I wrote in a recent blog about sacred values, which seem to have puzzling features. As noted, Scott Atran’s 2010 book Talking to the Enemy chronicles his anthropological work among Muslim terrorists in Indonesia. In one study he discusses, he gave his informants this problem, for which each informant would have to choose between two of the three options:

1. Complete a successful, effective suicide bombing attack.
2. Complete a successful, effective non-suicide bombing attack.
3. Complete a holy, once-in-a-lifetime pilgrimage to Mecca. 

The puzzle arose from the fact that informants generally chose 2 over 1 (non-suicide bombing is better than suicide bombing); they chose 3 over 2 (pilgrimage to Mecca is better than non-suicide bombing); but then they chose 1 over 3 (suicide bombing is better than a pilgrimage to Mecca). 

But both intuitively and according to standard axioms in formal theories of choice, we would expect our preferences to be transitive, meaning: If 3 > 2 and 2 > 1, then 3 > 1.

So how should we interpret the fact that Atran’s data indicate that some people seem to violate transitivity, when it comes to sacred values?

Atran himself seems to think that lack of transitivity is just a fact about sacred values. That might be a reasonable view. Lots of relations, after all, aren’t transitive. The blood relative relation, for example, isn’t transitive: if A is a blood relative of B and B is a blood relative of C, that doesn’t entail that A is a blood relative of C. So it could be that the comparative preferences that exist among sacred values just don’t have the orderly structure that transitivity would imply. 

But I myself find that view unsatisfying, since it is not clear to me what the benefits (evolutionary, individual, or cultural) would be of having non-transitive sacred preferences. 

One classic argument for why preferences are (or at least should be) transitive is known as the “money pump” argument, which is a reductio ad absurdum. It starts by making the assumption that some preferences could be intransitive, and it derives an absurd consequence from that assumption. It goes like this. Suppose you prefer an apple to a banana, the banana to an orange, but the orange to the apple [thus violating transitivity]. If that were the case, I could pump money from you in the following way: if you started with the apple and I started with the orange and the banana, you would give me the apple plus a penny for the orange, since you prefer the orange to the apple. But then I could sell you the banana for the orange plus a penny, since you prefer the banana to the orange. After that, however, I could get you to give me the banana plus a penny in exchange for the apple, since you prefer the apple to the banana. But then we’re back to where we started! So if you had intransitive preferences about fruit (or whatever), that would allow me to pump money from you gradually but indefinitely. But that absurd scenario (or even something similar to it) doesn’t happen, which is good reason to think that preferences in general and for the most part are transitive. 

The money pump argument of course can’t apply specifically to the sacred choices from Atran’s study, since some of those choice only could ever be made once. Furthermore, other studies on sacred values have shown them to be insensitive to “deal sweeteners” (e.g., offering extra money or whatever doesn’t get people to budge from their sacred values). So the money pump argument is nowhere close to being a compelling argument against the idea that sacred values are intransitive. Still, it does highlight that there is something at least strange about the idea of genuinely intransitive preferences, which in turn puts the burden on someone to say why there would be a preference system that violated transitivity. (Why, for example, would such a system have been supported by natural selection—either in biological or cultural evolution?) And after much trying, I still have a hard time seeing how that burden can be convincingly met. 

So here’s what I think is going on instead.

When it comes to making sense of the many psychological motivations that explain human choice, we should posit distinct psychological systems, one of which encodes ordinary, day-to-day utilitarian preferences—and the other of which encodes sacred values. So far, just about everyone who pays attention to sacred values will agree on positing these two distinct systems (or at least something along those lines). My former student Adrian Pecotic, for example, also discusses the positing of two systems in his excellent MA thesis.

But what Atran doesn’t consider is the possibility that the survey responses he collected (that seem to show intransitivity) actually are evidence of his informants’ oscillation between the two value systems. That is, for some of the choice problems he posed, his informants are relying on the sacred values system, but for others, they are relying on the ordinary utilitarian values system. 

Recall the Daniel Kahneman line I quoted in Part I: “People don’t choose between things. They choose between descriptions of things.” If this is right, then in some cases people could be describing a given choice in their minds in a way that activates one system or the other, and the same choice might be described internally in two different ways. 

Let’s focus on option 1, the suicide attack. 

That could be described internally as either 1u. or 1s.:

1u. Tactically effective attack in which one of our own soldiers is lost.

1s. Martyrdom. 

Importantly, 1u. is a description that will likely activate the ordinary utilitarian values system—the one that is responsive to numbers, quantities, and incremental costs and benefits. 1s., of course, is a description that will activate the sacred values system. 

What I think is going on, then, is that the two choice problems involving option 1 (1 versus 2, 1 versus 3) tacitly activate different internal descriptions of 1. 

When 1 is being compared to 2 (effective non-suicide attack), something like the 1u. description is activated in the mind of the person making the choice. So we have the following comparison as the first choice.

1u. Tactically effective attack in which one of our own soldiers is lost.

2u. Tactically effective attack in which none of our own soldiers is lost.

These in turn are evaluated not by the sacred value system but by the utilitarian system, which easily evaluates 2u. as better, since no one from the home team dies in 2u. But when 1 is being compared to 3, which will reliably be conceived in sacred terms, the choice will now be between these two descriptions.

1s. Martyrdom.

3. Pilgrimage. 

With this comparison, the sacred values system values 1s. more highly. 

With this framework in place, it’s also easy to explain why 3 > 2 appears in the data. The sacralizing description of 2 would be something like this.

2s. Jihad advances. 

So it appears to me that the sacred value systems of Atran’s informants may be transitive (read over descriptions) after all. They value in this order:

1s. (Martyrdom) > 3. (Pilgrimage) > 2s. (Jihad advances) 

So on my view, the violation of transitivity among sacred values is only apparent—a product of the facts (i) that the utilitarian values system can also be recruited to evaluate certain choices and (ii) that, when the utilitarian values system is activated, 2u. > 1u. 

That might all seem technically interesting. But you might still be wondering: what’s the take-away message? 

The deeper message, it seems to me, is not just that human psychological systems can value the same thing differently under different descriptions (that is a deep point, but it’s been known in psychology for a long time). It’s rather that in one mind, two distinct value systems, each with their own coherent rules (like transitivity), cover overlapping domains, such that the same event (like a suicide bombing) can be assigned different priorities by the different systems, in relation to other events.

It is as if a priest and an economist live in the minds of each of us (including priests and economists), each with their own values, preferences, and ways of describing events. Given that, it is no wonder that we often find ourselves conflicted about matters of deep importance. It is also no wonder that hypocrisy is so common. That, however, is a story for a different blog.