The World Makes More Sense if You're a Moral Realist

Or: Why the explanatory value of moral properties should be obvious to anyone

Pictured: Frank Jackson, who is also right about illusionism

A few months ago, I wrote a short piece explaining why I thought a classic argument against moral realism didn’t work. I won’t summarize the whole thing here, but it dealt with the idea that moral facts were “explanatorily inert” - that is, the idea that moral facts couldn’t play a meaningful role in our explanations of events, even if they did exist. My response, which was really just explicating a response by the exceptional philosopher Nicholas Sturgeon, ended up being more of a negative case for why we can’t rule out the explanatory value of moral facts. But now I want to make a more positive case for why I think moral facts actually can (and do!) play a central role in many of our most productive explanations. To do this, I’ll be talking a lot about some concepts that were first explored by Frank Jackson and Philip Pettit in their 1990 paper “Program Explanation: A General Perspective” and Mark Nelson’s more recent “Moral Realism and Program Explanation.” I hope it can help explain my confidence in moral realism and give other moral realists a handy tool for defending their ideas.

In order to defend the idea that moral facts can play a legitimate role in good explanations, it’s worth asking what makes a good explanation to begin with. Obviously, this is a huge question with no single agreed-upon answer! But at least one thing most people accept is that explanations should be sensitive to counterfactuals - that is, if one event explains another, then if the first event didn’t happen, the second shouldn’t have either. Think about something simple, like waking up and seeing a puddle of water on your kitchen floor. If you think the puddle is explained by the fact that your dishwasher broke, then you should also think that the puddle wouldn’t be there if the dishwasher was working fine; if someone said, “I think the broken dishwasher explains the puddle on your kitchen floor, but it would be there even if the dishwasher didn’t break,” then you would think their explanation wasn’t a very good one.

To tighten this idea up philosophically, we can say something like: If X is a good explanation for Y, then it should be the case that, in the closest possible world in which X does not take place, Y does not take place as well. Call this the counterfactual test. Now, there are a bunch of complications here that you would need to acknowledge in a broader theory of what constitutes explanation; for example, there are “over-determined” situations where multiple things occur that would each bring about the same result on their own, or “composite” situations where two events together bring about one result that neither one could bring about on their own. But none of these complications should impact the basic point I’m about to make, so I think it’s fine to ignore them for now.

The counterfactual test is a very common sense standard, but it can lead to some conclusions that you might not expect. Let’s say you step outside on a winter day and leave a little cup of water on your doorstep. You come back an hour later and see it’s frozen solid. What explains this fact? It might be tempting to pull out a thermometer, see it’s fifteen degrees Fahrenheit at the moment, and say “The fact that the air temperature is fifteen degrees Fahrenheit explains why the cup of water froze.” But this explanation clearly fails the counterfactual test! After all, the closest possible world where it was not 15 degrees outside would be one where it was 14 or 16 degrees outside instead, and in those worlds, the water would still freeze. While the water freezing was caused by contact with air that was 15 degrees Fahrenheit, that fact about the air in and of itself can’t be a sufficient explanation for the freezing; it’s just one way the freezing happened to take place.

So what should you do if you want an explanation that does pass the counterfactual test? No specific fact about the air temperature will work, because the closest possible world in which the temperature is different will also be one in which the water still freezes. If you want to escape this bind, you’ll have to jump up to a fact about some higher-level property that obtains in all (relevant) situations where the water freezes and none where it doesn’t - specifically, the fact that the air temperature was below 32 degrees Fahrenheit. This is the explanation that will pass the counterfactual test, because the closest possible world in which the air temperature is not below 32 degrees would be one in which the water didn’t freeze. To be clear, that doesn’t mean this higher-level property has some causal power over and above the actual temperature of the actual air that actually froze the water. It just means that this specific higher-level property really does tell you more about why the water froze than any individual reporting about the actual temperature of the actual air.

In general, most explanations of even remotely complex or contingent phenomena will ultimately incorporate higher-level properties like these. Another great example given by the philosopher David Brink is overpopulation: Even if any individual nation’s being overpopulated is entirely reducible to some large set of individuals being born in that nation, that set of individuals being born can’t properly explain any consequences so long as an entirely different set of individuals being born would have the exact same impact; rather, what truly explains those consequences would be the particular higher-level property of being overpopulated that any arbitrary set of births could realize. This sort of dynamic obviously generalizes to all sorts of domains, from physics and biology to economics and sociology. In fact, it’s actually very rare that anything other than the simplest kinds of straightforward phenomena are best explained by a singular description of the specific, first-level events involved.

These sorts of explanations are sometimes referred to as program explanations, because they give information about a particular higher-level property “programming” the lower-level behavior that actually causes the event to take place. This contrasts with process explanations, which just tell you the details of the first-level processes that actually occurred. In other words, program explanations are able to do more than just tell you what actually caused such-and-such to happen; they can also tell you the conditions under which such-and-such would or would not have happened in a way that identifies the truly relevant property at hand. This is obviously a very valuable feature for an explanation to have, and one that would make any higher-level property necessary for securing it more than justifiable in terms of our larger ontology. If there’s some plausible higher-level property out there that can get you a really good program explanation for something you care about, then you have every reason to accept that property into your general category of things that exist (in the same way that you would accept the property of being overpopulated or the property of being below freezing temperature as well).

But what exactly does this all have to do with moral realism? The answer is simple: If facts about higher-level properties have a place in our best explanations of events, then there’s an obvious and appealing role for facts about moral properties specifically in many situations that call out for explanation. Consider, for example, someone who donates $216 to the Red Cross and is commended for it by her friend. What explains the friend’s decision to commend her? The first-level fact that she donated $216 here fails the counterfactual test for the same reason the first-level fact that the air was 15 degrees in the first example also failed: Because, in both cases, the event to be explained would still have occurred in the closest possible world in which those facts were different. The nearest possible world where the first-level fact about her donation is different would be one in which she donates $217, or $215, or maybe the same $216 to Doctors Without Borders instead. But regardless, her friend still would have commended her. What actually passes the counterfactual test would be the fact that she demonstrated the higher-level property of generosity, because that’s the fact where, were it not the case, she would not have been commended. So if a program explanation is legitimately meaningful in the case of the freezing water - and I think it uncontroversially is - then it should also be legitimately meaningful here. And if moral properties play a legitimately meaningful role in a functional explanation, then we have good reason to accept that they actually exist, because non-existent entities shouldn’t be able to explain anything.

Of course, we shouldn’t need these technical concepts in the first place to accept the fact that moral properties can feature in good explanations. In our daily lives, we explain things by reference to moral properties all the time! We say that a particular manager’s success with his employees is explained by the fact that he’s conscientious and kind, or that a particular firefighter’s decision to volunteer for a dangerous rescue mission is explained by the fact that he’s brave. We bring up the moral character of dictators like Putin or peacemakers like Martin Luther King Jr. to explain why they made the decisions that they did, and how those decisions were accepted by the public at large. We even rely on the moral properties of systems and practices to explain historical events, like when (borrowing an example from Nicholas Sturgeon) we say the abolitionists were driven to fight slavery because it was an unjust and cruel institution. And all of these facts clearly pass our counterfactual test where reference to reductive first-level facts would fail, which is good reason to think they really do make for better explanations.

Statements like the ones above aren’t even remotely controversial outside the philosophy classroom, and excising them from our general speech would require radically restructuring how we analyze and interpret human behavior. The concept of a program explanation, therefore, isn’t something that finally gives us a reason for accepting moral explanations; it’s something that helps us better understand the justificatory foundation for the moral explanations we already accept in our daily lives. Nevertheless, many people are still surprisingly hostile to the idea that moral properties can truly pull their weight in an explanatory sense. So let me turn to one of the most common objections I see to this framework, which goes something like this:

Sure, we might talk about certain actions or events as having moral properties and even reference those properties in our explanations of other actions or events. But talk about moral properties is really just a shorthand for talk about properties that compel certain subjective judgments - whenever we say an action was (for example) cruel, we just mean it was the sort of thing that makes most people render the judgment, “That’s cruel.” So when someone says that (for example) abolitionist opposition to slavery was explained by the fact that the slave trade had the property of being cruel, what they really mean is that it was explained by the fact that the slave trade had the property of possessing many first-level features that lead most people to render the judgment, “That’s cruel.”

There are two reasons I think this objection clearly fails. The first is that, as I discussed in my earlier post, this move is generally available for pretty much any property you could possibly name, up to and including the property of existing externally in space. In fact, I’d go so far as to say that, given a willingness to bite some bullets, it’s always possible to take an objective higher-level property and “psychologize” it into a subjective one defined in terms of the judgments it invokes. The property of being a fish could always be replaced with the property of being the sort of thing that makes people say “There goes a fish,” the property of being French could always be replaced with the property of being the sort of thing that makes people say “That’s French,” and so on. Human beings necessarily encounter the world through a process of gathering experiential data and judging that data accordingly, so there will always be some experience and judgment accompanying any particular property that human beings know anything about. What we need is a reason to believe that, in this particular case, we’re justified in rejecting the external referents of our moral concepts in favor of a purely psychological account.

Furthermore, when we do frame this sort of argument specifically in terms of moral properties, we run into the second issue: It’s unclear how exactly you could give an account of the features necessary to instantiate the property of reliably compelling morally thoughtful people to make a judgment of (for example) cruelty that doesn’t collapse into a description of the features necessary to instantiate the property of cruelty itself. Imagine, as an analogy, someone who believed there were no entities with the property of being an apple tree, but only entities with the property of having features that make botanists believe them to be apple trees. If they were to actually try listing out the features necessary for the property of having features that make botanists believe an entity is an apple tree - a certain sort of branches and leaves, a certain sort of fruit, a certain sort of climate conducive to growth, and so on - then they would just end up describing an apple tree, at which point it’s unclear what exactly they’re rejecting.

The same is true for moral properties like cruelty. If someone did attempt to identify the features that instantiate the property of reliably compelling morally thoughtful people to make a judgment of cruelty, then the most obvious candidate would be something like callous indifference to another person’s suffering. But that’s all cruelty is to begin with! There are some properties where this isn’t the case, like the property of being a witch; in that case, the features necessary to instantiate the property of being the sort of person taken to be a witch by others are very clearly distinct from the features necessary to instantiate the property of actually being a witch, so it makes sense to affirm the former while rejecting the latter. But what makes people think an action is cruel just is cruelty. Any attempt to excise this property itself in favor of a related constellation of “judgment-making” features is going to either be woefully under-described or else amount to a roundabout description of the property itself, in which case the skeptic’s objection is entirely semantic.

There is obviously a tremendous amount more I could say, both about the argument I’m giving and the objection to it that I’ve raised. But hopefully, this piece can still function as a simple introduction to a sort of moral realism that avoids the pitfalls so many anti-realists assume are essential to realism itself. Holding this view, we don’t need to see moral facts as “extra bits” inertly layered over a world completely devoid of value; we can accept them as part of the world itself, in the same way chemical, biological, and sociological facts are, and build theories incorporating them that explain the full range of data we encounter. I find this vision of moral realism both intuitively plausible and practically productive, and would encourage opponents of moral realism to rethink the fundamental assumptions that lead them to reject it.