Intransitivity Is Invisible

Because human perception is unreliable, it simply cannot give us much evidence about very rare events. If roughly one in a million people looking at clouds think they see an alien spaceship, then even if one in a billion people looking at clouds actually do see an alien spaceship, we’ll just never know that by listening to sky-watching testimony. (At least not without enough data to distinguish a 0.1% effect on reporting rates.)

If there is such a thing as a human ability to perceive moral truth, independent of person, culture, moment, and mental context, it is a very noisy ability. Beliefs about moral truth vary greatly across cultures, who you ask, when you ask them, and how you ask the question. This high level of noise limits our ability to discern fine and unusual moral detail.

Larry Temkin has a new book arguing that moral intuitions are commonly intransitive, forcing us to reject either transitivity or some strong intuitions. Tyler summarizes:

Temkin is skeptical of transitivity. … The main contribution of this book is to show you that the transitivity postulate is far less intuitively appealing than it seems at first.

Tyler’s position:

I see the good is more holistic than additive-aggregative. … For many individualized normative comparisons there simply isn’t a right answer. I view “ranking” as a luxury, occasionally available, rather than an axiomatic postulate which can be used to generate normative comparisons, and thus normative paradoxes, at will. I see that response as different than allowing or embracing intransitivity across multiple alternatives and in that regard my final position differs from Temkin’s.

Whatever that means.

It seems to me that even if we accept that moral truth could be intransitive, the fact that specific moral intuitions are typically transitive forces our best guess moral beliefs to also be almost always transitive. Let me explain.

Even if moral truth is in fact transitive, the huge error rates in our moral intuitions would still produce a high rate of intransitivity in our moral intuitions. A rate similar in fact to the rate we see. So we basically have little evidence that moral truth is intransitive.

Furthermore, the strong overall transitive tendency in our intuitions implies that there is at least a strong overall transitive tendency in moral truth. So in any specific case where moral intuitions seem intransitive, our best explanation for that intransitivity is moral error, even if in fact moral truth does have a lot of intransitivity. Our moral vision is just not clear enough to discern real moral intransitivity’s from among error-induced apparent ones. We should thus continue on as before with our transitivity-based policy analyses.

Added 6a: Here’s what I mean by “the strong overall transitive tendency in our intuitions.” When we pick random options A,B,C, (e.g., charities, foods, movies) and ask random people to rank A v B, B v C, A v C, and in each case ask people for a probability of a mistaken ranking, the fraction of intransitivity cases would be overall well explained by a model of mostly transitive preferences plus error. The error probability would fit the probability of ranking errors people give, once we allow for error in the probability numbers and make a usual correction for overconfidence.

GD Star Rating
Tagged as:
Trackback URL:
  • Jason

    Let me pose a question: what evolutionary benefit is derived from selecting for moral transitivity (or preference transitivity in general)?

    I’m going to jump out on a limb and say none; therefore nothing will sharpen our moral intuition error to the point of satisfying transitivity.

    If I prefer apples to oranges and oranges to pears, what benefit do I derive from ensuring I prefer apples to pears in the absence of oranges? If I prefer apples, I am happy and consistent with apples, but if I prefer pears I am happy with pears.

    • One benefit is that despite never having made a direct comparison between apples and pears you can nevertheless know which you prefer (and presumptively more beneficial).

    • Kevin Dick

      For preference transitivity, the benefit is that you aren’t a “money pump”. If you prefer pears and I can manage to dig up an orange, I’ve got you at a disadvantage. Given the social and trading nature of humans, this seems like a real potential concern.

      • Jason

        Ah, but then, given humans are social, why then would we not have well-defined preference transitivity if we could be taken advantage of like this? Shouldn’t someone have come up with a way to exploit the lack of transitivity, and thus select for transitivity?

        There is an old psychology experiment where the experimenter does something similar to this and the subject keeps trading back and forth for the same two objects.

  • matt

    Well said. The arguments in much of moral philosophy seem to rely heavily on long narrative setups, framing, and ambiguous terminology. Of course using these rhetorical techniques you can come up with clever scenarios where transitivity “feels wrong” or even “repugnant.” But this doesn’t do much to remove the very real possibility that our intuitions may often be incorrect or manipulable. Especially in light of the obvious mathematical/logical validity of transitivity.

  • Kevin Dick

    This may be one of your most awesome posts of all time. I knew there was something wrong with Temkin’s reasoning and contrived philosophy counter-examples have always bothered me. But I could never articulate my objection. Until now.

  • You may have to break down and actually read the book in question (which I haven’t read) or Katz’s book, which makes the argument Cowen would prefer Temkin to have made, as your argument begs the question. I doubt anyone would disagree with you that *if* nontransitivity were rare–perhaps even if transitivity were the rule–*then* the evidence for _some_ nontransitivity could be explained away. This trivializes the question because these nontransitivists maintain that nontransitivity is the *rule*. Katz, for example, argues that the law is thoroughly infected with nontransitivity.

  • Michael Wengler
    • Robert Koslover

      If I prefer Sarah to Veronika and Veronika to Jenna must I, should I, ought I prefer Sarah to Jenna?

      Well yes, I think you should, if you are using the same criteria for establishing your preferences when doing each set of comparisons.

      • Cyrus

        Arrow’s Theorem often applies within my own skull.

  • Pingback: Assorted links — Marginal Revolution()

  • arch1

    “…the fraction of intransitivity cases would be overall well explained by a model of mostly transitive preferences plus [self reported probability of] error.”

    0) Robin, you seem to define error as (reported, intuitive morality) – (moral truth), where the latter – assuming it even exists – is independent of culture, individual, time, & context; and the former depends on all of these things. Is that right, and if so how could people possibly appraise it meaningfully?

    1) Even if I’m wrong, and you’re using “error” to mean something that people could meaningfully estimate – why do you assume that they would? My experience from preschool on is that everyone but me tends to grossly underestimate anything suggestive of their own fallibility, and that the model will instead be

    intuitive intransitivity = reported error + (unreported error + moral_truth_intransitivity),

    which only places a (probably loose) upper bound on the final term (again, assuming that moral truth even exists in the first place).

  • CaptBackslap

    I always enjoy when people go through all sorts of contortions to avoid facing the fact that “moral intuition” is roughly as old as the atlatl, and about as applicable to modern life.

  • Philo

    On a minor point: In assessing disagreement, we must be careful about what counts as “asking the same question.” What is verbally the same question–say, “Is abortion usually wrong?”–posed to a 21st-century American and to a 15th-century Inca would really be two different questions, because of implicit indexicality. The two questions would be something like: “Is abortion usually wrong in the circumstances of 21st-century America?” and “Is abortion usually wrong in the circumstances of the 15th-century Inca empire?” One might get opposite answers with no real disagreement, since *really* they were answers to two *different* questions.

  • Douglas Knight

    The fairly concrete conclusions that Tyler draws could equally well be drawn from humility about error.

  • Pensans

    On this blog, the fact that people report different moral beliefs is taken as strong evidence that they have different moral beliefs?

    More naive thab I thought.

  • mjgeddes

    All moral confusion can be resolved using my triple-aspect ontology. if you make a clear division between the platonic (timeless ideals) level, the system (agent) level and the object level (cultural artifacts) this resolves the confusions.

    PLA Platonic timeless ideals
    SYS Systems, goal directed agents, extrapolated volition
    OBJ Cultural artifacts, memes

    The bits of the morality that are on the object level are totally human created (memes are creations of culture). On the system level, morals there are partially objective but not universal, it’s the CEV based on the agents cognitive architecture (evo psyche in the case of humans)

    But the parts of morality on the platonic level are in fact universal and timeless. I see them very clearly now, such that no error remains Robin. They are based on aesthetics and the creation of beauty, as I have clearly stated on this blog multiple times. Aesthetics is the ultimate grounding for morality and it is universal and timeless, as I always claimed. Now and forever.

    The fact that many readers lack the wit and wherewithal to perceive the clear truth of my revealed moral intuitions is no fault of mine 😉

    • Preferred Anonymous

      Classic. I’m in agreement.

      What you say is just common sense. Though I strongly object to claiming it as your own.

  • Any model that assumes actual moral truths is hopelessly flawed. It’s not as though our moral intuitions come from interpreting a noisy radio signal beamed from a Morality Quasar. Morality is just stuff the brain does. We can ask “what would person x consider good”, but if person x has circular preferences, we can’t say “well that’s probably just noise” because there’s nothing extrinsic for it to be a noisy representation of.

    • Anonymous

      Not sure. If the person cares about the quality of experiences of other beings, then having circular preferences in wishing experiences on other beings may well be an inconsistent representation of the quality of these experiences. Of course, there is no objective moral truth “you must care about anyone’s experiences”, except for your own maybe, while they’re immediately present.

    • Awesome comment.

    • Dave

      Immortality is made easier by buffering and noise and,don’t forget,necessity. It is logical to turn up the noise, so as to get a good nights sleep.

      Take for example the buffering caused by distance . Slaves produced sugar so Englishmen could sweeten their tea. Also necessity,- Slave drivers,expelled to penal colonies would have starved on miserable islands,unless they could trade with the motherland.

      First order imperatives,like eating are not considered by armchair philosophers

    • mjgeddes

      Godel, Godel, Godel.
      Lob, Lob, Lob

      Anti-platonists think that Godel/Lob shows that there’s no universal math truths, that math is just human inventions. In fact Godel/Lob implies the total polar opposite of this. The fact that there are *true* math statements not captured within any given formal system x , shows that math is more than any given formal system x … ergo, it’s platonically real. Analogous arguments hold for morality. There are no ‘circular preferences’.

      #awesomeness, 1. An unmeasurable amount of awesomenimity something can produce. 2. Something that qualifies as awesome 😉

  • Beliefs about moral truth vary greatly across cultures, who you ask, when you ask them, and how you ask the question.

    Which makes me think that we should a) consider how our own moral truths are or aren’t culturally dependent and b) read essays like Paul Graham’s “What You Can’t Say,” which is tangential to your post but still worth pointing you to.

  • When you say that intransitivity rates are about what you’d expect given error rates and a correction for overconfidence, are you referring to a particular study or studies that shows this? I believe the plausibility of that statement but would really like to track the citation if it exists.

  • Pingback: 7 Words I Learnt This Month | Lynley Stace()