Life-time epistemic utility

I’ve been thinking about the diachronic aspects of epistemic utility. In the case of non-epistemic utility, we can get a decent first approximation to life-time utility by adding up (or, if time is continuous, integrating) momentary utility. But I think this works less well for the epistemic case. For many things of purely epistemic importance, figuring them out is much more important than when one figures them out. (Granted, figuring them out earlier is instrumentally epistemically valuable, because it gives one more time to use leverage the knowledge to figure other things.)

Here’s an extreme version of encoding the value of “figuring something out”. Assuming one does not suffer from mental decline, the epistemic value of one’s life is the epistemic value of the very last moment of it. It’s interesting to note that this won’t work. For imagine that no matter what other credences you had at a given time, you always set the credence of “This is the last moment of my life” to one, while being careful to (inconsistently) make no use of this credence in update. If only the last moment counts, this modification to your credences would be a good idea: it makes sure that when the last moment comes, you get the epistemic utility credit for it.

I suspect that other weightings that favor later over earlier beliefs will suffer from a similar problem—they make it a good idea to err on the side of pessimism about how close death is.

But at the same time, I think some sort of favoring of later beliefs over earlier ones seems appropriate. I don’t know how to resolve this difficulty.

Gender/sex minimization and Christian sexual ethics

Here is a way to live a life: Generally strive to minimize the number of cases where one’s having a particular sex or gender functions as a reason for one’s actions or emotional attitudes.

An extreme version of this is not compatible with traditional Christian sexual ethics unless one is planning on celibacy. It is also not compatible with American law which requires one to correctly fill out various forms, such as census forms, that ask what one’s sex is. And it is not compatible with common-sense morality which requires one to respect things like sex- or gender-segregated bathrooms.

A moderate version of such a gender-minimizing practice, however, could be perhaps sustainable within the bounds of traditional Christian practice, American law and common-sense morality.

One might think that given the heterosexualism of traditional Christian practice, it is hard in romantic contexts to avoid basing decisions on reasons like “I’m a a man and she’s a woman.” In this post I want to explore the idea that one could instead base one’s romantic actions and attitudes on: “We are an opposite-sex pair.”

One might object: “That’s cheating. The reason why the two are an opposite-sex pair is that one is a man and the other is a woman.” But this is “reason” in a different sense of “reason” from that of reasons for actions and attitudes. That one is a man and the other is a woman is a metaphysical ground of the two being an opposite-sex pair, and it may well be one’s epistemic reason for thinking the two are an opposite-sex pair. But it need not be one’s reason for, say, asking the other out on a date—the reason for asking the other out on a date could just be “we are an opposite-sex pair”, and of course the delights of the other’s person, even if the evidence and metaphysical ground for “we are an opposite-sex pair” is the more fine-grained fact that one is a man and the other is a woman.

One could do the same thing when discerning a vocation to the priesthood. Instead of thinking “I’m a man, so I should consider the priesthood”, one might think “I am of the opposite sex to the symbolic sex of the Church (which in turn is the opposite sex to the sex of the incarnate Word), so I should consider the priesthood.”

Would formulating one’s reasons for action in such an unusual way have any benefits? I think so. “We are an opposite-sex pair” focuses one on a relation between the persons. On Trinitarian grounds, there is reason to think that relationality is central to personhood. Half of “I am a man and she is a woman” is self-focused. Better to use “we” than “I” in romantic thinking.

Moreover, we perhaps shouldn’t focus on what is morally irrelevant to a decision. Suppose two people are a good romantic match in terms of character traits, interests, etc., and one is male and the other is female. Supposing (perhaps per impossibile) that their sexes were swapped, but their character traits stayed the same, plausibly they would still be a good romantic match. That they are of the opposite sex is relevant romantically, assuming traditional Christian sexual ethics. But perhaps which one is a man and which one is a woman is not very relevant—what’s relevant is that the couple has one of each sex.

The picture in the above exploration is that the significance of men and women is largely relational: there is a relationship possible to a man and a woman that is not possible to two men or to two women. This is presumably because a man and a woman are an opposite-sex pair, or a potential mating pair, or something like that. This fact about a man and a woman is a relational fact. Granted, this relational fact is metaphysically grounded in certain biological features of the man and the woman, which features may not be themselves relational (say, the existence of body parts with a certain shape, or at least of activated genetic coding for them; though even there the teleology of the parts is relational). But even if the features are not themselves metaphysically relational, their ethical significance could still be largely relational.

Of course, in the end, the physical consummation of love in marital union will require each party to pay attention to the sexed nature of their own and the other’s bodies. That’s unavoidable. But perhaps that’s just a detail? I doubt it’s just a detail myself, but I could very well be wrong.

Another objection to the above story is that in love we focus on the specific features of the other, and in romantic love this includes sex-linked physical features. So when Juliet loves Romeo, she does not love him just as a “someone of the opposite-sex with character traits T₁, ..., T_n”, but also as someone with a rich set of lovely physical features, for which it is important that he is male, as many of them would be aesthetically and biologically unfitting in a woman. Agreed! But that doesn’t mean that Juliet’s own femaleness needs to be a part of her reasons for loving Romeo. Instead, she can love him as “someone of the opposite-sex to me with character traits T₁, ..., T_n, and with physical features Φ₁, ..., Φ_m which are splendidly fitted to his maleness.” Of course that Romeo is of the opposite-sex to Juliet and that Romeo is male implies that Juliet is female, but even though it implies this, it need not be a part of her reasons for love. I do like the thought that we should minimize focus on self in other-love.

Perhaps the above story isn’t right. I don’t endorse it. It’s entirely hypothetical. I find some features of the story attractive, but my credence in the story is well below 50%. There may well be ethically significant non-relational features of being and being female. Perhaps swapping the sexes of a well-matched romantic couple might in fact change whether they are a good match.

For instance, maybe there are specifically male virtues and specifically female virtues, and then swapping sexes while keeping character the same produces someone who lacks virtues that they should have. Maybe, but I am inclined to be skeptical of this suggestion. A more moderate view would be C. S. Lewis’s, that although men and women should have the same virtues, the lack of certain virtues in a man is worse than their lack in a woman and vice versa. That has more of a chance of being right. I could imagine future scientific research telling us that the typical hormonal make-up of men tends to make some virtues easier for them and the typical hormonal make-up of women tends to make other virtues easier for them. Lacking an easier virtue seems worse than lacking a harder virtue, other things being equal. If so, then if you swapped sexes while keeping characters unchanegd, the moral evaluation could change. Perhaps Alice and Bob are respectively a decent woman and a decent man, but Alice would make a terrible woman and Bob would make a terrible man. Maybe. But if they both got worse symmetrically in this way, while keeping the same actual virtues, maybe it wouldn’t make a big difference to their romantic relationships. They’d still have the same virtues between the two of them, the same vices between the two of them, it’s just that the evaluation of these virtues and vices would be a bit different. And in any case on a story like the hormonal one, it’s not clear that the specific sexes matter except causally, as tending to produce the hormones.

My thinking on this was sparked by two things. First, for a while I’ve been exploring imagining what life would be like if we were isogamous heterothallic organisms, which of course we are not. Second, I recently attended the defense of a very interesting dissertation arguing among other things that it is compatible with Catholic orthodoxy to hold that gender (though not biological sex) distinctions are a major result of original sin.

Epistemic utilities and death

In the previous post, I proved that we get a proper scoring rule if we compute epistemic utilities as follows. We start with our current credence assignment, consider what credence assignment we will have in the future after we update on some further evidence, and then score that. I then suggested that one could get a lifetime epistemic utility by adding up the epistemic utilities over all the moments of life, and as long as death wasn’t random—as long as the lifespan was fixed—this would generate a proper scoring rule. I then said that if death is random (as it is) it might be the case that you don’t get a proper scoring rule.

My conjecture was wrong. You still get a proper scoring rule despite random death. It’s easy to see this. The basic idea is this. Suppose that you might die the next moment. This partitions the probability space into two subsets, D and L, for death and life. Your current credence is p. Next moment, on D, your credence either doesn’t exist (because you don’t exist or you exist in some supernatural state where you don’t have credences) or doesn’t count (because I am after lifetime credences). Thus, the appropriate way to do a forward-looking scoring of your credence p is to score it s(p_L) on L and 0 on D, where p_L is the result of conditionalizing your credence on the evidence L (after all, if you are alive, you will conditionalize on being alive), and s is some proper scoring rule. In other words, your forward looking score is s_L(p) = 1_L ⋅ s(p_L).

Is this score proper? Yes! For by propriety of s we have:

• E_pL(s(p_L)) ≥ E_pL(s(q_L)).

But this is the same as:

• (p(L))⁻¹E_p(1_L⋅s(p_L)) ≥ (p(L))⁻¹E_p(1_L⋅s(q_L)).

Multiplying both sides by p(L) (I am assuming a non-zero probability of survival), we get:

• E_p(s_L(p)) ≥ E_p(s_L(q)).

We can combine this with a more complex set of future investigations as in the previous post, and things will still work.

It is crucial to the above argument that when you’re alive, you can tell you’re alive. I suppose that’s not always true. When you’re asleep, you are alive, but can’t tell you’re alive. So to generalize beyond the above toy example, replace death with unconsciousness or something like that.

Forward-looking scoring rules

An accuracy scoring rule assigns a score to a probability function representing an agent’s credences, ostensibly measuring how close that probability function is to the truth. The score s(p) of a probability function p is a random variable, because the value of the score depends on what is actually true, i.e., on where we are in the probability space.

A proper scoring rule (on probabilistic credences) satisfies the propriety inequality

1. E_ps(p) ≥ E_ps(q)

which says that the expected score of your current lights—your current credences p—by your current lights is optimal: you won’t improve your expected score (by your current lights) by switching to a different credence q.

You can think of a proper scoring rule as representing the epistemic utility of having a credence p.

But now let’s think about things dynamically. In the future, you will receive additional evidence. As a good Bayesian agent, you will update on this evidence by conditionalization. Perhaps instead of thinking about maximizing your current score, you should think about maximizing your future score. Maybe your true epistemic utility is the score you will end up with after all the future evidence is in.

A simple model of this is as follows. There is some finite partition I = (I₁,...,I_n) of your probability space Ω with each cell I_i of the partition representing a possibility for what you might learn given future evidence. Your current credence function is p, and p(I_i) > 0 for all i. There is then a random credence function p_I where p_I(ω) is the credence function you will have once the evidene is in if you are at ω ∈ Ω. In other words, p_I(ω)(A) = p(A∣I_i) where I_i is the member of the partition that contains ω. (Technically, the function that maps ω to p_I(ω)(A) is equal to the conditional probability p(A∣G) where G is the algebra generated by I.)

Now, given a proper scoring rule s, define a new scoring rule s_I as follows:

2. s_I(p)(ω) = s(p_I(ω))(ω).

Your s_I-score for p at ω then represents the score you will have at ω once you learn which cell of the partition I you are in.

Theorem: The scoring rule s_I is proper if s is proper.

Note that s_I won’t be strictly proper (i.e., (1) won’t always have strict inequality when p and q are distinct) if I has two or more cells, because p_I and q_I are going to be the same if p and q assign different probabilities to the cells, but have the same conditional probabilities on each cell. But it might still be the case that s_I is strictly proper with respect to some relevant subfield of Ω—that needs some further investigation.

Suppose now you are a Bayesian agent who is guaranteed to consciously live for n moments. In each moment, new information comes in. Thus, we have a sequence J₀, ..., J_n of finer and finer partitions, with J₀ being the trivial partition, and with p_Jk representing the credence you will have at time k. Your overall epistemic lifetime score is then:

3. s_Σ(p) = ∑_ks_Jk(p).

It follows from the Theorem that s_Σ is a proper scoring rule if s is. And if J₀ is the trivial partition, then s_J0 = s, and so if s is strictly proper, then the lifetime score s_Σ is strictly proper, since the sum of a strictly proper rule and a proper rule is strictly proper. So, lifetime scores are strictly proper if they are constructed from an instantaneous score—in the above toy model.

Alas, the toy model is not fully adequate, because it is random when we will die, and so our lifespan doesn’t have a fixed sequence of moments. Once we take into account the randomness of when we will die, the overall epistemic lifetime score might stop being proper: this needs further investigation.

Proof of Theorem: By the Greaves and Wallace Theorem, an optimal method of updating credences with respect to expected proper score is by Bayesian conditionalization. Apply the Greaves and Wallace Theorem to the scoring rule s and the starting credence p with the following two strategies:

A. Bayesian conditionalization on the true cell of I.

B. Switch your credence from p to q, then apply Bayesian conditionalization on the true cell of I.

Saying that (A) is at least as good as (B) is equivalent to the the proper scoring rule inequality (1) for s_I.

Aristotle on sexual reproduction

Infamously, Aristotle thought that in animal reproduction (including the human case) the male contributed form and the female the matter.

I think it would have made for a neater metaphysics if he thought it was the other way around. For his story involves the complication that there are three ways that a form of an animal exists:

1. as informing a piece of matter and making it an animal of the relevant type

2. as being transmitted by semen, without turning the semen into an animal of the relevant type

3. as being found abstracted in an intellect thinking about the animal.

But if he supposed instead that the form was contributed by the female, he could have said that the semen contributes a small packet of initial matter for the organism and the female’s form works upon that packet in utero. At no point in this process is there a form that is separated out from an organism with that form (such a case will still have to happen when an intellect abstracts that form from matter), and so case (b) disappears, along with all the consequent metaphysical complications to the system.

It’s not clear to me how exactly Aristotle handled this. In Generation of Animals, he talks of the “power of the male residing in the semen”. This sounds like a kind of virtual presence of the form of the animal, a presence of the form as implicit in a causal power. It would be nice to be able to skip this complication. (Granted, Aristotle would still have a similar problem with regard to his view of delayed ensoulment, on which the offspring of a human is first a kind of plant, then gains an animal form, and then gains a rational form, for then the plant has a power to make an animal and the animal a human.)

I guess one might try to find Aristotle a way of reducing (b) to (a) while making the male still be the contributor of form as follows. One might think the semen is a living part of the male’s body, metaphysically like other parts, but separated spatially from the rest of the body (given what we know about how we are made of molecules, it is clear that spatial separation cannot be a bar to bodily unity). Thus, the semen would still be informed by the father’s form, in exactly the same way that the father’s ear or heart is. But this would have a downside. Suppose the father apparently died while the semen continued living, as does happen. Then it seems that the father would not have actually died—instead, his body would be reduced to a packet of semen, and he will have literally gone to seed. This seems absurd.

Why did Aristotle not go for the female as source of form solution, which is so neat metaphysically? Maybe for two reasons. First, the semen obviously could only contribute very little matter. (But then why not suppose that the initial organism is very small?) Second, sexism.

A really big choice

A messy Irenaean world which leaves much to be desired and then grows in value due to our contribution has a great value.

A perfectionistic world where everyone is already perfectly virtuous and everything is good and we just maintain that level of good also has great value.

Quite plausibly the two worlds have incommensurable value.

God could just choose between these incommensurable values. But what God does is something else: he leaves it to the first humans to choose. This seems an especially great gift: we get to choose the basic shape of the value that our world takes.

We sin, and we get the Irenaean world.

Little evils

From time to time I wonder why it is that we focus on big evils in discussing the problem of evil. After all, first consider this argument:

1. As far as we can tell, Alice’s cancer is a gratuitous evil.

2. So, probably, there is a gratuitous evil.

3. If there is a God, there is no gratuitious evil.

4. So, probably, there is no God.

And then this one:

5. As far as we can tell, Bob’s stubbed toe is a gratuitous evil.

6. So, probably, there is a gratuitous evil.

7. If there is a God, there is no gratuitious evil.

8. So, probably, there is no God.

The same kind of backing can be given to (1) and (5). We’ve thought long and hard about Alice’s cancer and its connections to the rest of the world, and can’t find a theodicy. We’ve thought long and hard about BOb’s stubbed toe and its connections to the rest of the world, and can’t find a theodicy. And the rest of the two arguments seems to flow the same way. But as far as I can tell (!), nobody’s become an atheist because of arguments like (5)-(8) while many have become atheists because of arguments like (1)-(4).

What’s going on?

Here is one answer, which at times I’ve felt some pull towards: we shouldn’t take (1)-(4) as seriously as we do.

Here is another answer: we should take (5)-(8) more seriously than we do.

But there may be a good conservative answer, one that maintains our intuition that (1)-(4) should be taken seriously while (5)-(8) shouldn’t. There is a difference in the two arguments, and it lies in the difference between the inference from (1) to (2) and the inference from (5) to (6). In the cancer case, a good that would justify God’s permission of the evil would have to be a very large good. In the stubbed toe case, a justifying good could be pretty small. But it is much easier to overlook a small good than a very large good. Thus, (5) confers much less probability on the existence of gratuitous evil than (1) does.

When I first thought of this conservative answer, I had the image of an elephant: it’s hard to overlook an elephant in the room, but a smaller thing is easier to overlook.

Is that a good image? Are bigger goods easier to see?

Sometimes. A tiny increment in pleasure may be hard to see just as it’s hard to tell the difference in color between two nearby shades of green. If the stubbed toe led to Bob having a slightly greater appreciation of his health the next day, he might not notice.

On the other hand, some really big goods are things we may take for granted, and they need to be pointed out first by some wise person for us to expressly notice them. For instance, the value of human dignity. However, goods that we take for granted are likely irrelevant here—for goods that we take for granted are unlikely to be goods that need Bob’s cancer to exist.

Furthermore, it may not be all about smaller things being harder to see. Consider Rover the dog. How likely is it that there is a cat bigger than Rover? Well, it is more likely when Rover is small than when Rover is big. Similarly, if an evil is smaller, then it is more likely that it is connected to good big enough to justfy it than if the evil is bigger.

So, I think, one can make a case that the inference from a cancer to gratuitous evil is better than the inference from a stubbed toe to a gratuitious evil. But is it much, much better? That I am not sure about. Yet we take (1)-(4) very seriously and we take (5)-(8) not seriously at all. Even if we are right to take (1)-(4) more seriously than (5)-(8), perhaps we make too big a difference between them. Perhaps we shouldn’t take (1)-(4) quite as seriously as we do, and perhaps we should take (5)-(8) more seriously, and perhaps both!

Christ's concrete human nature

One of the old puzzles of Christology is why it is that in us the soul and body together compose a person, but they do not do so in Christ.

The problem is intensified if one thinks that there is an entity composed of the soul and body of Christ, an entity one might call “Christ’s concrete human nature”. For the person of Peter just is the entity composed of the soul and body of Peter, while in Christ the quite similar (except in sin) entity composed of the body and soul is not a person, but rather is united to a person.

The standard way out of this difficulty is to say that being a person is an extrinsic feature of a concrete human nature. Thus, Peter’s concrete human nature is a person because it is not united to a person distinct from it. But Christ’s concrete human nature is not a person because it is united to a person distinct from it. A problematic consequence of this is that personhood is not an intrinsic feature of an entity.

But there is a better way to get out of the difficulty. Just deny that in Christ there is any entity composed of his soul and body, and any statements about “Christ’s human nature” we need to make about his concrete soul and body should get paraphrased as statements about two entities, Christ’s soul and Christ’s body. Since there is no such entity as Christ’s concrete human nature, there is no puzzle as to why this entity isn’t a person.

This also solves another problem. If there is such a thing as Christ’s concrete human nature, then plausibly it thinks human thoughts, just as Peter’s concrete human nature thinks human thoughts. But Christ also thinks human thoughts. So in Christ there are two human thinkers: Christ and Christ’s concrete human nature. This is implausible! Granted, we can get out of it by making human thinking extrinsic as well, but each time we do the extrinsicness move, we intensify puzzlement.

But there is still a problem: Why is it that Peter’s soul and Peter’s body compose a whole—namely, Peter’s person—while Christ’s soul and Christ’s body do not compose a whole? Here we should make an extrinsic move. Christ’s soul and Christ’s body do not compose a whole because they are personally united to something else. Peter’s soul and Peter’s body, however, are not personally united to something else (I don’t count composition as personal union).

It may seem like at this point we have made no progress: we are back to an extrinsicness move. But this extrinsicness move is plausible in non-theological cases. Suppose Fred is missing a left leg. Then Fred’s soul, head, arms, torso, and right leg compose a whole. But Peter’s soul, head, arms, torso, and right leg do not compose a whole. Why the difference? Because Peter’s soul, head, arms, torso, and right leg are personally united to something else–namely, a left leg. But Fred’s soul, head, arms, torso, and right leg are not personally united to something else.

There is another worry. Suppose we embrace survivalism and think that Peter continues to exist in a disembodied way after death and before the resurrection of the body. But now Peter’s concrete human nature is reduced to a soul, and the soul does exist. So now Peter’s soul is a person, while Christ’s soul is not, even though both exist and are intrinsically alike (apart from accidental features). However, I think there is a way out of this, too. We deny that Peter’s soul is a person. The easiest way is to go four-dimensionalist: Peter’s soul is just a part of a four-dimensional entity that includes a body, though the body doesn’t happen to exist at this time. A less easy, but still tenable, way is to say that disembodied Peter has the soul as a proper part, with no co-part, and classical mereology needs to be modified to avoid Weak Supplementation (which is the axiom that says that if an object has a proper part, then it has another non-overlapping part).

We should not make human-like AI

1. Either AI with human-like behavior is a person or is not.

2. We should not deliberately produce AI with human-like behavior when that AI is a person.

3. We should not deliberately produce AI with human-like behavior when that AI is not a person.

4. So, we should not deliberately produce AI with human-like behavior.

Obviously, the big question is whether (2) and (3) are true.

In favor of (3), a non-person with human-like behavior evokes emotional responses from us that are only apt as directed at persons. These emotional responses are of great moral importance to our life, and some of them are constitute a recognition of the object of the emotion as a being with dignity, a sacred being, and to have such emotional responses to something that lacks the relevant dignity blurs the central moral distinction between persons and non-persons.

In favor of (2), we have several arguments. Start with some Kantian ones. AI is an artifact. When we make artifacts, we make them to serve our purposes. To make a person to serve our purposes is to treat the person as a mere means to an end. And that’s wrong. This argument, I think, applies no matter what our purpose is: even if our purpose is to make the AI live its own free life. That’s still our purpose for it, and we have no right to impose a purpose on a person’s life.

Furthermore, by designing a digital person, we are designing, in its fundamentals what its basic purposes in life are and we are thus exercising a mode of control over another person that we have no right to have.

A final but least principled Kantian argument is that even if we “set free” the digital person—whatever exactly that means—it is pretty much impossible to protect digital persons from being enslaved by other humans.

There are also some non-Kantian arguments. If an AI is a person, that person will eventually be very cheap to keep in decent existence as compared to a flesh and blood human being. Since we have the duty to protect the life of a person when doing so is not an undue load on limited resources, we would have the duty to keep any person AI that we spin up running indefinitely. This is problematic as it ties the hands of future human generations, by imposing on them what one might call “an unnatural duty” to keep running all the AI persons that we make, with little benefit to the future human generations from this. The problem is the worse the greater the numbers of digital persons that we spin up. This problem is akin to the problem of frozen embryos in fertility clinics if these embryos are persons (which I think they are).

There is also a dilemma. We have the duty to protect the life of persons. But at the same time, there is something deeply unappealing about something of the level of sophistication of an ordinary human mental life extending an order of magnitude longer than the typical human life-span (rescaled as needed to take account of differences in processing speed). The most appealing religious accounts of afterlife involve a radical transformation, e.g., theosis or parinirvana. I think many of us rightly would feel that living a thousand years of the kind of life we now have isn’t appealing, though we wouldn’t mind an extra ten or twenty or maybe even hundred years. If we were ever able to indefinitely extend human life without an undue resource cost, we would find ourselves in an inextricable moral dilemma: on the one hand a duty to protect life when doing so does not carry an undue resource cost and on the other hand the monstrousness of living an order of magnitude longer than ordinary human life should be. But with digital persons, indefinite life extension would be easy, and so the dilemma would be unavoidable.

Finally, we would take a significant moral risk in designing digital persons. Training processes involve vasts amount negative feedback. We just do not know how unpleasant that might be.

Rowe's argument from evil revisited

In his 1979 classic paper, Rowe says that a necessary condition for an “omniscient, wholly good (OG) being” to fail to prevent an evil s₁ is that at least one of the following three conditions holds:

1. there is some greater good, G, such that G is obtainable by OG only if OG permits s₁, or

2. there is some greater good, G, such that G is obtainable by OG only if OG permits either s₁ or some evil equally bad or worse

3. s₁ is such that it is preventable by OG only if OG permits some evil equally bad or worse.

We might put it this way: Rowe is saying that OG will prevent all gratuitous evils, and an evil is gratuitous if none of (i)–(iii) are satisfied.

First, note that (i) can be omitted, since if we have (i), we have (ii) as well. That’s a nice simplification: we only need (ii) and (iii).

More interestingly, however, note an ambiguity in “there is some greater good”. To see the ambiguity, consider this apparent counterexample to Rowe.

Suppose both theological compatibilism and Molinism are false, and God decides to give Alice a free choice between a great moral good G₁ and a tiny moral evil s₁, in circumstances where Alice is very likely to choose G₁. Absent theological compatibilism and Molinism, God cannot, however, ensure that Alice will choose G₁. Suppose, also, that free choice is not valuable as such, but significantly freely choosing well is very valuable (where significant freedom is freedom to choose between right and wrong), and much more valuable than choosing well without significant freedom. Finally, let’s suppose Alice goes for s₁. It is reasonable to say that s₁ is non-gratuitous, for preventing s₁ would lose the opportunity for the great good of Alice significantly freely choosing G₁. But an opportunity for a good may not count as a good. Condition (iii) is not satisfied—if God didn’t give Alice the choice between G₁ and s₁, no other evil would have to be permitted. But neither is (ii). For the good G of Alice significantly freely choosing G₁ does not exist, since Alice does not in fact choose G₁.

For many years I thought that this was a counterexample to Rowe. Today it hit me that this might be uncharitable. When Rowe says “there is some greater good”, he might mean a greater possible good or a greater type of good. Thus, in my example there is a greater possible good or a greater type of good, that of Alice significantly freely choosing G₁, albeit this is a good that isn’t actually instantiated.

Thus, charitably, let’s take Rowe to be quantifying over possible goods (or, pretty much equivalently, types of goods). A weak support for this is that the “some evil” in (ii) clearly quantifies over possible evils, so it makes the conditions neater to quantify over possible goods and evils.

Here is another problem for Rowe’s account. Suppose (ii) and (iii) are not satisfied, but instead we have:

4. there is a good, G, such that G is obtainable by OG only if OG permits s₁ or something at least as bad while the only way that s₁ is preventable by OG is by permitting some lesser evil s₂ or something at least as bad as s₂, and that G plus s₁ is better than s₂.

It seems that in case like this, s₁ need not be gratuitous. But note that a case like this does not require either G to be a greater good in comparison to the evil s₁ nor does it require s₂ to be at least as bad as s₁. For intance, suppose the value of G is 10, the value of s₁ is  − 11 and the value of s₂ is  − 3. Then G plus s₁ has value  − 1, which is better than the  − 3 we get by preventing s₁ at the cost of s₂ and loss of G. In cases like this, (ii) and (iii) need not be satisfied, and yet the evil need not be gratuitous. In other words, there cases where we have a kind of mix of (ii) and (iii). (It’s hard to think of examples, but it’s also hard to think of examples of (iii). I think the examples would have to involve divine promises.)

I also worry about incommensurability. For simplicity, let’s suppose there is no incommensurability: all values and disvalues are comparable. Also, let’s suppose that there are only finitely many scenarios in play, so we don’t have to worry about infinite sequences of less and less bad options and stuff like that.

Still Rowe’s necessary condition for non-gratuity have become more complicated: (ii) or (iii) or (iv), and with the quantifiers ranging over possible goods and possible evils. And once conditions get so complicated, one starts to worry that we are missing something. It would be nice to simplify the conditions.

Here is my attempt at such a simplification:

1. There is a possible scenario where s₁ is permitted by OG which is at least as good as any possible scenario where s₁ is prevented by OG.

Before continuing, I want to worry a bit about the following. Suppose that by preventing an evil s₁ you lose a good G, where G is not a greater good, but an “equal good”, i.e., the value of G equals exactly the disvalue of s₁. In that case, it seems plausible that a morally perfect being would prevent s₁, since there is a presumption in favor of preventing evils. I will take avoid this worry simply by assuming that it’s better to have neither G nor s₁ than to have both G and s₁ in cases where the value of G equals the disvalue of s₁, i.e., that axiology has a presumption in favor of no-evil. Otherwise, we need to slightly modify (1).

Suppose (1) is true. Let G be the sum total of the goods in the scenario A that’s at least as good as any possible scenario where s₁ is prevented by OG. It follows that any way of preventing s₁ involves either the loss of G or a scenario where the totality of evil is greater than s₁. If G is a greater good compared to s₁, then we thus have (ii). If G is empty, then it seems that there are no goods in the scenario, but we must have greater evils in the s₁-prevention scenarios than in A, and so we have (iii). Finally, if G is non-empty but not greater in comparison to $s_1 then I suspect we are going to have (iv) or something very close, but it’s a bit too complicated for me to think through in detail at the moment.

Moreover, (1) is intuitive as a necessary condition on God’s permitting an evil, in a way in which the triple disjunction (ii) or (iii) or (iv) is probably too complicated to be intuitive. I take it, then, that (1) is a friendly amendment to Rowe’s atheistic argument.

But at this point there is an interesting skeptical theist move. Rowe needs to be able to be epistemically justified in pointing to some evil s₁ and denying (1). But that’s pretty hard. Claim (1) existentially quantifies over all possible scenarios where s₁ is permitted by OG. It’s not that easy to know that one of them has the property indicated in (1), given the vast range of scenarios where s₁ is permitted by OG.

Aquinas on Incarnation as a rock or animal

Aquinas thinks it would be unfitting for God to become incarnate as a non-rational creature. In Summa Theologiae III.4.1, his reasoning is that human nature had an appropriate dignity, as it “was apt to contain the Word Himself in some way through its operation, by knowing Him and loving Him”, but not so for the non-rational creatures. In Summa Contra Gentiles 4.55 we have a different and more philosophical argument:

We must also observe that the rational creature alone acts of itself, for irrational creatures are more driven by natural impetus than act of themselves [magis aguntur naturali impetu quam agant per seipsas]; thus they are to be regarded as instruments rather than principal agents. Now it was fitting that God should assume a creature capable of acting of itself as principal agent, because an instrumental agent acts through being moved to action, whereas the principal agent acts itself and of itself. Accordingly, if God were to do anything by means of an irrational creature, all that is required in accordance with that creature’s natural condition is that it be moved by God, and there is no need for it to be assumed into personal union so as to act itself, since its natural condition does not admit this, it being admitted solely by the condition of rational nature [quia hoc eius naturalis conditio non recepit, sed solum conditio rationalis naturae]. (Shapcote, with italicized phrases retranslated.)

In neither text does Thomas deny the metaphysical possibility of God being incarnate as a nonrational creature, such as a rock, plant or animal. While this is an argument from silence, I think it is a pretty strong one: if he thought it was impossible, we really would expect him to say so, rather than merely saying it was unfitting.

Both arguments seem to me to be based on the idea that if it’s fitting for God becomes incarnate as X, if there is to be much of a point to this incarnation, this X needs to be of a sort that can express the divine person in a special way. If the Logos were to become a piece of granite, the petrous behavior of the rock would not differ from that of any other piece of granite. But if the Logos becomes a human, the rational behavior of this human is able to distinctively express the Logos in a way that ordinary human persons do not. We see this in the Gospels: Jesus as a human being, in his free human choices, movingly expresses the personality of the Second Person of the Trinity. He is just as human as all of us, but all of us humans are different from one another in our operations in a way that is expressive of the person. Of course, all pieces of granite are different from another, but these differences are not due to the operations of the rock differently expressing the rock’s hypostasis (we shouldn’t say person!). Rather they are due to the differences in the matter the rock was formed from and the external forces. God can be expressed in granite.

Of course, every rock reflects God’s goodness, and some rocks especially so, say because they are formed into a beautiful statue or inscribed with a meaningful text. But the special expression of God’s goodness is just a function of the rock’s initial matter and external influences. If God wants a rock to particularly well express him, there is no advantage to God’s becoming the rock: “all that is required … is that it be moved by God”—and all rocks are ultimately moved by God. But it is different for a human, that is a self-originator of its behavior. For the self-originated agency to be fully an expression of God, the person who is human, from whom that behavior originates, needs to be a person who is God. There is thus something that the Logos can accomplish by taking on a rational nature that cannot be accomplished in any other way. We might say that all of creation is God’s self-expression by creatures, but in becoming a human, God can be self-expressed in a creature—something that could not happen if God were to become a non-rational creature. This highlights how it it is that the Incarnation is also an ultimate completion of the act of creation. God’s becoming a non-rational creature would be just God showing off, rather than God manifesting his love by living that love as a human.

TWo closing thoughts. First, I think Aquinas’ Contra Gentiles argument works best if non-rational creatures function deterministically. Given quantum mechanics, we have good reason to think they are indeterministic. But nonetheless the basic point remains: randomness is not really a self-expression, so even if the Logos were to control the quantum processes from the inside, it seems that as long as that control was fully in accordance with the nature of the non-rational creature, it wouldn’t be distinctively expressive of divinity.

Second, I am not sure that Aquinas’ Contra Gentiles argument fits well with Aquinas’ theological compatibilism, on which God can through primary causation make us freely behave a particular way. At least we can say that the argument works better if we have theological indeterminism. And thus theological indeterminism has a theological advantage: it gives us a deep reason for why the incarnation as a human is fitting.

AI and emotion

In some of my work, I use the example of a pill which gives one that warm glow that one has when one has done something sefless and morally good, but which pill one can take when one hasn’t done anything like that, just to feel good about oneself. This is wrong, because that warm glow emotion is too important morally for it to be the subject of counterfeiting. Moreover, I think it remains wrong to take the warm glow pill even if one fully knows that one hasn’t done the morally good deed that it fakes the feeling of.

Generalizing, I think we shouldn’t deliberately induce emotions in contexts where they are inapt when these emotions have a significant amount of moral importance. We shouldn’t induce them in ourselves nor in others. For instance, we shouldn’t try to make others feel like we are their friends when we are not—even if they fully know that that the feeling is misleading.

Now, there is a multitude of significantly morally important interpersonal emotions that are only apt as reactions to another person’s actions. These include feelings of being the object of good- or ill-will, feelings of gratitude or resentment, a feeling of not being alone, and of course a feeling of being a friend. Such emotions have a significant amount of moral importance. We should thus not try to induce them deliberately.

But I think a plausible case can be that current AI chatbots are tuned (both through feedback from users and the system prompt) to produce emotional reactions that are of this interpersonal sort—the communications of the chatbot are tuned to make one feel that one’s concerns are care about. And since the chatbots aren’t persons, the emotions are inapt. The tuning is thus morally wrong, even if any sensible user knows that the chatbot has no cares.

One can, sometimes, have a double-effect justification of inducing misleading emotions, when doing so is an unintended side-effect. However, given that leaked system prompts do in fact have instructions about emotional cadence, it is very implausible to think that the induction of inapt emotions is an unintended side-effect.

A couple of days ago, Anthropic offered me a decent chunk of money for doing some part-time review of the reasoning capabilities of one or more of their models. I turned it down because of moral concerns along the above lines.

I note that double-effect can, however, justify using a chatbot when one does not intend an inapt emotion that one expects in oneself (e.g., I find myself feeling grateful when I get a good AI answer), when the goods gained from the use are sufficient in comparison to the significance of the inapt emotion. But I think the risk should be taken into account.

This is all rather similar to St. Augustine’s infamous concerns about stage drama. But I think one can make a distinction between the cases. Interpersonal emotions can be categorical or hypothetical. Categorical disapproval is apt only when a person has done something morally wrong. But we also have hypothetical disapproval: we can imagine someone hypothetically acting in some situation, and then have a feeling of disapproval towards that hypothetical action. I think there is a real felt difference between these two feelings, just as there is a real felt difference between seeing a sunset and imagining a sunset. And, perhaps, the audience of a dramatic performance one only has—or at least should only have—the more hypothetical feeling.

Metaethics and emotion

Suppose that someone notes that the digits of π starting at some position n start looking like they have an odd pattern, and it occurs to someone to turn them into ASCII characters (three digits at a time, say). The result is a surprising English list:

• Murder, cruelty, torture, unjustified promise breaking, ….

The list goes on in the same vein, until after a long time comes a period.

Moral philosophers think hard, and realize that anything that is clearly wrong is on this list, that nothing on the list is such as to be clearly non-wrong, and indeed that the Prohibition Hypothesis that necessarily π at position n has a list of all and only the morally prohibited actions is quite plausible. Most moral philosophers end up accepting the Prohibition Hypothesis, and applied ethics enters a golden age where we are able to figure out whether some action is wrong simply by deciding whether it falls under some description on the list.

Obviously, the Prohibition Hypothesis still presents philosophical puzzles. Is this just random choice? Is there a God who has power even over the digits of π? Is English metaphysically special?

But now imagine that some metaethicists hypothesize that what it is for an action to be prohibited is for it to fall under one of the descriptions in the list in π starting at position n.

I think this metaethical hypothesis is something we would have good reason to reject. That growing people for organs falls under one of the descriptions on the list (say, the first one) gives us good reason to think that growing people for organs is wrong. But it’s surely not what makes growing people for organs wrong. For what makes growing people for organs wrong surely has nothing to do with what the digits of π are.

How could we back up this intuition? It is tempting to say something like this:

2. That an action falls under a description in the list at position n in π does not give me a decisive moral reason to omit the action.

But that begs the question. On the metaethical hypothesis in question, what it is for there be to a decisive moral reason for ϕing is for non-ϕing to fall under a description in the list.

Is there a way to argue against the metaethical hypothesis in a non-question-begging way? I think so.

3. There is a certain set E of emotional attitudes that is appropriately had to the fact that ϕing is prohibited.

4. The attitudes in E are not always appropriately had to the fact that ϕing falls under a description in the list in π starting at position n.

Given 3 and 4, we should not say that what it is for an action to be prohibited is for it to be on the list. We are not begging the question here. The question of what emotional attitudes are appropriate is not in general a question of what is morally obligatory. Note also that one can hold 3 and 4 while accepting the Prohibition Hypothesis. For emotional attitudes are hyperintensional. Every theist agrees that a child is smiling if and only if God knows that a child is smiling, but that a child is smiling and that God knows that a child is smiling do not appropriately give rise to the exact same emotional attitudes.

The above suggests that we have an emotional test for a metaethical theory: when the theory claims that what it is for an action to be prohibited (or required or whatever) is for it to satisfy some predicate F, we can criticize the theory if we can plausibly argue that different emotional attitudes are appropriate to the action’s being prohibited (or required or whatever) and to its satisfying F.

I suspect one can run an argument against divine command theory along these lines. The appropriate emotional attitudes (say, of disapproval) with regard to ϕing in light of ϕing being morally prohibited are different from the attitudes with regard to ϕing in light of God commanding us not to ϕ.

Mental properties and union dualism

Suppose union dualism is true (I am a composite consisting of soul and body) and I am looking at a red cube. Then I am consciously aware of a red cube in virtue of my soul having some property instance M. Now, this property instance M makes me have a perception of a red cube, but it doesn’t make my soul have a perception as a red cube, because otherwise two things have that perception, and we have the too many thinkers problem. Thus considered as a property of my soul, we shouldn’t describe M as “the property (instance) of perceiving a red cube”. Instead, it is the soul’s property (instance) of “being an x such that x’s possessor perceives a red cube”.

Is M also a property (instance) of me, or just of my soul? Indeed, can the same thing be a property (instance) of two things? Maybe. It sure seems that a watermelon is partly red because its flesh is red, and yet we shouldn’t multiply instances or tropes of red, so there is a single instance of red—which makes the part red, thereby making the whole partly red. Thus, the redness can be thought of as both a property of the flesh and of the watermelon. It is an immediate property of the flesh and a mediate property of the watermelon. As a property of the flesh, it is the property of being red. As a property of the watermelon, it is the property of being partly red. Perhaps, though, this is a merely semantic question. We might say that “property possession” is immediate property possession or that it is mediate or immediate property possession.

In any case, insofar as M is a property of my soul, there is a sense in which it is a mental property and a sense in which it is not. As a property of the soul, it is not a property of thinking, perceiving, believing or of any other of the usually named mental activities, because the soul engages in none of these, since these are activities of me, not of my soul. But at the same time, it is a property that makes me perceive, and hence can be considered mental in that sense. Perhaps we might want to say this: M is a proto-mental property, and to have a mental property is to have a soul that has the requisite proto-mental property.

In any case, I think the above considerations show that arguing for Cartesian dualism—the view that I am a soul—on the grounds that the soul has mental properties and yet there are not two thinkers is too quick (Merricks argues like that here). In one sense, the soul has mental properties, but these properties do not turn the soul, only its possessor, into a thinker. In another sense, the soul has only proto-mental properties.

Maybe it’s a bit like this. Water isn’t moist, but it makes other things moist. Water does this by having some property—a property the physical chemist will have more to say about—in virtue of which anything that has the right association with the water is moist.

Antinatalism and proxy consent

Consider the consent argument for antinatalism (the view that human reproduction is morally wrong) that we shouldn’t impose great risks on people without their consent, but by having a child one imposes great risks on them (life may turn out really badly), and one cannot get consent.

One obvious response is that we are willing to accept proxy consent in medical decision making when the patient is incapable of consent. And of course in the case of a child, normally the first people we turn to for proxy consent are the parents. So it seems reasonable to think that proxy consent from the parents should take care of any worries about the child’s not having consented to come into the world.

What kind of a response can an antinatalist give?

One response is that the whole idea of proxy consent in medical cases is itself bogus. We just should not do medical procedures on children or other non-competent persons that are of a sort that would require consent—say, procedures that are highly invasive or significantly change the course of future life in ways that are not unambiguously for the better. If it is a choice between amputating a child’s leg or letting the child die, we should just let the child die, because a choice whether to live without a leg is the kind of choice that a person should make, and if they cannot make that choice themselves, no one should make it for them.

This seems implausible. It seems pretty clear that we shouldn’t let a child with a gangrenous leg die when the parents consent to amputation. But perhaps the antinatalist can try to distinguish the cases of procreation and amputation. In the amputation case, the child already exists and will suffer a painful death if we do not amputate. Perhaps that overrides the autonomy concerns to a sufficient degree that proxy consent becomes sufficient to justify the amputation. I think that doesn’t work. We can suppose that the child is unconscious, and that in the absence of amputation death will be painless. And it still seems permissible to amputate with parental permission.

A response that I’ve seen is that the fact that standard medical proxy consent cases differ from the procreative case because one cannot even give proxy consent on behalf of someone who doesn’t exist. But this response seems to me to be rather problematic. For if one thinks one cannot give proxy consent on behalf of someone who doesn’t exist, it seems that by the same token one cannot violate the autonomy of someone who doesn’t exist.

But actually both ideas seem false. We can violate the autonomy of someone who doesn’t exist. For instance, we can manufacture a robot that is going to be dormant for 25 years, and will then activate, choose a random 20-year-old and forcibly perform cosmetic surgery on them, radically changing their appearance. By manufacturing the robot we are violating the autonomy of someone who doesn’t exist. And similarly it seems quite possible for prospective parents to provide proxy consent for emergency medical procedures prior to the existence of their children. For instance, supposing (in violation of antinatalist norms) they have several children and expect to have several more, and are afraid that they will forget to sign emergency medical consent forms for their future children, so they sign a blanket consent form for single-leg amputation in case of otherwise untreatable gangrene for all of their present and future children.

Another possible response is that while it’s acceptable for parents to give proxy consent for medical procedures for their incompetent children, the life they create in procreating will one day be the life of a competent person who never agreed. However, this just doesn’t create a disanalogy, because in a number of cases of proxy consent for medical procedures, the child’s entire future life is affected, including the future life once the child reaches competency—this is true in typical cases of amputation.

Furthermore, the idea that one cannot give proxy consent for things that can have significant negative effects on the future life of a competent person, but one can in the case of things that only affect the parts of life when the person is incompetent, leads to the following really weird consequence. A couple is ordinarily not permitted to have a child, but if they find out that they have a genetic condition which will result in the painless death at age four of any child they have (after an ordinary life, without any suffering beyond the ordinary), and if there is no afterlife, then they are permitted to have the child, because their proxy consent does not affect the future life of a competent person. That finding out about this condition would make procreation permissible, while it would be impermissible absent the condition, seems absurd.