

(b) Parfit himself is aware that some Kantians have a different take on FUL. He therefore emphasizes that the problem of mixed maxims reoccurs as a matter of fact even if we grant that FUL is supposed to evaluate not an action but rather what Parfit calls an agent’s “doing of it” [15: 290f.].⁹ Hence, the second version of the Mixed Maxims Objection states that, in some cases, FUL wrongly condemns some innocent ‘doings of it’. A closer look at this version will, pace Parfit’s own intentions, illuminate why maxims should be seen as morally relevant after all.

Parfit introduces the idea that Kant distinguishes between the statements “what [an agent] is doing is wrong” and “his doing of it is [wrong]” [15: 290]. The former refers to the action itself, and the latter to how someone is acting in a broader sense, including their motivation and underlying attitudes. According to this distinction, when someone helps a friend out of egoistic motivation, ‘what the agent is doing’ (helping a friend) might not be wrong, but the agent’s ‘doing of it’ (acting egoistically instead of acting from duty) is wrong. At first glance, it seems that the Mixed Maxims Objection misfires if we understand FUL as making statements about the agent’s ‘doing of it’ because Kant could agree that there is a real sense in which helping a friend is not wrong even if the agent is acting on an impermissible maxim (if their ‘doing of it’ is wrong).

However, Parfit believes that this distinction does not solve the problems that mixed maxims pose. According to Parfit, even if we understand FUL as making claims about the agent’s ‘doing of it’, FUL still yields many false verdicts, since impermissible maxims may correspond to clearly permissible ways of the agent’s ‘doing of it’. According to that understanding, the Mixed Maxims Objection could now be stated as:

Kant’s formulas fail, as these formulas condemn some ‘doings of it’ that are clearly permissible or morally required.

Unlike the first version, this second version of the Mixed Maxims Objection is in line with Parfit’s own wording of FUL as it pertains to acting on maxims as opposed to actions (“[i]t is wrong to act on maxims …”). The second version of the Mixed Maxims Objection again resembles a version of the objections based on false positives/negatives, although the false verdicts that the formula supposedly issues pertain to the agent’s ‘doing of it’ (the way they are acting).

Parfit offers several paradigmatic scenarios for maxims that, while impermissible according to FUL, nonetheless correspond to intuitively permissible ‘doings of it’. We will discuss two of these here as, pace Parfit, they illustrate why we should hold on to the notion of maxims in Kant’s sense. Both scenarios refer to situations in which agents do not have moral duties. Parfit assumes [15: 290] that if there is no moral duty, then there is no demand for a specific ‘doing of it’, and, in particular, there is no demand to act from the right motive. The first paradigmatic scenario features an agent who performs prudent actions that do not affect anyone else. If somebody performs a prudent, self-oriented action on an impermissible maxim, then it seems that they do not violate the demand to act from duty because there is no duty pertaining to this situation. Hence, there is no sense in which they are acting wrongly, or so one might think.¹⁰ However, FUL implies that they are in fact acting wrongly.

“When this Egoist takes some medicine, or puts on warmer clothing, he may be acting on his maxim ‘Do whatever would be best for me’. Since this man could not will that this maxim be universal, Kant’s formulas again mistakenly imply that he is acting wrongly. Nor could we claim that, though what he is doing is not wrong, his doing of it is. There is no sense in which, when this man puts on warmer clothing, his acting in this way is wrong.” [15: 291]

The second scenario is a situation in which there is no moral duty, yet certain options would be considered supererogatory. Suppose a child has fallen into a fast-flowing river, and rescuing the child would be too risky to be morally required. Nonetheless, a man jumps into the water and rescues the child, acting on the maxim of maximizing the prospect of reward. Again, according to Parfit, there would be no sense in claiming that the man’s doing of this is wrong:

“On the suggestion we are now considering, if this man saves this child’s life at this great risk to his own life, what he is doing is not wrong, but his doing of it is. That is clearly false. This man is not failing to fulfil any duty, or acting wrongly in any sense.” [15: 291]

However, let us look critically at these two examples in turn. Contrary to what Parfit believes, there is a sense in which, although the Egoist who takes the medicine does not perform an impermissible action, they fail to act in the right way. After all, it is only contingently the case that taking the medicine on the maxim of egoism does not harm anyone. Suppose the situation changes: someone enters the room and informs the Egoist that all of the remaining medicine is needed for a third party who faces imminent death. In that case, the Egoist, following their maxim of unlimited egoism, would nonetheless take the medicine, accepting the risk of death to the third party, even if their own life was not endangered and the medicine would merely cure their cold. Suppose now that there is, in fact, no third party facing death and, consequently, no one tells the Egoist that the medicine is needed elsewhere. Nevertheless, the Egoist’s acting is in this case defective and, in an important sense, morally wrong, because of their disposition to allow others to die if this were a side effect of their action. That is why there is something wrong with their ‘doing of it’ and, hence, FUL does give us the correct verdict.

Similarly, the man standing by the fast-flowing river is disposed not to rescue the child: he would not save the child unless there were a realistic prospect of reward for doing so. As Parfit describes the case, there is no duty to rescue the child (whereas, in the medicine case, there is a duty not to deprive the hypothetical third party of the much-needed medicine). However, we think that the rescuer’s acting is not beyond criticism, because, under different circumstances, the man would have performed actions that were clearly morally wrong. For instance, following his maxim of maximizing the prospect of reward, he would have thrown the child into the river if this would have created a situation that holds out the prospect of reward. Even if the man saved the child, his acting is defective because he acts on a maxim that would lead him to commit immoral actions in other situations.

Barbara Herman, in her critical reply to Parfit in the second volume of On What Matters, draws on a similar example to expose the shortcomings of Parfit’s departure from Kant [6: 87-89]: a gangster feels remorse after a moral conversion not just for their actions, but for the person they were and for the actions they could easily have performed. The gangster was a person who did not respect the rights and moral status of others, and who acted based on wrong attitudes towards their fellow human beings. Herman argues that Parfit’s theory, with its focus on actions as opposed to maxims and character, cannot account for cases where agents regret things that they might (easily) have done to others, given the person that they were at the time. Providing a framework to evaluate dispositions or priorities is a desideratum for ethics as it can capture the significance of counterfactual self-evaluation and feelings of remorse for one’s attitudes and character.¹¹ Parfit wrongly neglects the moral importance of such dispositions.¹²

(c) Notwithstanding our argument so far, the Mixed Maxims Objection can be formulated such that it constitutes an even more pressing challenge to FUL. The problem revealed by the existence of mixed maxims might not be that FUL gives false verdicts about actions, but that it gives no answers concerning the morality of actions at all. We can state the revised version of the Mixed Maxims Objection as:

One of the central tasks of ethics is to answer the question of which actions are morally permissible. If there are mixed maxims, then the evaluation of maxims is of no relevance for this central task, as maxims do not reveal anything about the permissibility of actions.

The idea that one crucial question of ethics is concerned with the moral status of actions constitutes a general challenge to an ethics of maxims. This challenge is not limited to the problem of mixed maxims, as there are other potential gaps between the evaluation of actions and maxims that might leave an ethics of maxims unable to say anything meaningful about the morality of actions. For instance, it could be the case that maxims are so abstract and general that they leave underdetermined which actions can be performed on them. It might therefore be impossible to evaluate every possible action, even if we could avail ourselves of a principle that infallibly evaluates maxims. However, mixed maxims are the most immediate and pressing challenge for an ethics of maxims, since, if there are mixed maxims, it seems that we cannot transition from the (im)permissibility of maxims to the (im)permissibility of actions. After all, any maxim could in principle be a mixed maxim. It then falls to the Kantian to show that the evaluation of maxims can indeed inform our evaluation of the morality of actions. We should note that the problem that FUL might say nothing about the wrongness of actions might be even worse than false negatives. It would mean that FUL missed one of the central points of ethics, and not that it is a deficient or less than ideal answer to the question of how we are to act.

From the existing literature on Kant’s ethics, we can identify three different ways of responding to the real challenge that the Mixed Maxims Objection poses on Kant’s behalf. We will show briefly why the first two are unsatisfying, before focusing on the third.

The first way to deal with the Mixed Maxims Objection is simply to reject the assumption that one of the central tasks of ethics addresses the permissibility of actions, and to maintain that Kant was right to focus on maxims rather than actions. Onora O’Neill sometimes suggests such a view. For example, she holds that Kant was interested ultimately in the good will and thereby in the rightness of maxims, and that judgements about the rightness of actions are only “a second and derivative part of his ethical theory” [13: 87]. Accordingly, Kant is able to talk about the morality of actions, but only insofar as they are actions that someone who has a “morally worthy” [13: 87] maxim would perform. If this were Kant’s position, then we would lack sufficient criteria to make an exhaustive distinction between right and wrong actions. It is only if we know that someone acts on a worthy maxim that we can infer that the action is permitted. Yet, we would not know for every action whether it is right or wrong: in some cases, we do not know what maxim an action was performed on,¹³ and in other cases, though we might know that an action was performed on an impermissible maxim, this is not sufficient to establish whether the action was right or wrong. However, according to this first proposal, only maxims are relevant to determining the quality of the will, and ethics does not have to ask questions about the permissibility of actions.

This is not convincing, either as a reconstruction of Kant or as a way of making sense of Kant’s ethics. Firstly, as Parfit correctly points out [15: 295, 16: 298f.), Kant often does make claims about the permissibility of actions, and not only about the permissibility of maxims (see, for instance, IV:402.16-7). Moreover, we have strong intuitions about some actions that ought to be committed or omitted, such as saving people in our immediate vicinity from grave danger if we could do so without danger to ourselves. Those intuitions are partly independent of the question of what maxim a person is acting on as in some cases our primary concern is that certain actions be performed (such as actions of easy rescue) or omitted (such as murder). The maxim on which this happens is of secondary concern in those cases for which we have strong intuitions about the action. If we can, we should avoid reading Kant as though he were unable to accommodate these intuitions.

The second way is to conceive of maxims as very specific descriptions of actions. As we have already mentioned in section 1, this is how Parfit himself proposes that we respond to the Mixed Maxims Objection, and some Kantians are also sympathetic to this reading of Kant. Iain Morrisson, for instance, claims that it is “relatively uncontroversial” that “a Kantian maxim consists of a statement about performing a type of action, in a certain situation, and for a particular end or object” [9: 74].¹⁴ Proposals like these would do away with the conception of maxims as policies. Instead, a maxim would then take the form “When in situation S, do A!”, whereby S and A are described with all the morally relevant properties that apply. “For Kant’s Formula to succeed, I claimed, this formula must refer, not to maxims in the sense that can refer to policies, but to maxims more narrowly defined as the morally relevant descriptions of someone’s act” [16 (unpublished original English text); see also 15: 311]. The restriction to morally relevant properties, as opposed to all properties, is necessary to avoid maxims that contain names, dates, etc. and that lead to rarity problems, which we will discuss in section 4.

If we accept that maxims are action descriptions, we could restate FUL’s approach to ethical evaluation as follows:

Morally Relevant Description Reading (MRDR):

(1) An action is morally permissible if and only if it is performed on a permissible maxim.

(2) A maxim is a full description of an action that an agent intends in a specific situation, containing all and only morally relevant properties of the respective situation and the action.

(3) A maxim is morally permissible if we can rationally will it to be a universal law.

MRDR is prima facie attractive as it avoids the Mixed Maxims Objection by having each maxim contain morally relevant properties that make it a description of a specific action in a specific situation. Thus, MRDR evaluates people’s specific actions in specific situations.

Of course, how to identify morally relevant properties is a difficult question, and, for those Kantians who emphasize the ‘formal’ character of FUL, this could turn out to be a serious problem, as a purely formal principle alone might not afford a substantive notion of moral relevance. We cannot discuss this issue here in detail, but, since the notion of morally relevant properties will be important for us, a brief remark is necessary. FUL makes assumptions about what we have reasons to will, as the formula asks which maxims we could rationally will to be universal laws. This means that we must ask ourselves whether we could will a world in which certain rules are universally accepted. FUL can only work if we have an account of what gives us reason to will certain events or states of affairs. This implies that we should grant that, even without any of the Categorical Imperative formulas, such as FUL, we have a normative basis upon which to judge things as better or worse (although duties only enter the picture after we have applied the formulas). Once we have an account of our reasons to judge something as rationally willable, we are also able to define morally relevant properties: these are the properties that make a difference to whether we have reasons to will that certain states of affairs obtain. For instance, if a person could obtain a benefit for herself, then the strength of our reasons to rationally will her to obtain this benefit might depend on a number of factors: whether that person is better or worse off, whether others need the benefit more or are more deserving of it, and whether the person has contributed to creating the benefit or not. It does not depend on whether the person is male or female, white or black, etc. The former, not the latter, types of factors are representative of the properties that are morally relevant here.¹⁵

The problem with MRDR is that it neglects an important element of ethics: namely, the principles or policies fixed in one’s character and acted on regularly. MRDR does not inform us about the general policies that we should adopt or abandon because general policies, as per its subclause (2), are not to be fed into the universalization test. According to MRDR, the only kind of disposition that can be assessed morally is a person’s disposition to perform action A when in situation S, whereby both S and A are very specific. The notion of a maxim, as presented by MRDR, leaves no room for determining whether the person would have performed A if the situation had been somewhat different – for instance, if the incentive structure were different (e.g. there was no prospect of a reward). The maxim of our Egoist who rescued the child from drowning would be indistinguishable from the maxim of a moral saint who saves the child out of duty, as both had chosen to perform A in S. In consequence, the evaluation of maxims would no longer allow us to identify permissible actions that are still morally deficient, because under different circumstances agents would have acted in ways that were morally wrong. This would undercut our distinction introduced in section 1, according to which the Egoist’s ‘doing of it’ can be wrong whereas ‘what he is doing’, the action itself, can be permissible.

Yet, we argued that we should hold on to the distinction between the ‘doing of it’ and ‘what an agent is doing’. Parfit’s own examples, as well as Herman’s example of the gangster who regrets their past priorities, reveal that ethics should (among other things) have something to say about character, stable dispositions, and priorities. Thus, part of the agenda of moral philosophy should be to provide a framework for evaluating agents’ policies, something that MRDR fails to do.

The third response to the Mixed Maxims Objection views maxims as policies, just as the first does. However, unlike the first response, the third holds that Kant’s ethics can and should evaluate actions. According to the third response, an action is permitted if and only if it can be performed on at least one permitted maxim. This solution was proposed by Pogge and, in more detail, by Nyholm in direct response to Parfit.¹⁶

Any-Permissible-Maxim Reading (APMR):

APMR allows us to evaluate stable dispositions or maxims understood as general policies, as it leaves open how general or specific maxims might be. At the same time, it overcomes the gap between evaluating maxims as policies and actions, as it determines the permissibility of actions via the permissibility of maxims. Finally, APMR allows us to understand how agents can perform a permissible action on an impermissible maxim: namely, when they perform an action on an impermissible maxim and this action could have been performed on a permissible maxim.

However, APMR is also faced with an important objection: namely, amongst the possible maxims that an action might be performed on, some will run into so-called “rarity problems”.

Parfit objects to APMR on the grounds that for any action, we can find a maxim such that it is the case that the action could be performed according to that maxim and that this same maxim passes the universalization test embedded in APMR subclause (2).¹⁷ If this were true, APMR would fail to prohibit any actions and would therefore not be a viable proposal for understanding the relationship between the evaluation of actions and maxims.

More specifically, Parfit is concerned that, if maxims are understood as policies, they could potentially contain an arbitrarily large number of very specific and morally irrelevant properties. He expresses this concern in the Rarity Objection, arguing that there could be very specific maxims, such as “I should steal the wallets of any women dressed in white I meet, who are eating strawberries while reading the last page of Spinoza’s Ethics!” (see [15: 289]). Although policies are typically quite general, they could also be quite specific and contain references to any number of morally irrelevant properties. Maxims as action descriptions require that they be specified by all and only morally relevant properties that pertain to their application. By contrast, for maxims as policies, there are no set restrictions on the number or kinds of properties to which they make reference.

APMR seemingly allows that we ask whether very specific policies can be willed as universal laws. If an action would be advantageous for me, and the action is described with very specific and morally irrelevant properties, it is likely that I could rationally will that everyone acts on a maxim that allows exactly this type of action. After all, it is unlikely that there will be future situations to which the maxim applies. Thus, the universal acceptance of the maxim would make me better off in the present situation, and I would not have to fear future disadvantages. In Parfit’s example, I could rationally will that everyone acted according to the very specific stealing maxim. If the stealing maxim were permissible, APMR would issue the verdict that it is a permissible action to steal the wallet (even if I stole the wallet on a different maxim, such as the maxim of unconstrained egoism). This reveals a general problem for APMR: namely, that, potentially, it would render all actions permissible. After all, for any action to be permissible, we only need to find one sufficiently specific maxim on which this action could be performed, and which we can will to be a universal law because the maxim is so specific.

There are two ways out of this problem. We will discuss and reject the first one in some detail (a), as it draws on a standard assumption in the literature. We will then develop our own solution (b). In both cases, the idea is that the maxims that we put to the universalization test have to meet further conditions.

(a) A number of authors claim that maxims are by definition very general rules or principles that must not contain specific properties of situations (see, for instance, [19: 171]). Most recently, Nyholm has defended APMR against Parfit on these grounds [12]. Since, according to this view, very specific rules could never be maxims, the plausibility of APMR would not be jeopardized by very specific rules such as the one that Parfit advances as an example of the Rarity Objection.

However, understanding maxims as very general rules or principles does not solve the problem. We have good reason to believe that most very general maxims are not as a matter of fact universally willable. Note that we must assume that these very general maxims, if they are to determine actions in a given situation, contain a “no-matter-what” clause. Take a very general maxim, such as “Keep your promises to friends!”, and assume that this maxim contains not a “no-matter-what” clause, but rather an implicit condition: “as long as there are no opposing moral goods at stake”. In that case, this maxim would not apply, for instance, to a situation in which, my financial resources being limited, I could either keep my promise to lend a friend money to buy luxury goods, or pay my debts to a person who is in financial need and whom I actually owe money to. In this situation, there are opposing moral standards (paying one’s debts). The maxim would therefore not be one that I could put to APMR’s test to evaluate the morality of actions in this situation. Now, if both of the maxims at stake in our example – keeping promises to friends and paying debts – lacked “no-matter-what” clauses and contained implicit “as-long-as” conditions, then they would both be inapplicable to the situation at hand. A very general maxim without a “no-matter-what” clause would therefore not prescribe actions for situations such as the one outlined, but merely make a general declaration such as: paying one’s debts is a good thing. But maxims are not just general statements of what an agent deems good or approves of. They are action-guiding.

For this reason, we should assume that general maxims contain a “no-matter-what” clause. Now, it seems that the universal acceptance of general maxims with such a clause cannot usually be willed rationally.¹⁸ In the future, it could be the case that I would be the person who suffered great disadvantages from the universal acceptance of, say, “Keep your promises to friends!”, and thus I could not will everyone to act on it (the disadvantage of this or other “no-matter-what” maxims could simply be too great). For instance, if there really were no condition on the maxim of keeping a promise to a friend, then this could, in some cases, result in a situation in which someone acting on that maxim does not pay back the money that they owe me, even though this might be extremely detrimental to me. Thus, I cannot rationally will as a universal law a maxim of honouring promises no matter the costs.¹⁹

For most situations, general maxims therefore either do not apply because of implicit conditions or, if they are unconditioned, they cannot be rationally willed as a universal law. It seems that APMR now condemns almost all possible actions, as for most actions there is no permissible maxim on which the action could be performed; maxims are either not applicable or too general to be rationally willable.

That is why, when we apply APMR, we need maxims that are general enough to exclude rare maxims, and specific enough to be rationally willable. The standard Kantian response of insisting that maxims have to be extremely general overlooks the fact that the problem does not merely lie with overly specific maxims. Overly general maxims pose problems, too.

(b) There is, however, a solution to rarity problems that allows us to maintain the idea of maxims as (quite) general policies and dispositions. In contrast to MRDR discussed in section 3, we suggest that not all morally relevant properties need to be included in maxims that are to be put to a universalization test. To address the Rarity Objection, it is sufficient to require that permissible maxims contain only morally relevant properties. Maxims that contain only, but not all, morally relevant properties can avoid rarity problems and serve for our evaluation of an agent’s dispositions, as they are not extremely specific.

Accordingly, we can specify APMR as:

Any-Permissible-Maxim Reading* (APMR*):APMR* is different from MRDR because the latter implies that there is one maxim for each situation. Unlike MRDR, neither APMR* nor APMR restricts the notion of a maxim with which it operates to an action description. APMR* can thus be used to evaluate maxims as policies. When applying APMR*, we can put maxims that do not include all morally relevant properties to the universalization test, and the universalization test reveals whether or not it is permissible to adopt each maxim. It will often be possible to find a maxim that does not contain all morally relevant properties and that can be rationally willed as universal. Those maxims are of special interest since they apply not only to one action in one concrete situation, but to more situations. Such maxims are policies in the sense that we introduced above, and they allow us to evaluate people’s priorities and dispositions.

Most importantly, APMR* has one significant advantage over APMR: namely, it avoids rarity objections. APMR* renders impermissible all maxims that include morally irrelevant properties. It thus imposes limits on how specific a permissible maxim can be, and ensures that morally irrelevant properties do not yield implausible results when put to the universalization test.

To illustrate APMR*, we can revisit Parfit’s own examples. An Egoist who pays their debt performs a permissible action because this action could have been performed on at least one maxim that refers only to morally relevant properties and that we can rationally will to be a universal law. However, there is also a sense in which the Egoist’s acting is wrong: namely, in the sense that the maxim on which the Egoist acts in this specific situation, the maxim of egoism, would have made the Egoist act wrongly if circumstances had been different. Since a maxim of egoism is applicable in many situations, this maxim is of great interest to ethical assessment, as it reveals something about the person’s general priorities and dispositions to act.

In the case of the woman dressed in red and eating strawberries, I will not be able to find any universalizable maxim that allows me to steal her wallet and that refers to morally relevant properties only. In that situation, permissible maxims might allow me to greet her politely or simply to leave her alone, but none of the applicable maxims will render the action of stealing her wallet permissible.

Finally, let us address potential objections to our APMR*.²⁰ Doing so will also serve to clarify our proposal.

First of all, it might be questioned whether APMR* does indeed differ from MRDR. It is more likely that more general maxims fail the FUL test, as the case of “Keep your promises to friends no matter what!” demonstrated. Our discussion above might even suggest that, when applying APMR*, we do in most cases have to include all morally relevant properties in order to find a maxim that can be willed as universal. As stated above, the universal adoption of general maxims can be detrimental for me in some circumstances, and thus I cannot rationally will them to be universal laws. The maxim “Keep your promises!” is too general to be willed universally, and requires restrictions, such as “Keep your promises, as long as you do not violate anyone’s right to physical integrity!”. However, the same point could again be made against that more restricted maxim: even though it is less general than the first, it might not be universally willable because this maxim would allow persons to steal my property in order to keep their promises. Thus, we have to refine that maxim once again, adding the condition that property claims are not to be violated. Likewise, it could be necessary to eliminate the possibility of insults, humiliation, extremely detrimental effects on a third party’s well-being, and so forth. It seems that any universally willable maxim must be very specific in the sense that it describes a concrete situation. This gives rise to the objection that APMR* is superfluous for ethical evaluation, as, whenever we apply APMR*, we de facto apply MRDR.

However, this objection to APMR* overlooks the fact that, as soon as a maxim becomes specific enough to apply to a particular situation, it is possible to add an “as-long-as” condition to the maxim. This addition would make the maxim universally willable without making it a description of only one specific action. As discussed earlier, my maxim could be “Keep your promises, as long as no further goods are at stake!”. Let us stipulate that the maxim is universally willable. This maxim may already be action-guiding in some situations, whilst in many other situations there are further goods at stake and, thus, the maxim does not tell me what to do in these situations. I could refine the maxim further as: “Keep your promises, even if a third person will suffer a small disadvantage, as long as there are no further goods at stake!” This maxim will apply to a much broader range of cases than the first. I will be able to apply that maxim in many situations, although, of course, not in all situations where I have promised something. At the same time, the maxim is universally willable, since I do not have to fear great disadvantages when others act according to the maxim, but I will benefit much from universal promise-keeping. Since this maxim is both sufficiently specific to be applied to many situations, and sufficiently restricted to the point that it can be rationally willed, this maxim is a good candidate for the kind of disposition that moral agents should cultivate. APMR* allows for the evaluation of that maxim and other maxims of this kind, while MRDR does not.²¹

Thirdly, one might also object that the evaluation of maxims embedded in subclause (2) of APMR* is too restrictive, as any maxim that contains morally irrelevant properties would be impermissible, as permissible maxims must be both universalizable and not contain morally irrelevant properties.²² Maxims such as “If someone urgently needs my help, and I am able to help and this would be easy for me, and the moon orbits the earth, then I will help the person!” might sound odd and it is unlikely that anyone will have such a maxim. But why, as follows from APMR* subclause (2), should we judge this maxim as impermissible? The answer is that adopting such a maxim would condition my willingness to help others on morally irrelevant (and very arbitrary) facts about the orbit of the moon. Adopting such a maxim should make us wonder how deep and informed an agent’s commitment to others’ welfare really is. When other morally irrelevant properties occur in our maxims, then similar concerns arise; and, as such, it is reasonable to assume that these maxims are impermissible.

Consider first a situation in which I am holding a gun in my hand that is pointing at you. APMR* allows that we put the maxim “I move my finger if I am in the mood to do so!” to the universalization test, and that the action described in the maxim is permissible if the maxim survives this test. This maxim lacks any reference to the fact that moving my finger would bring about your death, and it seems that, without this particular fact, we do not have reason to believe that this maxim is impermissible. However, it would be grotesque if we were permitted to ignore this decisive fact.

We would like to remind the reader that one reason why a maxim cannot be willed when universalized is precisely because it is too general and too unspecific. Leaving out certain relevant moral properties can therefore be grounds for rejecting a maxim. The maxim “I move my finger if I am in the mood to do so!” is indeed impermissible for that reason. If everybody could act on that maxim, I would live in permanent danger of being killed by someone choosing to move their finger. Therefore, I cannot rationally will the universalization of this maxim.

A similar case can be made for maxims that seem less arbitrary and less unlikely than the moving-finger maxim. Take the apparently innocent maxim “Make my friends happy!”. We might think that this very general maxim passes FUL’s universalization test, making the actions that could be performed under this maxim permissible according to APMR*. However, it would then seem permissible for me to make a friend happy by for example lying to him or maybe even killing someone. However, if the maxim of making friends happy does not include implicit conditions, but is supposed to mean “Make my friends happy, no matter what!”, then universalizing this maxim will lead to a world where others will be prepared to kill me in order to make their friends happy. This is something that I cannot rationally will. The maxim would have to be at least something like: “Make my friends happy, as long as I do not violate anyone’s right to physical integrity!”.

Matters are more complicated when turning to an issue that has a long and chequered history in Kantian ethics: namely, the issue of lying.²³ As we pointed out in section 1, Kant’s own path for justifying perfect duties is no option here, as FUL, in Parfit’s reformulation, is only concerned with what I can rationally will as universal. A maxim such as “Make my friends happy, as long as I do not violate anyone’s right to physical integrity!” seems to be permissible at first sight, and APMR* therefore seems to allow lying to your friend, if lying makes your friend happy: being an act based on a permissible maxim, lying to a friend would then be a permissible action. However, we should question once more whether the maxim really is universalizable. After all, in a world where everyone would act according to the refined maxim of making friends happy, thousands of people would lie to their friends to make them happy, thereby undermining important features of friendship, such as trust and honesty, which make friendship the particular good that it is. It is not clear that we could rationally will a world such as this one.

We cannot settle this issue for good here, given that the question of the moral status of lying has such a rich and complicated history in Kantian ethical theory. Some of the problems that might beset APMR* are in fact problems of the universalization test and raise the question of how best to formulate this procedure. This is an important issue not just for APMR*, but is one that is beyond the scope of our paper.

The strongest version of the Mixed Maxims Objection challenges Kantians to explain how they can proceed from evaluating maxims to evaluating actions. The best approach for this task is to ask whether an action could be performed on at least one permissible maxim. This approach requires limiting the FUL test to maxims that contain only morally relevant properties. However, we should not constrain the test to maxims that contain all morally relevant properties, thereby eliminating maxims as policies as Parfit does. If we followed Parfit, the test could not accommodate the fact that, besides evaluating actions, ethics is also concerned with evaluating dispositions or priorities. Such dispositions or priorities can be causes of shame and remorse, and may warrant moral criticism. Hence, when understood as policies, maxims help us make good on an important desideratum. Since Parfit fails to establish that we necessarily have to abandon this notion, Kantians are well-advised to develop principles that can evaluate both maxims and actions.