1 Introduction

In his book, Normative Reasons. Between Reasoning and Explanation, Artūrs Logins presents an ingenious new view concerning normative reasons, the so-called Eroteric View. This view is reductive – by contrast to reasons-first views – since it provides an account of the nature of normative reasons that goes beyond the platitude that normative reasons to φ are facts that favor φ-ing. On the Eroteric View, all normative reasons are (parts of) answers to normative ‘why’ questions. This is their unifying feature. Normative ‘why’ questions are questions that ask, ‘Why φ?’ or ‘Why ought S to φ?’, with respect to a certain subject S for some response φ, which might be a belief, an action, or an affective attitude. Still, Logins insists that “there is no one single sort of thing that corresponds to normative reasons”, for there are two types of normative reasons that are “fundamentally distinct” (Logins, 2022, p. 165). The two types of reasons are due to two distinct types of ‘why’ questions: The first type of question assumes the truth of the explanandum – the fact that S ought to φ – and requests an explanation that will help the inquirer better understand why the explanandum is the case. An appropriate answer is any fact that explains or that makes it the case that S ought to φ. The corresponding reasons are normative explanatory reasons (NERs) to φ, according to Logins. The second type of question challenges the truth of the explanandum and demands an argument that will support the explanandum as its conclusion. Regarding this second type of normative questions, they ask for an argument that will establish that S ought to φ. The premises in such an argument will be normative reasoning reasons (NRRs) to φ.

Let me briefly illustrate the distinction with the help of one of Logins’s own examples. Artūrs is serving his children soup for dinner. The normative reasons for the children to eat the soup include, on the one hand, normative explanatory reasons, or NERs. These are answers to the question ‘Why ought the children to eat the soup?’ that take it as given that they ought to eat the soup, and that provide us with facts that make it so. Logins gives as an example the fact that Artūrs decided to buy soup for dinner, or that there is soup in the house. Both of these partially explain why it’s the case that the children ought to eat the soup, and so are normative reasons to eat it. Logins denies that the fact that soup is healthy is a reason in this sense, since that doesn’t explain why they should eat the soup rather than, say, eggplants or carrots, which are also healthy. The children’s normative reasons further include normative reasoning reasons, or NRRs; these are answers to the ‘why’ question that challenges the claim that the children ought to eat the soup and asks for an argument for this claim, and so they are premises in a good argument supporting the conclusion that the children ought to eat the soup. NRRs exclude the fact that Artūrs bought the soup, since this fact has no place in a good argument supporting the conclusion; by contrast, that it is healthy to eat soup is a premise in a good argument supporting the conclusion that the children ought to eat the soup, and so it is a normative reason in this sense (Logins, 2022, p. 173).

So much for a brief exposition of the Eroteric View. I now first turn to one friendly comment on the exact form of the normative question which is central to the view, a comment that I hope might help to strengthen the Eroteric View. Then I will focus on my more critical comment that calls into question how NERs and NRRs relate.

2 The Shape of the Normative Question

My friendly comment concerns whether it is correct for proponents of the Eroteric View to confine themselves to the question of why S ought to φ as the relevant normative question, which delineates normative reasons of the two kinds. In my view, another plausible candidate is the question, ‘Why is S permitted to φ?’ or similarly ‘Why is it permissible for S to φ?’ It seems as intuitive that reasons why S is permitted to φ are reasons to φ, as it is that reasons why S ought to φ are reasons for her to φ. For instance, in the soup case, the fact that soup is healthy is part of an explanation why it is permissible for the children to eat soup, and at the same time intuitively and contrary to Logins’s claim, a NER for the children to eat soup. Think about it: Why is it permissible for the children to eat soup? – In part, because soup is healthy. It may help to contrast this with cases where the food Artūrs bought is very unhealthy. In such cases, the fact that he bought the food does not make it permissible for the children to eat it. This indicates that the fact that Artūrs bought the soup alone does not suffice to fully explain why eating it is permitted, and that the fact that the soup is healthy contributes to the explanation of why the children are permitted to eat it. It is also an NER to eat the soup.

I recommend a broader understanding of the relevant normative question, including as NERs answers to both the ‘ought’ question and the permission question. This understanding is preferable because it allows adherents of the Eroteric View to count more facts as NERs. Moreover, it helps the Eroteric View to identify normative reasons in cases where there is nothing the agent ought to do. Consider cases in which an agent has several options which are equally supported by reasons. For instance, Homer has normative reasons to stay home that are just as good as his reasons to go out. Plausibly, both options are permissible for him, but it is true of neither of them that he ought to perform it. The understanding-seeking questions ‘Why ought Homer to stay home?’, ‘Why ought he to go out?’ don’t seem to get a grip, since they are asking for something that is not the case. By contrast, the questions ‘Why is Homer permitted to stay home?’, ‘Why is he permitted to go out?’ are perfectly suitable to the situation. My suggestion also alleviates the worry that the Eroteric View cannot account for the existence of reasons to kill civilians with gunfire rather than with a nasty torture device (Logins, 2022, p. 187). It is much easier to swallow that it is permissible to kill an innocent person with gunfire if the only alternative is killing them by torturing them, than it is that the agent ought to kill them in this way, given that the only alternative is much worse.Footnote 1

I have focused on the understanding-seeking normative question here, but similar points hold for its argument-seeking counterpart. Sometimes a questioner may challenge even the assumption that an action is permitted, and plausibly, premises in a corresponding argument to the conclusion that φ-ing is permitted are reasons to φ. This description fits the bill for Homer’s case, for instance.

3 Normative Explanatory Reasons versus Normative Reasoning Reasons

I now want to focus on the question of how unified or distinct NRRs and NERs are, according to the Eroteric View. How do we read Logins’s claims that – though their being answers to the same normative question unifies them – these two kinds of reasons are not one kind of thing, but fundamentally distinct? (See the quotes in the introduction.) Note that in discussing this issue, I will work with Logins’s version of the normative question as it can be found in the book.

To begin with, note that Logins’s claim is not to be read as stating that one fact that p cannot be both a NRR and a NER to φ. Logins (2022, 176) explicitly allows this, and rightly so. It is highly plausible that often the same fact that explains why an agent ought to φ can also be adduced by her in an argument towards φ-ing. What I understand Logins to say instead is that NRR and NER constitute two different kinds under which the same specific fact can fall. These kinds are distinct because being an answer to an understanding-seeking question is a different matter than being an answer to an argument-seeking question. But what is implied by saying that these kinds are fundamentally distinct? I take this to mean, first, that there is no substantial connection between the two kinds and, second, that falling under these kinds implies very different normative properties of a fact.

As to the first point, for example, being an NRR does not entail being an NER. Moreover, the two kinds don’t stand in the relation of determinable and determinate to each other, as do the kinds bird and owl, say. As I read it, Logins’s claim is comparable to metaphysical disjunctivism in the philosophy of perception. That view has it that genuine perception is fundamentally distinct from hallucination – the two types of mental states do not share the same fundamental kind (e.g. Martin, 2004). On the Eroteric View, the commonalities between the kinds NRR and NER are exhausted by their being answers to the same (or same-sounding) normative question, which is, however, asked with different intent, viz. ‘Why ought S to φ?’

I reject this stark separation of the two roles of explaining why an agent ought to φ and being a premise in a sound argument towards φ-ing (or, equivalently, being an answer to an understanding-seeking normative question and being an answer to an argument-seeking normative question). I agree with Logins that the fact that p is a premise of good reasoning towards the conclusion that S ought to φ nowhere near guarantees that p also explains why S ought to φ, or makes it right that S φ-s (Schmidt 2021). This is because an agent, in her sound reasoning, can use many facts of a situation as indicators of what she ought to do which go beyond the facts that make it the case that she ought to give the response in question. However, that the fact that p explains why S ought to φ does near enough guarantee that p is a premise of good reasoning towards the conclusion that S ought to φ – or so I want to argue here. In general, if a certain fact p contributes to its being the case that another fact q obtains, and so contributes to an explanation why q (and is a good answer to the corresponding understanding-seeking question), p is thereby available to be exploited by reasoners as a means of figuring that q is the case. It can be used as a premise in a sound argument towards q. Reasoners can naturally exploit explanatory relations between different facts to reason towards previously unknown facts about the world. For instance, that water molecules form a crystalline structure at 0° C explains why water freezes at 0° C; a person who knows that water molecules form a crystalline structure at 0° C can use this fact as a premise in sound reasoning towards the conclusion that water freezes at 0° C. It is a reason for her to believe that water freezes at 0° C. That such a substantial connection obtains between explanantia and premises of good reasoning is hard to deny, and should be the default position. The burden of proof is on theorists who want to deny it, including for the special case of explaining, versus reasoning towards, the fact that S ought to φ.

Now Logins can point out that he does deliver an argument against this connection by providing counterexamples to it, such as surprise-party style cases or cases involving Moore-paradoxical considerations. Take Suri, who loves surprise parties. The fact that a surprise party is taking place at her friend’s house explains why she ought to go to her friend’s house, but she cannot use this fact as a premise in good reasoning towards the conclusion that she ought to go there, since the party ceases to be a surprise party once she knows about it and is able to use it in reasoning. Turning to Moore-paradoxical considerations, Maurice is in a burning building, but does not believe that he is. This fact – that Maurice is in a burning building, but he does not believe that he is – is not something that he can rationally believe, and so he cannot use it as a premise of good reasoning towards the conclusion that he ought to check whether there’s a fire. At the same time, this fact explains why it is the case that Maurice ought to check whether there’s a fire (Logins, 2022, p. 78, 71). So in both cases, we have an explanans of why S ought to φ, which cannot play the role of premise in sound reasoning towards the conclusion that S ought to φ. We have an NER that is not an NRR.

I concede that these and similar cases undermine the claim that being part of an explanation of why S ought to φ entails being a premise of good pattern of reasoning towards the conclusion that S ought to φ (that is, the claim that being an answer to an understanding-seeking question entails being an answer to an argument-seeking question). But I submit that the counterexamples don’t undermine the claim that explanantia are, in principle and barring exceptional circumstances, available to reasoners as premises for the relevant conclusion. What the counterexamples show, in my view, is just that this in principle availability can be blocked in particular when some unusual interactions between the reasoner’s psychology and the facts in question prevent her from making use of these facts in her reasoning. This happens, for instance, when believing the premises would falsify them (surprise party case) or undermine their use in reasoning (Moore-paradoxical case). But this does not count against the in-principle connection. Note that my claim is not about how we can define what it is to be a reason. For that project, I do believe that Logins’s counterexamples are problematic. I merely want to insist that there is an intimate connection between being a NER and being a NRR: The fact that p explains why S ought to φ near enough guarantees that p is a premise of good reasoning towards the conclusion that S ought to φ, the exceptions being rare outlier cases that can be independently explained.

Now turning to the second point, does a fact’s falling under the kinds NRR and NER imply very different normative properties for the fact? I take it that Logins will allow at least the normative commonality that both kinds of reasons count in favor of the relevant action. One important difference he diagnoses between NRRs and NERs, however, is that only the former have a weight and can be weighed (Logins, 2022, p. 178, 184/5). That NRRs have a weight and can be weighed against each other is explained by the defeasibility of reasoning patterns, according to Logins. Even where it is good reasoning to move from a certain set of premises to a conclusion, it may nonetheless be bad reasoning to move from the original set plus an extra premise to the same conclusion. If this is the case, the extra reason outweighs the reasons corresponding to the premises in the original reasoning pattern. The weight of reasons is due exactly to this feature of reasoning, where different reasoning patterns, constituted by different sets of premises, need to be considered and compared to determine what the agent really ought to do (Logins, 2022, p. 185).

By contrast, NERs don’t come with weights and cannot be weighed. There are only pro toto NERs. There are no pro tanto NERs that interact in weighing explanations à la Broome which together determine what the agent ought to do. Logins allows that an NER may be merely “a part of … the content of an appropriate explanation providing answer” to the question ‘Why ought S to φ?’ (Logins, 2022, p. 170, my italics) I understand his claims as follows: There is no weighing-explaining, but only standard explaining of ‘ought’ facts. Any NER is a pro toto reason, but this doesn’t mean that only the total explanation is the reason. Rather, any good answer to the understanding-seeking question, ‘Why ought S to φ?’ is an NER to φ, even if it partial. The important point is that these partial explanations can never stand in weighing relations to one another. Each of the reasons corresponding to such a partial answer favors φ-ing in a self-contained way, as we might say.

I am not convinced by this picture. To begin with, it is highly intuitive that all normative reasons, including NERs, come with weights and can be weighed. Weighing is essential to normative reasons not (just) because we weigh considerations in deliberation by considering where our reasoning will take us if we include different premises, as Logins suggests, but (also) because different features of a situation together determine, objectively speaking, whether a certain action ought to be performed. Take the following case, in which private investigator Priya is just starting to investigate a case in which a rich aristocrat was shot and killed. Here are some facts about the case:

  1. (a)

    The butler’s fingerprints are on the murder weapon and Priya doesn’t believe this.

  2. (b)

    The butler has no alibi for the time of the murder and Priya doesn’t believe this.

  3. (c)

    The butler is a convinced pacifist and Priya doesn’t believe this.

None of these facts can be used by Priya in her reasoning, on Moore-paradoxical grounds. So each of them must be an NER for Priya. Each of (a) and (b) is individually a good pragmatic/practical reason for Priya to focus her investigation on the butler; but (a) and (b) taken together provide stronger pragmatic reason to focus her investigation on the butler than each of the facts do individually. Moreover, (a), (b), and (c) together provide weaker reason for Priya to focus her investigation on the butler than (a) and (b) alone. As this illustrates, the relevant features of the situation, even where they are unavailable to the agent’s reasoning, together make up the overall normative profile of her options, and thereby determine which option/action she ought to perform. The weights of the relevant facts as reasons – for or against the action – together fix what the agent ought to do.

So, contrary to Logins, weights and weighing are relevant to normative reasons across the board. The normative properties of NERs and NRRs are not so different. If this is right, this raises the further question whether it is possible for NERs to be weighed against NRRs.Footnote 2 Specifically, what if NERs favor φ-ing, whereas NRRs favor not φ-ing? Consider the case of Gerhard and his cheating partner, in which the following is the case:

  1. (d)

    Gerhard’s partner cheated on him and (successfully! ) made sure that Gerhard doesn’t believe that she cheated on him.

  2. (e)

    A highly reliable friend of Gerhard’s says that his partner would never cheat on him.

In this case, (d) is an NER for Gerhard to break up with his partner, as it is an answer to the understanding-seeking question, ‘Why ought Gerhard to break up with his partner?’ (Not only did Gerhard’s partner betray him, she did so in a particularly conniving way.) By virtue of this reason, it is the case that Gerhard ought to break up with his partner. At the same time, (e) is plausibly a good answer to the argument-seeking question, ‘Why ought Gerhard not to break up with his partner?’, and thus an NRR for him not to break up with her.Footnote 3 It appears that, this reason favors not breaking up with his partner.

What are the connections between Gerhard’s NER and his NRR in this scenario? I see three stances that proponents of the Eroteric View might take: First, they might say that the case involves a dilemma. No matter how Gerhard acts, he violates a practical ought, either the one that is due to his NER (d) or the one that is due to his NRR (e). Second, they might argue – in line with my thesis above that NERs can be weighed – that there is a practical all-things-considered ought for Gerhard in this situation, which is determined by the joint weight of his NER and his NRR. Third, they might deny that the given reasons make it the case that Gerhard both ought and ought not to break up with his partner. Rather, since both options are arguably supported by reasons of roughly equal weight, both actions are permitted.

However, none of the options is attractive. The first option is unhappy because it seems unmotivated to introduce dilemmas of practical normativity wherever there are unbelieved pertinent facts (explaining why S ought to φ) on the one hand and mere indicators of a conflicting practical ought (supporting that S ought not to φ) on the other. Some philosophers do want to allow that there are moral dilemmas in rare cases where each available action violates some fundamental moral norm. But allowing for practical dilemmas in the described kind of cases (which presumably are going to be quite common) takes being open-minded about dilemmas too far – it overgenerates practical dilemmas. The second option is unattractive because it is difficult to fathom how facts as disparate as the fact (d), that Gerhard’s partner cheated on him and made sure that Gerhard doesn’t believe that she did, and the fact (e), that a highly reliable friend of Gerhard’s says that his partner would never cheat on him, could be comparable, or weighed against each other. Since the former cannot enter Gerhard’s reasoning, while the latter is irrelevant to explaining the fact that Gerhard ought to break up with his partner, it’s not clear how they could interact. The third option is problematic seeing as the Eroteric View individuates NERs and NRRs, each in isolation, as answers which explain or support the fact that S ought to φ. Claiming, with respect to my scenario, that NERs and NRRs each give rise merely to a permissible option, creates a tension with the Eroteric View’s characterization of these types of reasons via normative ‘ought’ questions.

One way out of this conundrum for Logins is to adopt my friendly suggestion above and to reconceive the normative question as, ‘Why is it permissible for S to φ?’ By contrast, my favorite escape – which I do not have the space to develop here – is to deny that all good answers to the argument-seeking normative question (however we phrase it exactly) are normative reasons. Instead, let me summarize my critical comment. I suggested that there are substantial connections between NRRs and NERs, and that they share fundamental normative properties such as having a weight. If this is on the right track, it leads to problem cases like Gerhard’s, in which NRRs and NERs conflict and for which it seems that the Eroteric View struggles to provide a good account. In the end, I think that such problem cases raise the question whether the Eroteric View should be abandoned.

Be this as it may, Logins’s Eroteric View is an exciting proposal, with which it has been an honor and a pleasure to engage in this commentary – and which I am confident will stimulate much further debate about the fundamental nature of normative reasons in the future.