In his recent book, Morality and Revelation in Islamic Thought and Beyond, Amir Saemi tackles an important problem that he calls the “new problem of evil”: the problem that Islamic Scripture seems to command actions that our modern moral sensibilities find immoral. For example, Islamic Scripture seems to claim that men are in charge of women, that women’s testimony is less trustworthy than men’s, and that men are allowed to beat their wives in certain situations. Other examples include allowing slavery and killing apostates. While this book is written in the Islamic context, it is clear that many religions face a version of this problem: their Scriptures, on the seemingly most plausible interpretation, command actions that seem to modern readers to be immoral.

Saemi notes two standard solutions to this problem. One solution is linguistic reinterpretation: when we pay more attention to the historical and linguistic context of Scriptural commands, we discover that the actual commands are different than they seem. But matching up Scriptural commands with modern sensibilities requires a radical departure from their apparent and historically understood meaning. Another solution is contextualization: while the commanded actions are wrong in our current context, they were not wrong in the radically different context in which they were issued. For example, in seventh-century Arabia – when the Qur’an was revealed – it was common to be excessively violent towards women; and so the command to only beat women in particular circumstances was a restriction on permissible violence towards women. The problem with this solution, according to Saemi, is that even if restricted wife-beating is better than unrestricted wife-beating, it is still not morally justified, even in the context in which the command was issued.

We are left with two options. One is to hold that Scripture takes priority over our moral judgments. Saemi argues that this view (the Scripture-first view) is the dominant view in the history of Islamic philosophy. The other option is to hold that our moral judgments take priority over Scripture (the ethics-first view). This is Saemi’s view, and he spends much of the book arguing for it.

On the ethics-first view – under the assumption that Scripture is divinely inspired – we have to explain why God would command or allow immoral actions. To answer this challenge, Saemi proposes a view of Scripture that he calls the Legal Interpretation, which says, roughly speaking, that Scriptural commands are legal commands but not moral commands. In Saemi’s words:

To explain why God or the Prophet might legislate immoral actions, Saemi draws on Scott Shapiro’s “planning” conception of the law, according to which the law is a social plan designed to solve moral problems (Saemi, p. 217). A key fact about planning is that it is irrational to make plans that are infeasible – that we know cannot be carried out. Therefore, a legislator is constrained by which options are feasible, and must sometimes legislate rules that are morally non-optimal, if there are no morally optimal rules that are feasible.

One example of this comes from the ethics of war (Saemi, p. 220). It might be true that, in a war, if one side is just and the other is unjust, then it is morally acceptable to kill both combatants and “active” civilians on the unjust side, but it is not morally acceptable to kill anyone on the just side. However, it would be infeasible to enforce these principles: in practice, most countries will think they are on the just side, so the rule “kill only those on the unjust side” would lead to a lot of deaths. The Geneva Convention holds that combatants can kill each other, but not civilians. Even if this is not the morally optimal rule, it leads to less death than alternative rules.

Saemi argues, then, that the rules of Scripture are not morally optimal, but that it is nonetheless good for God or the Prophet to legislate them. Not only are they the best rules that are feasible at the time of legislation, they can help move society towards the morally optimal rules. In the particular case of wife-beating: according to Saemi, Scripture allowed it in certain circumstances, because although it is morally impermissible, allowing it in those circumstances was morally better than the other feasible options (e.g. allowing it in all circumstances), and this legislation can help Islamic society gradually move towards a complete prohibition on wife-beating.

In this article, I will examine these claims more closely. I will first identify four conditions that must be met in order for Saemi’s account to explain why Scripture prescribes actions that are morally non-optimal. To see whether these conditions are in fact met, I will introduce a formal account of coordination. I will then show what function legislation must have in order for it to successfully legislate the non-optimal rules but not be able to legislate the optimal rules. Roughly: it must not function by introducing rewards and punishments, but instead by creating expectations about what others will do and about what others will expect others to do. Given the ethicsfirst solution, the account must also say why the optimal solution is available in the modern day, even without Scripture legislating it. The best way forward for this solution, I argue, is not to hold that people have gotten more prone to virtue; instead, when we examine the process of moral learning, we come to see why legislating a nonoptimal solution can sometimes lead people to eventually converge on the optimal one. I take my solution to be a friendly amendment to and expansion of Saemi’s proposal. I close with some potential worries.

To explain his account of religious laws, Saemi draws on a story by Ibn Tufayl (1105-1185) (Saemi pp. 210-214). In this story, a person named Hayy knows all the moral truths. He travels to a nearby island, full of vicious people who follow the teachings of a prophet sent by God, but these teachings surprisingly don’t align with the moral truths. Hayy tells the people of the island the moral truths, but because they are vicious, it actually makes them behave worse. Hayy thus discovers that non-ideal laws can sometimes have better outcomes than ideal laws, in particular when people are vicious. In Saemi’s words:

According to Saemi, nascent Islamic society is like the island in the story. Legislating morally correct principles would have made people behave worse than legislating morally non-optimal principles; and legislating morally non-optimal principles made people behave better than they would in the absence of any laws.

This account has something in common with the contextualization solution: it points out that Scriptural commands, at the time they were issued, were morally better than existing behavior. But unlike that solution, it doesn’t hold that the contents of the commands are moral truths. According to the Legal Interpretation solution, restricted wife-beating is not, in fact, morally permissible – but issuing the command to do it is nonetheless the best God can do.

To make this account more precise, assume there are three possible norms a society could adopt. The first is a deeply inegalitarian norm, which allows men to beat their wives even for trivial reasons. The second is a moderately inegalitarian norm, which allows men to beat their wives only in extreme circumstances. The third is the egalitarian norm which never allows men to beat their wives. And assume that the egalitarian norm is in fact the morally best norm, followed by the moderately inegalitarian norm, followed by the deeply inegalitarian norm.

Saemi’s account rests on several key claims. The first two:

Together, these claims imply that by legislating the moderate norm, Scripture made things better than the status quo. The third claim:

Along with the first two claims, this claim implies that legislating the moderate norm in fact leads to the best societal outcome, even if it is morally sub-optimal.

Saemi also seems to hold that the moderate norm is not the final resting place: that present-day people could instead adhere to the egalitarian norm. First, Saemi holds that the laws don’t just improve the status quo – they also move us towards the morally optimal solution (Saemi, p. 221). Second, he holds that we currently understand that sexism is wrong, and so permission to beat one’s wife even in extreme circumstances would be a bad social plan in our society today (Saemi, p. 222). Finally, he holds that we should be guided by the ethics-first view in our present society, rather than the Scripture-first view: presumably, following our own ethical ideals will result in a better outcome than following Scripture.

Therefore, Saemi’s final claim is that:

The key question is whether these four conditions can be simultaneously vindicated. To answer this question, we need to look more closely at individual motivation and rationality – and, in particular, at how individual preferences interact to produce the relevant facts about what people will do, both in the absence of Scripture and in the presence of various Scriptural commands. In particular, we must examine the general nature of coordination and convention, and how norms are maintained and changed.

In his influential book Convention, David Lewis uses simple game theory models to analyze what conventions are, how they solve coordination problems, and how they are formed and maintained. A coordination problem is a situation in which two or more individuals are trying to choose actions in light of interests that coincide. For example, we are each trying to decide which side of the road to drive on, and our primary aim is to drive on the same side of the road as each other, to avoid accidents. We can describe our situation using a game matrix, where the numbers represent the “utility” payoffs to each of us for each combination of actions, Row’s payoff followed by Column’s (Figure 1). The magnitude of the numbers is inessential: what matters is their relationship to each other.

In this scenario, we don’t care which outcome we coordinate on, as long as we coordinate. In other scenarios, we might have a preference about which outcome we coordinate on, even if coordination is more important to us than otherwise satisfying our preferences. For example, consider the choice of what to wear to work. We are most interested in wearing the same type of clothing, so as to not be embarrassed; but beyond that, we both find sweatpants more comfortable than khakis, and we both find khakis more comfortable than suits. Our situation can be represented in Figure 2.

The numbers represent three important facts. First, holding fixed what the other person does, your highest payoff can be found by matching their action. Second, if you come close to matching them, but don’t match them exactly, you still get a moderate payoff: if you wear khakis, then I am only mildly embarrassed by wearing sweatpants or a suit, since each of these “isn’t too far off” from khakis. Third, being far away from matching the other person is the worst of any of the options: if you wear a suit when they wear sweatpants, or vice versa, you will be very embarrassed.

What each of us does depends not just on our own payoffs, but on our expectations of what the other will do. If I expect you will wear khakis, for example, then that is a reason for me to wear khakis. But it is also important that you expect that I expect you will wear khakis. If you expect that I expect you will wear a suit, then you will expect that I will wear a suit, and so you will wear a suit. In this way, coordinating on the same action both requires and is supported by higher-level expectations.

We say that we have “solved” a coordination problem when we end up in one of the “coordination equilibria”: one of the outcomes in which neither of us could do better by either one of us unilaterally acting otherwise (both wearing sweatpants, khakis, or suits).

One way to solve coordination problems is explicit agreement: if we say to each other that we will both wear khakis, then we will both expect that the other will wear khakis, we will both expect that the other will expect us to wear khakis, and so forth. But coordination problems can be solved without explicit agreement. They can also be solved by precedence: if we both wore khakis to the office last week, we will have higher-order expectations that we will wear khakis. They can also be solved by salience: if we have no other reasons to coordinate around any particular equilibria, we might each choose the one that “sticks out.” An important way for an equilibrium to stick out is for it to be the unique equilibrium that everyone prefers.

Coordination problems don’t just take place between two people; they also occur in populations. Lewis says that a convention arises among members of a population in recurrent situations of the relevant form when everyone conforms to some regularity and expects everyone else to do so, and when these facts are common knowledge.

Importantly, once a convention arises, it is rational for each individual to stick to it, and irrational to deviate. For if I can reasonably expect that others will do their part, then I do best by doing my part too. Thus, conventions are maintained because it is individually rational for each person to do their part. It is also possible for conventions to change, similarly to how they are formed in the first place. If I come to expect that others will do their part in a different convention, then it will be rational for me to do my part in this new convention, rather than the old one. The rational force of conventions thus comes from each individual’s preferences, plus each individual’s higher-order expectations about each other’s behavior.

According to the Legal Interpretation, Scriptural commands help people coordinate around an outcome that is superior to what they would do if left to their own devices. We can clearly see this in the example of commands telling people when to pray, if it is important for everyone to pray at the same time (Saemi, pp. 229-230). But this example has two specific features: we are neutral about what times to coordinate around (as we are about which side of the road to drive on); and when to pray is morally neutral. Can the theory of convention also explain why Scripture would legislate morally non-optimal actions rather than morally optimal ones? More generally, can we give a model of the situation that satisfies all four of the conditions outlined in the previous section? I turn to this question now.

Let us assume that each individual man has three options: inegalitarian actions, moderate actions, or egalitarian actions. Now, recall the four claims that Saemi’s solution rests on. Given PAST VICE, let us assume that – in the absence of any facts about what others will do – men in nascent Islamic society preferred inegalitarian actions to moderate ones to egalitarian ones. But let us also assume that men care more about coordinating – taking the same actions as each other – than about satisfying their personal preferences. This is exactly analogous, in terms of each man’s preferences, to the situation of dressing for work (distasteful as the comparison may be!). As in the workwear example, we will assume that if men don’t do the same actions as their fellow men, they fare better insofar as they are closer to doing what their fellows do rather than farther away. The assumed situation is represented in Figure 3.

These numbers represent men’s preferences, and the example of Hayy and the island certainly implies that these preferences arise from viciousness. But I note that this is not essential to the model itself: preferences could, for example, be “moral” preferences about what would be best for the family unit. For example, the preferences of someone who thinks inegalitarian actions are morally better and wants to act in accordance with moral good, but personally dislikes carrying out the inegalitarian actions, could be represented by these numbers.

It is natural to think that in the absence of Scripture, everyone will take the inegalitarian actions, and the inegalitarian norm will thus become a convention. For given that everyone prefers the inegalitarian actions, in the absence of reason to coordinate around anything else, everyone will form higher-level expectations that everyone will take the inegalitarian actions.

Finally, let us assume that the correct moral ordering is Egalitarian ≻ Moderate ≻ Inegalitarian. And let us imagine we are in the position of legislators, and we are trying to get people in this society to adopt the morally best convention we can get them to adopt.

There are two ways in which legislation can produce conventions. I argue that the first way cannot vindicate both of the conditions about effectiveness, but the second way can.

We have seen how three of the conditions on which Saemi’s account rests can be vindicated. The final piece of the puzzle is to vindicate PRESENT VIRTUE. We need to explain how individuals in the present day will choose egalitarian actions merely as a result of thinking they don’t need to follow Scripture in our current time – rather than as a result of, say, new revelation legislating a new convention.

I’ve provided an account of how to vindicate the four claims on which Saemi’s Legal Interpretation is based. I briefly summarize its elements. First, men in nascent Islamic society preferred to be less egalitarian rather than more, but cared more about following the same norms as each other than satisfying their preferences. Second, they were able to form higher-order expectations about following moderate norms, but not about following egalitarian norms, perhaps because these norms seemed inefficient or they believed others would believe they were inefficient. Third, following moderate norms allows men test out egalitarian norms, and they eventually come to realize they prefer them. Therefore, in the present day, men will follow egalitarian norms in the absence of Scriptural injunctions.

A few worries remain about the account sketched here, and about the Legal Interpretation as a solution to the new problem of evil. First, why didn’t God create a world in which men have a more efficient way to learn which actions are better? After all, knowing the true commands sooner would result in less harm to women. The account provided holds fixed what men are like and considers which commands are optimal in light of their nature, but we could instead ask why they have this nature; this is an instance of the (old) problem of moral evil.

Second, as mentioned, Islamic Scripture is supposed to be the final revelation. There is something unsatisfying about the final revelation containing non-optimal commands, especially if we are now in a position to coordinate around the optimal commands.⁶ Indeed, if Saemi’s story is correct, then it would certainly be less confusing if God were to issue new commands (or provide an expiration date for the old ones), especially given that some people think Scripture does describe the morally correct norms.

Perhaps a proponent of the Legal Interpretation solution could argue that the morally correct norms are only feasible given that the sub-optimal ones are recommended. The idea might be that each individual’s moral journey must necessarily follow the historical journey described: a person early in their moral development is inclined to the bad actions and can only be moved by Scripture to the better but sub-optimal ones. By accepting Scriptural constraints, each individual will experiment with the morally correct actions and come to see that they are preferred – a fact that (a proponent of this view might claim) can only be learned through experience rather than Scriptural testimony. That is: that Scripture is a kind of scaffolding in our individual lives as well as in society. This is a fairly revisionary view of Scripture, but it is an available view.

Third, as we saw, if religious laws are to have the function they need to have to vindicate the solution, then their motivational force cannot be the force of reward and punishment; instead, it is the force of higher-order expectations. This also seems like a fairly radical departure from how Scripture is traditionally conceived. In addition, it makes it difficult to receive guidance from Scripture in the present day if all it tells us is the optimal way to resolve social problems in light of the feasibility constraints produced by past societies. So this kind of view would need to say more about the role of Scripture in guiding religious people today.

The Legal Interpretation is a novel and interesting view of Scripture. But the devil (or God) is in the details. A solution to the new problem of evil must explain how Scripture can be efficient but not too efficient, and how past men could want the worst but present men could want the best. I’ve provided an account on which all these conditions can be satisfied. It rests on the recalcitrance of expectations when “outlandish” actions are recommended, and an account of moral learning where these actions are tried out when costs are not too high, and found to be the best. Whether this story is a fully satisfying religious account of Scripture is a further question, but I hope I’ve at least articulated an account that rests on plausible facts about humans and their development, and illuminated how non-ideal norms can play a role in discovering the ideal ones.