Sam Harris, the Naturalistic Fallacy, and the Slipperiness of “Well-Being”
This post is about the main argument of Sam Harris’s new book The Moral Landscape. Harris argues that there are objective truths about what’s morally right and wrong, and that science can in principle determine what they are, all by itself. As I’ll try to demonstrate here, Harris’s argument cannot succeed. I call the argument “scientistic” because those who take (a variation of) its first two premises to be obvious are led to exaggerate the importance of scientific measurement for determining what’s morally right, and correspondingly to underestimate the importance of moral reasoning and moral philosophy.
Harris commits what philosophers call “the naturalistic fallacy”: of attempting to draw conclusions concerning what we ought to do (normative conclusions) directly from premises that are purely factual, or scientific, and value-free (purely descriptive premises). Today I will show why such a move is fallacious, and draw attention to the way that Harris’s use of the ambiguous term “well-being” masks the fallacious move that his argument makes use of.
Here, then, is a schematic version of Harris’s argument: The Scientistic Argument
Premise 1.1) The right action is whichever action maximizes well-being.
Premise 1.2) Well-being is the balance of [conscious states C].
Premise 1.3) Scientists can measure the balance of [conscious states C].
Conclusion) Scientists can (indirectly) measure the rightness of actions.
Harris does not come down very clearly in favour of any one particular set of conscious states that he takes to constitute well-being, so I’ve left a placeholder in the argument in square brackets. You should imagine the placeholder as having been determinately filled in in whichever way Harris thinks appropriate. It has to be filled in one way or another so that scientists will know which states to measure. One traditional account of well-being that Harris seems sympathetic to in places is the classic utilitarian definition in terms of pleasures and pains: the greater well-being one has, the greater the balance of one’s pleasures over pains.
Here’s another argument; I’ll call it The Accountancy Argument:
Premise 2.1) The most economically successful business is whichever business makes most profit.
Premise 2.2) Profit is the balance of income over outgoings.
Premise 2.3) Accountants can measure income and outgoings.
Conclusion) Accountants can (indirectly) measure the economic success of businesses.
To be truly compelling, an argument needs to have both a set of undeniable premises, and a conclusion that logically follows from them. The Accountancy Argument has these features. And The Scientistic Argument seems superficially analogous to The Accountancy Argument. If it is genuinely analogous, then it too must be a compelling argument, and we will have to accept its conclusion.
Roger Crisp’s post on this blog last week points toward one important disanalogy between The Scientistic Argument and The Accountancy Argument. Premise 2.1 of the Accountancy Argument is undeniably true because the most economically successful business is by definition just the same thing as whichever business makes most profit. In contrast, premise 1.1 of the Scientistic Argument does not seem to be true by definition (though the possibility that it is will be considered later).
This need not be a fatal objection: premise 1.1 could still be defended by showing that its two concepts “right action” and “action that maximizes wellbeing” refer to the very same property (just as the concepts “heat” and “total kinetic energy of the atoms in an object” refer, scientists have discovered, to the very same property, though the terms were not defined to mean the same). But Crisp raises the worry that the property of being the right action and the property being whichever action maximizes well-being are not the very same thing: they might be, as Derek Parfit claims, “too different” to be the same thing. After all, we can’t possibly recognize that something is the right action without recognizing that there is some sense in which we ought to do it. Being the right action, as Parfit says, is a normative property. But we surely could know that some action would maximize well-being (in the sense defined by premise 1.2: maximizing the balance of [conscious states C]), and still legitimately ask whether there is any sense in which we ought to do it. Being the action that maximizes the balance of [conscious states C] seems, unlike the property of being the right action, to be a purely descriptive property: it describes how the world is or could be, without yet telling us what we ought to do about it. It seems unlikely, then, that the two concepts in premise 1.1 could refer to the same property.
As Crisp points out, there is still a third possible way to defend premise 1.1: Even if its concepts refer to two different properties, premise 1.1 might still be true if the rightness of an action “is anchored” in, or supervenes on, its maximizing well-being (i.e. maximizing the balance of [conscious states C]). This is to say that if two actions differ as to their rightness, then they must also differ as to whether they maximize well-being, although rightness and maximizing well-being are not the very same thing. On one view of the mind, this is similar to the relationship between mental states and brain states – you can’t have a change in whether you feel tired, for example, without a corresponding change in the state of your brain – even though feeling tired is not the very same thing as having a brain state of a certain kind.
The difficulty with this move for Harris, as Crisp correctly recognizes, is that in both these cases – mental and moral – scientists are unable to measure the supervenient property directly. So the question arises: How do we know about the supervenient property, or about the supervenience relation (and hence about the truth of premise 1.1)? We only find it appealing to think that feeling tired supervenes on brain states because each of us starts from our own individual experiences of feeling tired. Scientists cannot directly measure these conscious feelings in other people; they can only measure their behaviours and brain states (for example, by using surveys, or MRI scans). If our world were populated with zombie or unconscious robot scientists and a few unscientific conscious people, we might reasonably wonder whether the scientists would have any idea that feelings of tiredness even exist, let alone know anything about their supervenience relations to the other, measurable properties.
A magnified form of the same problem arises when it comes to moral rightness. Moral realists sympathetic to The Scientistic Argument might want to claim that we can have individual “experience” of the property of rightness, just as we have individual experience of the feeling of tiredness. If this claim could be defended, perhaps we could use it to argue for premise 1.1. However, there are a couple of serious difficulties for the claim that we “experience” the property of rightness: First, we often seem to deeply disagree about which are the right actions. This is so not just in real world cases, but in hypothetical ones where the natural facts can be agreed on by stipulation (e.g. Suppose that over their whole lifetimes, Blue would have a well-being of 10, and Red a well-being of 5, all other things are equal, and you could either give an additional 6 units of well-being to Blue or 5 to Red. Which would be right? Those who care most about equality will answer one way, those who care most about the total will answer another.) This disagreement makes our “experience” of rightness look, at best, highly unreliable. Secondly, moral realists have provided no plausible explanatory account of how it is that human beings have the ability to experience the supervening property of rightness. This makes it difficult to see how human experience of such a property could could even be possible. (My own view is that we should respond to these worries by abandoning the moral realist claim that rightness is discovered by human beings rather than constructed by them. Pace Harris, this need not mean that morals are relativistic, or just a matter of opinion, or that there are no moral truths.)
Unlike in the case of The Accountancy Argument then, we have difficulty justifying the very first premise of The Scientistic Argument. But Harris might now reply that I am just being difficult: perhaps I should accept premise 1.1 as true because it is obviously true. How could anyone possibly doubt that the right action is the action that maximizes well-being? To say that some action would generate more well-being than another, Harris might say, is just to say, in different words, that it would be better. And how could the right action be any other than the best one?
There are some questionable steps in this reasoning to do with aggregation, which I’ll set aside. I admit that premise 1.1 does understandably invite the kind of reading on which it can easily be seen as tautologically or obviously true, just like the first premise in the Accountancy Argument which is true by definition. This makes the Scientistic Argument seductive, but misleadingly so. The trouble here is in the slipperiness of the term “well-being”. For “well-being” is, in our ordinary language, a fundamentally normative term. Understanding “well-being” in the ordinary way, suppose there is a population of morally praiseworthy individuals in World1, in which each has a well-being of 1, and an otherwise equal World2, in which each member of the population has a well-being of 2. Then World2 is better, and (importantly) it is better by definition. If we use this ordinary, normative definition of “well-being” to understand premise 1.1 of The Scientistic Argument, then that premise may seem obviously true. But then it becomes tempting to jump to a different definition of “well-being” when we come to premise 2.1: for example, we may then be tempted to define “well-being” as “the balance of [conscious states C]”. Without defining “well-being” this way, premise 2.1 cannot be obviously true: we would, at the very least, need an argument for it. But if we do define “well-being” as “the balance of [conscious states C]” in premise 2.1, our Scientistic Argument argument then commits the fallacy of equivocation. Moreover, if Parfit is correct in his very plausible claim that normative properties and descriptive properties are just too different to be the same thing, then these two definitions of “well-being” are flat-out incompatible.
In The Accountancy Argument, the three premises could all be true by definition. But for moral realists at least, the first two premises of The Scientistic Argument cannot both be true by definition. For at least one of these premises, a difficult question has to be faced: How do we know this to be true? And science cannot answer that question. Above, I showed the difficulties that arise for defending premise 1.1 if we define “well-being” in a purely descriptive way. If we instead define “well-being” in a normative way – for example, as “the measure of that which makes a person’s life better” – then similar difficulties will arise for defending premise 1.2, as Kwame Anthony Appiah’s fine review of Harris’s book points out. The question then is: Which kinds of things increase or decrease a person’s well-being, and how much does each of them count in relation to the others? Conscious states? Knowledge? Activities? What about things that you don’t know about, or that even happen after your death, such as your book becoming famous, or your reputation being impugned? If well-being is understood as a normative property, then these questions are, precisely, questions about the nature of the supervenience relation between the moral and the natural; questions that cannot be answered by science. Yet the answers to these questions are anything but obvious: we must engage in moral reasoning, and moral philosophy, to find them.