Constructivism, realism, and our knowledge of God's will

Before we can understand why constructivist readings of Kant face severe problems in accounting for his religious commitments it is worth reviewing what it even means for a reading of Kant to be constructivist. It is perhaps helpful first to get clear on the opposing position of moral realism both generally and as a reading of Kant. Moral realists hold that there is some sort of independent order of values that human beings become aware of through practical reason. There is of course tremendous variety in what realists take these independent values (or value) to be, but according to the realist reading of Kant the independently existing value is the value of rational nature itself. According to such realists rational nature in all its possible forms is valuable in and of itself independently of what human beings or anyone else might think about it.

Constructivists deny that there is any such independent order of values, and hold that whatever values there might be depend in crucial ways on us and our practical reason, or as Rawls says in explaining what he takes to be Kant's view, ‘[P]ractical reason constructs for the will its own object out of itself and does not rely on a prior and antecedent order of values’ (Rawls (Reference Rawls and Herman2000), 230). While it has the virtue of being concise, this characterization of constructivism is a bit gnomic as well, and it is helpful to turn to a fuller characterization of constructivism's attitude towards the nature of value offered by Sharon Street:

As with realism there are widely varying ways of working out this basic claim about the nature of values, and constructivists differ on what attitudes of valuing creatures and facts about their viewpoint are important, how these attitudes and facts about their viewpoints generate normative demands, and what level of objectivity those normative demands have. Kantian constructivists are quite ambitious on this last point; they hope to highlight certain basic commitments that all human beings share in virtue of facts about the nature of human agency itself and show that these attitudes commit us to certain moral standards. If the Kantian constructivist is right about the nature of our agency and the commitments it entails and the moral standards those commitments themselves entail, then she can generate moral standards that hold for all human beings.

How exactly can the Kantian constructivist use these commitments to generate such moral standards though? The usual way that they do so is through some sort of procedure that they claim models these commitments as well as some other shared facts about the human condition. The moral principles that human beings are bound by are whatever principles these procedures generate. John Rawls presents one way of doing this on Kant's behalf in his discussion of the categorical imperative procedure (or CI procedure in Rawls's abbreviation) in his Lectures on the History of Moral Philosophy. Rawls quite plausibly claims that Kant's own statements of the CI procedure implicitly assume a good bit of background information and attempts to present a more fleshed-out version of the CI procedure that makes these assumptions explicit:

After some further discussion Rawls sums up what he takes to be Kant's CI procedure in the following terms:

Though they disagree on the details a number of Kantian constructivists try to generate moral demands from such idealized procedures on Kant's behalf. (One of the most ambitious, and perhaps after Rawls most influential, attempts to do this is Christine Korsgaard's work, in particular her Sources of Normativity).Footnote ² Not all constructivists make recourse to such idealized procedures, however. Onora O'Neill rejects the reliance on idealized procedures and conceptions of agency that she believes are characteristic of Rawls, Korsgaard, and many other Kantian constructivists and invokes a much more minimal account of practical rationality. Rather than rely on such idealized procedures and attempting to work out moral principles by answering hypothetical questions about such models, O'Neill thinks that constructivists need to work out moral demands with reference to the actual answers that real procedures generate. O'Neill's version of constructivism then relies on the actual political processes in ways that Rawls and Korsgaard's do not.Footnote ³

One may very well be sceptical whether constructivists can find any such commitments that human beings share simply in virtue of our practical viewpoint or any other facts about our agency or the human condition. However, let us suppose for the time being that they can in fact find some commitments all human beings share in virtue of the nature of our agency or some other facts about us and present some procedure, whether real or hypothetical, that accurately models such commitments.

Even if they can do so, the Kantian constructivist's principles still fall short of the level of the objectivity that Kant holds that moral principles have. Kant thinks that moral principles hold not just for us or beings like us, but for all possible rational beings. In fact he goes so far as to repeatedly claim that God follows the same principles that human beings are aware of.Footnote ⁴ For instance, in the second Critique Kant claims that, ‘[The principle of morality] is therefore not limited to human beings only but applies to all finite beings that have reason and will and even includes the infinite being as supreme intelligence.’Footnote ⁵ And in the Religion Kant reiterates this claim. Further, Kant's dismissal of any putative revelation that is at odds with morality also depends on the claim that God adheres to the same moral principles that our reason reveals to us.Footnote ⁶

Now the moral realist interpretation of Kant faces no challenge in explaining why God would follow the same moral principles that we find to be valid for us. Rational nature has value independently of the attitudes that one has or the point of view that one adopts, and both God and human beings are aware of this value through reason. However, the claim that God and human beings are cognizant of the same moral law is extremely difficult for a constructivist reading of Kant to even make sense of much less justify. Consider the proceduralist reading of Kantian constructivism that we find in Rawls. Rawls thinks that we should set our procedure up so that it models the most important features of human agency and our moral relations with others. So when Rawls explains what he takes to be Kant's CI procedure in the Lectures on the History of Moral Philosophy he assumes a number of features about human being, such as that we live in relations of social dependence with other human beings and that we have limited information. Neither of these is true for God, so there is no reason to think that any moral principles that might result from the CI procedure as Rawls sketches it would hold for Him.

Rawls actually admits as much, and notes that Kant thinks that the moral law applies to all possible rational beings:

Rawls has presented an accurate statement of Kant's own position here, but the problem is that this position does not fit within the constructivist framework. If moral principles are constructed out of a procedure that models facts about our own reason then how can we really put any distance between the moral law and the way that law applies to us? On Rawls's reconstruction of Kant the moral law is an entirely empty ideal until we perform some sort of constructivist procedure, and the content that the moral law has depends on the exact nature of that procedure. If that procedure is supposed to model key features of the human will and the human social condition, then a different sort of procedure would apply to God (or any rational being radically different from human beings), and a different procedure would yield different principles. Proceduralist interpretations of Kant like Rawls's lack any obvious way to give real content to the statement that the same moral law holds for God and human beings. If the statement means that roughly the same moral principles hold for God and human beings, then constructivism cannot justify it. If it does not mean this, then it is it is hard not to conclude that it is empty verbiage.

The problem is bad enough for constructivist interpretations of Kant like Rawls's that involve some level of idealization, but it is even worse for someone like O'Neill who thinks that we must look at actual human beings and our societies and political processes to derive moral principles through Kantian constructivism. If idealization is allowed in the constructivist procedure we might hope that we could draw out some very general features of the will that might hold for all rational beings as such – we probably shouldn't hold out too much hope for this, but it is at least not ruled out in principle – and create a procedure that models those features. However, if the content of moral principles is dependent on the features of actual human wills and the contingent features and histories of the societies those wills find themselves in, then there is even less reason to think that they would hold for any sort of being that differed substantially from human beings. O'Neill's insistence on human finitude and our lack of transcendent insight into any sort of moral principles further widens the gap between the human and divine wills. Throughout her work on Kant O'Neill repeatedly emphasizes that we must make do with what she sees as Kant's modest conception of practical reason because we are finite, material beings, who lack the sort of transcendent insight into morality that philosophers like Plato claimed.Footnote ⁷ However, if God exists He would be an infinite, immaterial being, who possesses exactly this sort of insight. So why think that He would be forced to accept the modest conception of reason that we must make do with?

I realize that at this point these worries are rather abstract, and there are several questions I need to answer if my argument against the constructivists is to be compelling: in what ways exactly might the principles that apply to human wills differ from those that would apply to a divine will on the constructivist interpretation? And even if the worry turns out to be legitimate, just how serious would it be if Kant's claim that the same moral law applies to divine and human wills turns out to be unsupported? Does this claim do any real work in Kant's moral philosophy? I now turn to showing both why the claim that God follows the same moral principles we are aware of is important, and a concrete way that the constructivist procedure would yield different principles for God and human beings.

The most important reason that we cannot lightly dismiss Kant's claim that God follows the same moral principles as human beings is that he insists that morality sets ends for us that are possible only with divine assistance. So if we are truly to pursue those ends we must have faith that God will provide such assistance. Kant makes a number of different claims about why we need divine assistance, and his views on this issue seem to evolve over time. However, for our purposes the most important such claim is the doctrine of the highest good that Kant first presents in detail in the second Critique.Footnote ⁸ According to this doctrine we must at least be able to hope that it is possible to bring about a condition where rational beings are happy in proportion with their worthiness to be happy. There are a number of open interpretative questions on the issue of the highest good, and I think it quite likely that Kant's views on this issue changed somewhat over time. But Kant's commitment to the highest good as an end that morality necessarily sets for us remains constant after the second Critique.Footnote ⁹ And in all his discussions of the highest good he insists that it is only possible if God exists. According to Kant the only way that happiness and worthiness to be happy could match up perfectly in this way is if God brought it about or set up the order of the world in such a way that it would come about.

For this to be so though requires that God concerns himself with human happiness and worthiness to be happy. God must have some moral reasons to assist human beings in achieving the highest good. On the moral realist reading of Kant it is not hard to sketch why this is so. Human beings are rational beings and rational beings are valuable. What is more, the morally acceptable ends of rational beings have value in virtue of the fact that they are the ends of rational beings, and so any rational being has reason to further the morally acceptable ends of other rational beings. Since the highest good is not just a morally acceptable end, but one Kant thinks that moral law sets for human beings, God or any other rational being that could help human beings in achieving this end would have especially important reasons to do so.

However, on a constructivist reading there is no obvious reason why God would have any moral reason whatsoever to assist human beings. To see why this is so consider how Kantian constructivists argue that human beings have a duty of mutual assistance. Because we are finite beings who may well find ourselves needing help from others, we could not will a maxim never to help others in need as a universal law. Therefore, we owe others at least some assistance. One should already see reasons to doubt whether this reasoning could possibly apply to God. After all, God is an infinite being and presumably could not find Himself in a situation where He needs help in accomplishing His ends. Further, as Rawls notes on this point the argument for duties of assistance among human beings depends on the fact that real human agents have limited information. If I knew that due to my own temperament and extremely fortunate life circumstances I would never in fact require aid from others, then I could in fact universalize a maxim not to assist others. But unlike us God is omniscient and there are no limitations on His knowledge. So if our having duties to assist others necessarily depends on the facts that we are finite vulnerable creatures who lack full information, then one can only conclude that an omniscient and omnipotent being would have no such duties. In fact, O'Neill explicitly says as much, and concedes that even beings who were merely much more intelligent, knowledgeable, and powerful than us would lack any such duty of assistance:

So a constructivist cannot make sense of Kant's claims about the highest good. The very idea depends on the claim that we have reason to believe that God will assist us if He exists, but the constructivist can give no account of why God would have any reason to do so.

Can we dispense with Kant's religious commitments?

How might a constructivist handle these problems? Rawls for one admits that one cannot connect the constructivist reading of Kant to the claims about the highest good that he makes in the second Critique, but he simply summarizes what he takes to be Kant's doctrine of the highest good rather than trying to explain why Kant would adopt a doctrine at odds with what Rawls takes to be his method of moral philosophy.Footnote ¹⁰ Many other constructivists devote little or no attention to these aspects of Kant's philosophy, which is remarkable given the problem they pose for their readings of Kant.

O'Neill represents a notable exception here.Footnote ¹¹ In her ‘Kant on reason and religion’ she attempts to deal with the difficulty that Kant's religious commitments pose for the constructivist reading of Kant. However, on closer examination it turns out that O'Neill's attempt to deal with the religious elements of Kant's thought fails on the terms O'Neill herself sets for success.

Now it is practically impossible to imagine how the highest good could be brought about without divine assistance. For one, thinking that we could even roughly approximate such a state in this world would require that we think that the natural and social order of this world is extremely amenable to the ends that morality sets for us. Kant argues that this requires that we believe in a God who has coordinated natural laws with the law of morality. However, O'Neill argues that nothing in Kant's philosophy requires that we believe in the highest good. Instead, all that Kant's moral philosophy requires is that we have some grounds for believing that moral agency can be effective, and these hopes need not take the ambitious form they do in the highest good.Footnote ¹² She argues that more modest forms of hope such as a hope in political progress can also suffice, and these more modest forms of hope can have purely secular sources. Further, she believes that we can in fact find two rival accounts of hope in Kant's own work. While Kant sometimes argues for the highest good, in other works he argues for only a more modest belief in the possibility of some moral progress. O'Neill implies that Kant himself believed in only this weaker form of hope, but presented the distinctively religious form of hope because he wanted to appeal to traditional religious believers like his servant Lampe. On O'Neill's reading, though Kant's preferred form of hope is purely secular, he wants to allow space for distinctively religious hopes for those who require them. As readers we are free to choose either the stronger religious form of hope (if we cannot do without such hopes) or the weaker secular forms of hope.

There are a number of problems with O'Neill's argument both as a reading of Kant and as an independent philosophical argument. However, I want to focus on only two of them. First, her argument that Kant presents two rival interpretations of reasoned hope with the secular version being his preferred one simply does not fit his actual works. Second, while O'Neill argues that Kant at least wants to make a space for religiously based hopes, her own constructivist reading of Kant cannot do that. Instead, the only hopes that could possibly have any support on the framework she erects are purely secular ones.

The first claim is the easiest to show. O'Neill argues that while we find religious hopes in some of Kant's work, there is also a purely secular form of hope in other works that is in fact Kant's preferred alternative. However, O'Neill quotes only one work in support of this, the essay ‘On the common saying: that may be true in theory but not in practice’, but this is a relatively brief, popular essay rather than one of Kant's developed philosophical works. Even if O'Neill were right that Kant occasionally flirted with a rival secular form of hope, we should give weight to longer, more developed, philosophical works such as the second and third Critiques and the Religion when deciding what Kant's considered views were, and in those works he clearly opts for religious hopes. In fact, in the third Critique Kant expressly denies that the hope that morality requires is possible without belief in God:

If the two accounts of hope are in fact rivals then the religious account has a much better claim to represent Kant's own considered views. But the deeper issue with O'Neill's reading as an interpretation of Kant is that there is no reason to read the secular and religious accounts of hope as rivals and quite a bit of reason not to. If we really do find two accounts of hope in Kant, then the most charitable interpretation would be to find a way to reconcile them rather than reading them as rivals, and it is not too hard to read Kant in a way that does reconcile the secular and religious accounts of hope. For instance, John Hare argues that we should read Kant as holding that we can make some progress towards the highest good in this world and this life and should aim to approximate it as closely as we possibly can, but that we can only really achieve the highest good in an afterlife.Footnote ¹³ On Hare's reading we could very well find a secular, political account of the hope morality requires in Kant that is itself part of the highest good, which does require religious support. Accounting for Kant's own views, then, requires that we have some grounds for believing that God would assist us if He exists.

One may of course wonder how bad it is for the constructivist if she cannot in fact account for Kant's highest good. After all, many philosophers find Kant's arguments for the highest good entirely unconvincing. Personally, I believe that we can reconstruct a better case for the highest good in Kant's work than many interpreters admit, but I need not argue that point here. This is because even if we dismiss the highest good entirely, O'Neill's reconstruction of Kant fails to do what she promises. While O'Neill thinks that we find a purely secular account of hope in Kant, she also holds that he wants to make a space for religious forms of hope. However, the problem is that on the constructivist reading of Kant there can be no grounds whatsoever for any form of religious hope.

Religious hopes require that we have some grounds for believing that God would assist us in achieving the ends that morality sets for us whatever we might take those ends to be. But what possible grounds are there for such a belief on the constructivist reading of Kant? As we have seen, the constructivist reading of Kant cannot itself show why God has any moral reason to assist human beings. What other sources could religious hopes have? One obvious candidate is of course revelation. But despite the role that revelation plays for many religious believers, it is clear that Kant thinks that revelation in and of itself carries little weight. Consider what Kant says about direct revelation in the Conflict of the Faculties: ‘For if God should really speak to a human being, the latter could still never know that it was God speaking. It is quite impossible for a human being to apprehend the infinite by his senses, distinguish it from sensible beings, and be acquainted with it as such’ (Conflict of the Faculties, 7:63; Kant (Reference Kant, Wood and di Giovanni1996b), 283). And while Kant takes pains not to rule out the possibility of revelation in the Religion, he is also clear that on its own revelation gives one hardly any positive grounds for belief. Instead, religious beliefs must have the support of practical reason. The only convincing positive grounds that Kant believes we can find for any revelation are moral ones. But again thinking that our own moral principles give us any way of evaluating putative revelation supposes that God would follow the same moral principles that we do if He exists.

One could answer that the mere possibility that God would provide assistance to human beings in achieving the ends that morality sets for us is sufficient for religiously grounded hopes. Granted, the constructivist reading does not rule out the possibility that if God were to exist he could have some inscrutable moral reason to assist us, but independently of Kant's philosophy this is insufficient for any form of hope worthy of the name. Let us consider a mundane situation here. It is at least possible that one of my neighbours has some mysterious reason, say, to pay off my mortgage in five years’ time. Even if I couldn't rule this out (and I can't) that would hardly suffice for me to hope that I might be able to pay off my mortgage in five years. Contrast this with a situation where a candidate for governor has promised all state employees a raise, and the extra money from the amount she proposes would be enough that I could pay off my mortgage in five years. Even though I could hardly be certain that I would get the money (she might lose the election, she might abandon her promise, the legislature may block the proposal, etc.) it makes sense to speak of hope here since I have some positive reason to believe I might get the money I need to pay off the mortgage in five years’ time.

Kant himself would not think that bare possibility is enough to underwrite hope here. The hope that O'Neill has in mind requires faith, but for Kant faith is not mere belief in something that is logically possible. Instead, Kant is clear that faith (Glauben) is a sort of belief that is supported by some positive reasons, albeit reasons which fall short of supporting full-blown knowledge (Wissen).Footnote ¹⁴ But again what positive reasons can we possibly have for religious hopes within the Kantian framework that constructivists like O'Neill present? According to Kant revelation cannot provide sufficient grounds, and on O'Neill's account practical reason can tell us only about the moral principles that hold for beings like us, but not about the moral principles that God would follow. If the constructivist reading of Kant is right, then there is no possible source for religiously based hopes within the sphere of morality. Only secular forms of hope could possibly find any sort of epistemic support.

O'Neill wants to give a reconstruction of Kant's views where either religious or secular versions of faith and hope are live possibilities. However, the constructivist reading of Kant gives us grounds only for secular hope. Even if we find his arguments for the highest good lacking, this ought to trouble us. Even O'Neill thinks that Kant wants to allow some space for religious hopes, and it would be mind boggling that Kant would write so much about religion if he did not at least allow that religion can play some role in the moral life.

This should also give us pause when thinking about the merits of Kantian constructivism as a moral theory. In other pieces O'Neill has argued that one of the main merits of Kantian constructivism is that it provides a firm foundation for a robust form of toleration, and she insists that toleration requires more than simply putting up with various forms of belief. Instead, toleration in its fullest form requires that we respect those beliefs.Footnote ¹⁵ Kantian constructivism cannot do this when it comes to religiously supported ethical hopes, though. At best it must hold that such hopes are entirely lacking in support. This could allow for toleration in a weak sense, but it does not allow for any sort of robust respect for religious beliefs. This is a problem for constructivism more generally and not just for O'Neill. Constructivists want their moral theories to be religiously neutral, and there are good reasons for this desire. Any moral theory that rules out any positive space for religion will greatly limit its possible reach, and simply as a logical matter a theory that needs to assume less is in better shape than a rival that must assume more.

Unlike constructivism, moral realist forms of Kantianism can be religiously neutral. According to the realist, if God does exist He would have reason to assist us in bringing about our morally legitimate ends, so there are some grounds for religiously based hopes. However, a Kantian realist would also hold that moral requirements exist independently of God. So if we were, say, unconvinced by Kant's moral arguments for God's existence, there is no reason that one could not develop purely secular forms of Kantian ethics within the realist framework.Footnote ¹⁶

Formosa's constructivism and reasons for divine assistance: a test case

Now one might charge that my argument is not a knock-out blow against constructivist readings of Kant's ethics. I would admit this – decisive proof is rare in philosophical debates after all – but it does set a challenge for constructivist readings of Kant: they need to be able to give some account of why God would have positive moral reasons to assist beings like us. If they cannot do this, they can neither account for Kant's own religious views nor even allow space for distinctively religious sources of hope within the supposedly Kantian framework they erect. What are the prospects for a constructivist account meeting this challenge? Might some more recent forms of constructivism more successfully meet this challenge than do O'Neill, Korsgaard, or Rawls? In this section I will show that more recent forms of constructivism have not met this challenge, and that the prospects of doing so look quite bleak.Footnote ¹⁷

How might a constructivist explain why God has reasons to assist us? The only way that I can imagine that a constructivist account could show that God and human beings follow the same reasons is by focusing on the mere idea of a rational will as such. If one can successfully derive concrete moral principles from features that any rational will must have, then such principles would apply to any rational will including God. However, the worry is that such an account would be so abstract that there is little reason to think that one could use it to generate any duties whatsoever, much less positive duties of assistance. To see how this plays out let us examine the Kantian constructivism that Paul Formosa formulates in some recent works. Formosa tries to work out moral principles from the mere idea of a rational will, but as we will see, there is some reason to doubt he can generate any moral principles, and by his own tacit admission it cannot generate any principles that dictate moral assistance. To do that not only must Formosa bring in facts about human finitude as O'Neill or Rawls do, but like O'Neill he explicitly admits that beings who did not share our vulnerabilities would have no reasons to provide assistance to other rational beings. (I focus on Formosa since, unlike other recent versions of Kantian constructivism such as those of Timmermann or Sensen, he explicitly tries to account for Kant's claim that God follows the same moral rules that we do. If I can show that even his version of constructivism cannot show we have any reason to think that God would assist us, this shows that the problem is still outstanding for constructivist interpretations of Kant.)

Formosa contrasts his own version of constructivism with Rawls's and Korsgaard's, and tries to develop a form of Kantian constructivism that does not depend on contingent facts about our own commitments as he believes theirs do. He starts with an extremely minimal idea of practical rationality and practical reasons and tries to derive the categorical imperative from these. He begins by drawing on Kant's discussion of rational principles in the Groundwork, ‘A rational law or principle is one that “hold[s] for every rational being as such,” and so must be based on grounds that are valid for every rational being as such’ (Formosa (Reference Formosa2013), 181). He interprets that in the following way:

He then continues that by the very definition of rational beings as beings that freely and intentionally act for reasons, a rational being could not will to be used as a mere means. He holds that this gives all rational beings a veto over the principles that could be adopted by other rational beings, and so any principle that all rational beings could not share must be rejected. So only those principles that all rational beings could agree to count as rational. If successful then, Formosa's version of constructivism would generate concrete moral principles from the mere idea of rationality, and would not require that we assume contingent features of human beings to do so. Formosa's primary worry is that previous formulations of constructivism like Rawls's open the door to relativism, but he claims that if successful his form of constructivism would account for Kant's claim that God follows the same moral principles that we do (ibid., 189–190).

Does it succeed? I am convinced it does not. For one thing, he does not say nearly enough about how we answer the question of what principles a rational agent could not share or regard as valid. He is clear that this question is not to be interpreted in terms of what an agent actually does choose, but that leaves the question of what his positive account is wide open. In speaking of reasons rational beings could agree to or share he cannot mean reasons that they would agree to if they were presented to them, as this would make the matter completely subjective, and it would rule out a number of perfectly acceptable maxims. For instance, many narcissists would in fact reject any principle that allowed others to criticize them in any form in any circumstances. So these narcissists could not share or regard as valid any principle that allowed them to be criticized, but that hardly means that it is immoral to level any form of criticism whatsoever at a narcissist. On the other hand if all that could share means is that it is merely conceivable that a possible rational being would choose it, then Formosa's formulation rules out nothing. For instance, Formosa says, ‘your incentive to use me as a slave to enrich yourself is not a reason I could share as a rational agent since it involves using me as a mere means’ (ibid., 182). However, it is easy enough to imagine instances where a rational being would consent to being used as a slave. The sad truth is that there are a good many actual people who idolize say some movie star, businessman, pop star, or athlete, who would be overjoyed at the prospect of serving as their idol's slave.

How then do we cash out ‘could’ in such a way that Formosa's procedure does not yield bizarre results but still yields results at all? Admittedly he does not seem to mean by ‘could’ what actual rational beings would consent to or what it would be possible for some imaginable rational being to consent to. Instead of what reasons rational beings could or would share or regard as valid he sometimes speaks of what we could consent to as rational agents. The question then is: what does it mean to choose as a rational agent? It is hard to give any account of this that is not viciously circular. Does speaking of ‘what one could will as a rational agent’ mean ‘what one would will if one were fully rational’? But this leaves us an account of rational principles that says nothing more than that rational principles are the principles we would choose if we were fully rational. That tells us nothing about what concrete principles could actually count as reasons, though.

Formosa may give us an account of reason that as a purely verbal matter holds for all rational beings as such, but as far as I can tell it does not yield any determinate content. The reason Rawls, Korsgaard, O'Neill, and other constructivists have brought actual facts about human beings, assumptions drawn from a particular human culture, or, in O'Neill's case, the results of actual political processes, into their procedures is that doing so allows them to yield determinate content. As O'Neill and Rawls both note, without these assumptions Kantian constructivism is, at best, underdetermined in its results and at worst completely without content. There is a price for doing this, which is that the results of any such procedure are limited in the range of agents from whom they hold true, but the alternative is to get no results at all.

Now it is entirely possible that my own realist sympathies have led me to read Formosa's account through an uncharitable lens, and I must confess that interpretative moves that seem obvious to many constructivists hardly seem so to me. However, even if Formosa's account were more successful than it seems to me to be, it would not adequately address the problem I have raised in this article. If successful, Formosa's account would show that God has moral reasons not to do certain things to us; He would not use us as mere means or order another human being to do so. This would make sense of some of Kant's claims about God, such as that God would never order us to kill an innocent person.

What it does not account for is the claim that God would have reason to assist us. In ‘Is Kant a moral constructivist or moral realist?’ Formosa touches on assistance only very briefly. However, he addresses the issue of positive duties of assistance in much greater detail in a later essay, ‘The rule of vulnerability in Kantian ethics’, and what he says there is extremely revealing. In this piece Formosa derives human duties of assistance entirely from the facts of human vulnerability and dependence. In fact, he claims that we have a duty to promote the happiness of others and help them achieve their ends only because:

Now one might answer that just because Formosa invokes facts about human vulnerability and dependency in deriving duties of assistance here this does not mean that he must do so to derive duties of assistance. Perhaps there are multiple ways of doing so open to him, and he only chooses the one here that most easily allows him to do so. The problem with this reply is that later in this same passage Formosa expressly denies that beings who were not vulnerable and dependent would have any moral reasons to provide assistance:

While Formosa is quoting Barbara Herman in much of this passage, he clearly endorses what she says and goes so far as to provide his own gloss on why ‘angels’ would have no duties of mutual aid. If such angels would have no duties of assistance towards us, God, who is completely invulnerable and independent, would not either.

If the constructivism Formosa presents in ‘Is Kant a constructivist or moral realist?’ has any way to derive duties of assistance that does not invoke facts about human limitations, then why would he tacitly deny that any such derivation is possible here? If his version of constructivism can make sense of what Kant says about the divine will, why would he, contra Kant, deny that a being like God would have any reason to assist human beings? Even if Formosa can derive some moral reasons from the mere idea of the rational will, by his own tacit admission the constructivist cannot derive reasons for assistance from the mere idea of a rational will. So even if successful, Formosa's constructivism could only make sense of one half of what Kant says about God following the same moral reasons that we do. It could show that God has moral reasons to refrain from doing certain things to us, but it would not show that He has any reason to do anything for us. In fact, if we take Formosa at his word in this later essay, it would seem that his constructivism shows that God and similar beings would not have moral reasons to assist us.

Now there are other recent constructivist readings of Kant such as Sensen's and Timmermann's that I have not considered in detail. However, neither Sensen nor Timmermann even try to address Kant's claims that the same moral principles hold for God and human beings, and I see no obvious way that their constructivisms could do so.Footnote ¹⁸ I am sceptical that Sensen, Timmermann, or any other Kantian constructivist could do this. This is due to a deeper problem with constructivist attempts to explain God's assistance that Formosa's attempt to formulate a version of Kantian constructivism that holds for all beings makes manifest.

The more that a Kantian constructivist abstracts away from particular features of human agency, the wider will be the scope of beings for which that account holds. So if an account can abstract away from enough features of human agency, then such an account may well hold for any sort of will including God's. However, the more features of human agency that a form of Kantian constructivism abstracts away from, the more difficult it becomes for it to generate any concrete moral principles and the fewer such principles it can generate. This is why Rawls and O'Neill work distinct features of human agency into their versions of Kantian constructivism; doing so allows them to generate sufficient moral principles for a complete and satisfying moral theory. It is revealing that even Formosa eventually takes the same route; when it comes to positive duties of aid he assumes such features of human agency rather than the more abstract account of practical reason and willing he works out in ‘Is Kant a moral constructivist or a moral realist?’

The price for assuming features of human agency in one's account of practical reason, though, is that its scope becomes narrower. Such a theory may not even apply to all possible finite rational beings, much less to an infinite rational being. Given the huge gulf between human and divine agency, only a very abstract and general account of practical reason and agency of the sort Formosa develops in ‘Is Kant a moral constructivist or a moral realist?’ could possibly hold for both sorts of will, but such an abstract and general account of practical reason and agency cannot explain positive duties of assistance and so cannot explain why God would have reason to assist human beings. On the other hand, an account that assumes concrete features of human agency may be able to explain duties of assistance, but such an account would not apply for the divine will. Kantian constructivists seem stuck with choosing one option or the other, and neither of them can explain why God has reasons to assist us. Realist readings of Kant simply avoid this dilemma.