Disagreement Case Study – Hawk Bias

Two weeks ago I asked for posts "describing disagreement case studies;" here is my first such post.  In January I posted on Daniel Kahneman and Jonathan Renshon’s Foreign Policy article claiming all known cognitive biases favor hawks over doves.  Foreign Policy then invited me to write this letter:

Kahneman and Renshon … fail to understand that inaccurate beliefs do not necessarily equal bad outcomes. We have evolved most biases on purpose, and many give us advantages. The best salespeople really believe in their product no matter its quality, the best lawyers believe their guilty client is innocent, and the best lovers are confident of being attractive. Similarly, people often attribute successes to their own ability and character, and their failures to circumstance. Our ancestors evolved these biases because they can provide strategic advantages by instantly signaling intentions, resolve, and abilities. Perhaps nations are better off biased. Of course, every belief bias is not exactly designed for maximum advantage in every situation. But we could be too biased, or not biased enough. Until we know the direction of our errors, we certainly do not know, as Kahneman and Renshon claim, that hawks "are likely to be more persuasive than they deserve to be."

Kahneman and Renshon respond:

As Robin Hanson’s comment illustrates, evolutionary arguments are sometimes used rather naively to defend the conclusion that the mind is perfect. Evolutionary psychology contends that human cognition evolved in the Pleistocene era, when humans lived in bands of fewer than 100 members. Assume for the sake of argument that cognitive biases evolved to yield just the right level of aggressiveness toward other small bands of humans. We still would not see why Hanson is confident that the cognitive machinery that was optimal in prehistoric times is still perfect in an age of world wars, ballistic missiles, and nuclear weapons.

So I find myself disagreeing with one of the few Nobel-prize winning psychologists ever (Kahneman), co-founder of modern cognitive bias studies, on the subject of bias.  Could I be more arrogant?  Surely he considers the possibility of his own biases, and can make use of many bias indicators.

The issue here is the relative mix of error and adaption in our hawk bias.  The most charitable spin I can find is that Kahneman and Renshon think the optimal hawk bias is distributed across environments with a mean of near zero bias, and either

  1. our ancestor’s hawk bias had little relation to the optimal level for their environment, or
  2. the optimal hawk bias for our ancestors has little relation with the optimal bias for us.

In either case, our hawk bias would be mostly error, and not adaptation.   These would not be my best guesses, as our hawk bias seems too systematic, deep, and functional to be accidental, and in my theoretical analysis optimal biases tend to be more positive than negative.  But Kahneman and Renshon may know things I do not about how accidental biases can seem systematic and functional.      

I now put roughly 50% weight on this charitable spin, that they believe one of these two cases, for good reasons.  But the fact that they say I assume "cognitive machinery that was optimal in prehistoric times is still perfect in an age of world wars" when I explicitly said otherwise (though more clearly in my submitted draft) suggests they are not paying much attention here. 

They likely think me a nobody, and their less-expert co-author Renshon (BA in ’04) probably wrote their reply.  Neither has written anything I can find on evolutionary psychology; perhaps Kahneman resists this new trend in psychology that eclipses his.  And their staking out an extreme position on a politically-charged topic in a high-profile political forum weakly suggests politically-muddled thinking.   

So I put roughly 50% weight on the conclusion that our hawk bias is more adaptation than error.  Half of this weigh goes to Bryan Caplan’s suggestion that the adaption helps the individual but not the group; other group members may take hawk talk as a signal of loyalty.   The other half goes to the possibility that our hawk bias helps the group, similar to the way it may help individuals in dealing with each other.

All of these conclusions could change if Kahneman or Renshon responded further.  But a lesson here is that people feel freer to disagree with relative experts who seem to pay little attention to their arguments.   But how rational this is as response to silence?

GD Star Rating
Tagged as: , ,
Trackback URL:
  • Bruce K Britton

    Herbert Simon (Nobel Prize in Economic Sciences, 1978) regarded himself as primarily a psychologist rather than an economist (his prize was based on his research on decision making) based on a talk I heard him give, but said he enjoyed the licence to it gave him to ‘lord it over the economists.’

    Evolutionary biologists have learned to be very cautious about claims for evolutionary explanations for psychological tendencies, for several reasons.

    First, behaviors do not fossilize, nor do brains, and it is hard enough for paleontologists even to describe the evolution of bony structures, for which we have physical evidence because they do fossilize. And of course, first describing the evolution of something is a useful preliminary step before trying to explain that evolution. Explaining something for which you do not even have a description runs a risk of error.

    We may of course have a description of the present state of the psychological tendency, but reasoning backward from the present state of something to its state at various times in the past is prone to error unless we have some constraints, in the form of actual evidence about the intervening events, to guide our reasoning. So evolutionary biologists have become very reluctant to venture into explaining things for which physical evidence from the past is entirely missing, like behaviors. In this connection it is useful to remember that humans did not evolve from present day primates, although presumably humans had common ancestors with present day primates. To the extent that we have physical evidence for those common ancestors of humans, it might be possible to say something useful, but the amount of such evidence is extremely small, even for bones.

    Second, it has been embarrassing that explanations of even fossilized structures by evolutionary biologists have turned out often enough to be unmistakeably wrong when later fossilized evidence has been found, contradicting the earlier explanation. Just think how wrong we might be in the abscence of any physical evidence at all! Unfortunately, there are so many such occurrences that evolutionary biologists have become chary of such claims.

    Third, for many fossilized structures it is not only entirely possible to come up with a large variety of different stories explaining its evolution in completely different ways, but the option of making up a large number of such stories has been widely exercised. Quite often some sets of stories are mutually contradictory and so cannot all be true, so some of them must be false. Since there are commonly not enough constraints in the existing physical evidence to choose between the stories, using any particular story as an explanation is questionable.

    Just think how many stories could be made up about traits for which we have no physical evidence at all, like behaviors.

  • Putting this all aside, I find the exchange quite bizarre. Robin’s comments seem germane and rational, the response reads as if it was made to some other letter.

  • Bruce K Britton

    Sorry to be unclear, I was responding to question Robin’s use of evolutionary arguements, which Kahneman and Renshon also question, including his claim that ‘we have evolved most biases on purpose,’ his references to ‘our ancestors hawk bias.’ and his claim that evolutionary psychology is a ‘new trend in psychology that eclipses’ Kahneman’s.

    The first two claims seem ones that the experts in evolutionary biology that I have known would not buy in to, and the third one I happen to know that several prominent psychologists besides Kahneman do not subscribe to, nor do I (for what that’s worth).

    Also I was suggesting that Herbert Simon was the first psychologist to win the Nobel Prize, not Kahneman as Robin parenthetically remarks, although one could argue that some neuroscientists were even before him, Katz in 1970 for example, or Hubel and Weisel.

  • Bruce, thanks for the correction regarding other Nobel winning psychologists – I have corrected the text.

    Everyone, Bruce is an academic psychologist of Kahneman’s generation. Perhaps Kahneman shares his low regard for evolutionary psychology, helping to explain the gulf between they and Tim and I.

  • michael vassar

    Hmm. Ironically I find myself generally agreeing with Kahneman, but find the exchange to be very clearly reminiscent of several exchanges between myself and Robin only with Robin playing me. Does my agreement with Kahneman here mean that I should place greater credence in some of Robin’s positions which I disagree with?
    OTOH, I’m pretty sure I know more than Robin does about evolutionary biology and evolutionary psychology, and am much less convinced that Robin knows more than I do, and than the experts that I have consulted do, about, for instance, historical human population sizes and contemporary rates of scientific/technological output.

  • Bruce K Britton

    To Robin, Tim and Michael:

    It seems to me that this exchange might profit from some input from evolutionary biologists, who would be qualified by virtue of their knowledge of the field to evaluate the claims about evolution made by Robin and their apparent denial by Kahneman.

    Are there any such who are in this blog, or is there some way this could be submitted to such people? I know some evolutionary biologists at the University of Georgia, but since they are the same ones who influenced my original comment, they might be regarded as biased. (IThey don’t, by the way, have a low regard for evolutionary psychology in general, nor do I.)

    Does GMU have some evolutionary biologists who might be willing to comment?

  • The most conspicuous aspect of this exchange is that Kahneman and Renshon’s reply claims that Robin said that cognitive biases from primitive times would be optimal today, when he made no such statement. This does not appear to be a disagreement so much as a case of misunderstanding. In effect, the conversation goes:

    Robin Hanson: What makes you so sure the bias is wrong?
    Kahneman and Renshon: What makes you so sure the bias is perfect?

    It is an unproductive conversation but not yet a disagreement at this point.

  • Robin,

    Kahneman and Renshon probably only read the comment that was published, not the comment that you submitted. I don’t see anything in the published comment that distinguishes the ancestral environment from the modern one.

    I cannot conceive that Kahneman is resistant to evolutionary psychology. Heuristics and biases draws on natural selection as its fundamental explanation and background – and his response to you is pretty much exactly the one that I would expect a standard evolutionary psychologist to give.

    Furthermore, as nearly as I can tell, Kahneman and Renshon are in the right of this argument. Suppose – as certainly seems plausible – that some particular level of hawkishness advantaged a tribe twenty thousand years ago, and that the innate level of hawkishness was tuned in the presence of side-effect hawkishness from other cognitive biases. If other cognitive biases support hawkishness, this doesn’t mean they were tuned to support hawkishness unless you suppose that gradient descent along the bias side proceeded faster than gradient descent on actual, directly hawkish emotions. Think of a multiple regression where some variables compensate for other variables.

    Most importantly of all – from my perspective – when Kahneman and Renshon refer to a criterion of optimality by which biases are suboptimal, they are not talking about reproductive success now, let alone ten thousand years ago. They are talking about ordinary morality. If our biases lead us to be more hawkish than our moralities would seem to support, then that is, from their perspective and mine, something to be concerned about. I think, Robin, that your training in economics – where you often ask “Is the market secretly rational?” – has led you to take a poor moral attitude toward biases, adopting moral stances that would make biases appear less horrible than they should appear given our stated moral philosophies.

    But even if reproductive success were the criterion. Or even if the goal were things that correlated to reproductive success then, and are desirable under our stated moralities now. Even so, the gap between the ancestral environment and modern international diplomacy is so enormous as to render the inference from possible ancient “benefit” to possible modern “benefit” null and void in the absence of further and detailed justification. Not just the inference of maximization or optimality, which evolutionists never make, but the inference of any “beneficial” optimization whatsoever that would carry over from now to then. In this particular case, K&R’s phrasing is slightly poor, but the logic of the argument carries over very well to a similar argument for genetic benefit (never mind moral benefit) and a charitable reading should embrace this.

    IMO, Daniel Drezner has the most compelling objection, and Drezner’s objection again fails to understand K&R’s goals; while constant bias may not explain variance in peace, it may imply that we constantly get less peace than we want!

  • Michael, perhaps you will post somewhere a disagreement case study about one of your disagreements with me.

    Bruce, you say you don’t have a low regard for evolutionary psychology, but you also say we can’t believe anything without supporting physical evidence of ancestral behavior. That seems to reject almost all current evolutionary psychology work.

    Hal, it sure looks like a disagreement from a distance.

    Eliezer, I meant that our net hawkishness may be tuned, not that each component was tuned. And for nations at war, I think ordinary moral concepts align pretty well with reproductive success.

    Everyone, doesn’t much of the overconfidence in the people around you seem to benefit them in many ways? If so, why is it so strange to think that overconfident nations also benefit?

  • Eliezer raises a good question: what goal is this pro-hawkish bias supposed to advance? Is it personal reproductive success, or tribal advantage? I think most evolutionary arguments focus on individual reward rather than group selection. Hence selection would focus not on whether people’s hawkish bias was good for the tribe, but was it good for them personally.

    In that context, translating this bias into the modern world is especially problematic, because we really don’t care much at all about the reproductive success of our diplomats and political leaders. Now, if the hawkish bias were supposed to help the tribe, and if we thought of our nations as big tribes, then maybe we could see some advantages to having such a bias. But it doesn’t sound like the evolutionary story goes this way.

    Another point arises regarding bias vs strategic behavior. It has been suggested in earlier posts here that people are biased so that parts of their minds can act as a de facto “public relations department” for the person. People have to hide the truth from themselves in order to do a better job of selling the story they want to tell, hence it works better to have a bias rather than to simply be manipulative.

    But this analogy then illustrates that when dealing with modern diplomacy, such bias may not be necessary after all. Modern governments do in fact have public relations departments who can present whatever story they are told as official policy. Government officials don’t need a hawkish bias in order to sell their policies; they would do better to be unbiased and to reason strictly in strategic terms. If this requires bluffing, they can set up internal institutions so that bluffs are sold with whatever degree of human credibility is required. So it is unclear that any bias is desirable in government organizations.

  • Hal, a lot of war is about citizens’ resolve and willingness to suffer for a conflict. Government public relations departments can’t present that anyway they want, when other nations can interact with citizens more directly.

  • Everyone, the point of these case studies is not so much to rehash the disagreement to decide who was right, but to evaluate how rational each side was in disagreeing with the other.

  • Kahneman plausibly doesn’t know who you are, and didn’t see your original message, and possibly wasn’t even the one who replied to your edited message. By Aumann logic, then, your observation that Kahneman failed to adjust his opinion, need not imply too strongly that your own opinion was deficient.

    However, your “most charitable spin” is not anywhere near charitable enough, and I’m not sure you realize what seem to me the most powerful grounds for disagreeing with you – in particular, the basis on which one is to evaluate the so-called “optimality” of the bias. One should be aware, in enumerating the “most likely” members of a huge space, that when the “most likely” hypotheses don’t seem to fit the observed situation very well, that the real answer may be something else entirely. I’ve been zapped by this a few times and now endeavor to be wary. Aumann logic makes this consideration especially relevant – if your best hypotheses for explaining why a skilled rationalist would argue a certain way, don’t strike you as very good, it should be a strong cue they might know something you don’t.

    The notion that Kahneman dismisses evolutionary psychology seems like a particularly unlikely excuse for dismissing him, barring specific presented evidence of such a failure on his part.

    I agree that K&R’s position seems unusually extreme on a politically charged question, and that this is suggestive of bias on their part. But neither is their position obviously wrong to my brief scrutiny, and my protocol for disagreement says to find the object-level errors first before daring to think of metalevel argument. Part of your and my disagreement on how to handle disagreements, is that I feel the object-level argument should screen off most of the meta-level argument, if you are doing everything else correctly.

  • PS: Symmetrically, given that Kahneman doesn’t know who you are, and saw only the edited version of your comment, I don’t see any flaw in his rationality for daring to disagree with you.

  • Robin, when they write, “Hanson is confident that the cognitive machinery that was optimal in prehistoric times is still perfect in an age of world wars, ballistic missiles, and nuclear weapons”, do you see that as an accurate description of your position? If so, then I guess I misunderstood your original point. But if you don’t think it is accurate, isn’t this evidence that they failed to understand you, and so no true disagreement has occurred?

    What exactly do you see as the disagreement here; how would you summarize the issue in dispute, in a few words? Is there a sentence that you are confident that you would agree with but that they would disagree with, based on what they have written?

  • Bruce K Britton

    Signs of Irrationality in Kahneman vs Hanson Exchange.

    Robin, you are definitely right that the point of these case studies is not to rehash the disagreement, but to evaluate how rational each side was in disagreeing with the other. Along those lines:

    Kahneman disagrees with Hanson on two grounds:

    1. He claims Hanson is ‘naive’ in his use of evolutionary arguements ( implying that Hanson’s evolutionary claim does not have enough evidence to justify it) .
    Is Kahneman showing irrationality or bias here?

    It seems to me that Hanson and Kahneman could rationally disagree about whether there is ENOUGH evidence to justify a claim

    (unless the claim were necessarily true: Kahneman would be irrational if he said ” the fact that someone is married is not enough evidence to say they are not a bachelor, ” or “there is not enough evidence for the claim that ‘what will be will be’ “or ‘there is not enough evidence for the claim that ‘ the weather will either change or stay the same.'” ).

    But to say that there is not enough evidence for a claim about something’s evolution is not necessarily irrational.

    2. Kahneman claims that even if Hanson’s evolutionary claim were true, it would not necessarily lead to the conclusion that Hanson takes it to; in other words Kahneman claims that Hanson implies that something that is optimal in the past is also optimal in the present, and Kahneman says that implication is not true, that it might be that something that is optimal in the past is not optimal in the present.

    Is Kahneman being irrational here, or is he showing any of his biases? I don’t see any irrationality here, because it is rational to say Hanson implied that, and rational to say either that something that was true in a past state of society is true now too, or to say that it is not true now.

    Robin’s corresponding claims are, mutatis mutandis, also not irrational.

    That said, the disagreement may still have some irrationalities and biases, as follows:

    First, we have to realize some biases might be present but not showing: for example, myside bias (confirmation and disconfirmation bias) might be well be present in both Kahneman and Hanson, but might not show on the face of their arguements.

    Second, we see some overt signs of irrationality:

    A. Kahneman says Robin is ‘naive’ which in this academic context has the force of an insult. Hurling insults is often a sign of irrationality, but I can’t assign it to a particular bias. (Kahneman might say it was intended just as descriptive, not as an insult; but he could have referred to it with an emotionally neutral term, such as ‘mistaken’ or ‘questionable’)

    B. Robin’s submitted draft says Kahneman is from ‘the last psychology revolution and just doesn’t get the new revolution’ and ‘perhaps Kahneman resists this new trend in psychology that eclipses his.’

    These might partake of the fallacy of Appeal to Novelty (Newer is better), otherwise why say ‘…last revolution… new revolution’, and ‘…new trend. ‘
    (Again, Robin might claim descriptiveness.)

    C. Robin says Kahneman has not …’written anything … on evolutionary psychology…’ which partakes of the fallacy of Attacking the Person (Ad Hominem), because Kahneman’s arguements might still be correct whether he had or hadn’t written anything on evolutionary psychology. Evaluating the truth of his arguements is not dependent on his other writings. The following two statements are equally in-apropos: “Here’s a true claim, but its author hasn’t written anything on evolutionary psychology” or “here’s a false claim, but its author hasn’t written anything on evolutionary psychology.”

    Another possible case of Attacking the Person is ‘…their less-expert co-author Renshon (BA in ’04) probably wrote their reply.’ Evaluating the truth of Kahneman’s or Renshon’s reply can be done in the absence of information about which of them wrote it, their level of expertise, the level of their education, and the date of receiving their degree.

    D. Robin’s “They likely think me a nobody…” seems a possible sign of irrationality: Suppose they do, or suppose they don’t — how does that affect the validity of their claims?

    E. Robin also says they are ‘…staking out an extreme position on a politically charged topic…’ and suggests ‘…politically muddled thinking.’ How are these relevant to the validity of their claims? This seems another possible sign of irrationality.

    Kahneman only wrote 4 sentences and so had relatively little scope for showing his biases here.

    Note that these signs of irrationality are around the edges of the disagreement, not at its core.

    Note that both authors could scrub their text for signs of irrationality, fallacy and bias, but they could still be influencing their thinking.

  • Hal, the disagreement is over K&R’s claim that cognitive biases make hawks “more persuasive than they deserve to be.”

    Eliezer, I am not claiming that these meta-considerations are the best way to “handle” disagreements. But if we want to study these meta-considerations, concrete examples like this can help.

    Both of you, K&R’s paper makes no mention of the possibility that a hawk bias might be useful, and no other commenter on the paper I can find mentioned it either. When I do mention it, K&R do not say anything like “yes, it might be useful, but only for immoral purposes.” They instead question the relation between what was useful for our ancestors and what is useful for us now.

    So it still seems to me that Bruce’s reaction is a reasonable guide to K’s reaction – he doesn’t think much in general of evolutionary speculations not backed by hard fossil data, or he doesn’t see much relation between ancestral and modern environments.

  • Bruce, there are many considerations that you consider to be irrelevant, and call introducing them a “fallacy,” that I think are quite relevant. Eliezer calls them “meta-considerations,” and thinks them less useful, but still grants their relevance. It does matter who is naive or has written on a topic, for example.

  • “It does matter who is naive or has written on a topic, for example.” Yes indeed, but only insofar as the object-level evidence does not already settle the issue. A correct argument is just as correct no matter who makes it. Airplanes do not literally fly because credentialed scientists say they should.

    “K&R’s paper makes no mention of the possibility that a hawk bias might be useful.” Hanson, it appears to me that K&R (and myself) may not be viewing the problem according to your criterion of optimality, using a different criterion instead, which is why the question “Are we just as biased as we should be?” seems so much less plausible for us. I would not say, “Useful, but only for immoral purposes”; rather I would just say, “Promoting undesirable outcomes.”

    When I, and perhaps K&R, decide how persuasive an argument deserves to be, we are evaluating it relative to normative belief and to stated moral philosophy. To define “optimal” persuasiveness under this standard, I ask what the persuasiveness would be if we had perfectly veridical belief and unlimited self-control. Neither ancestral reproductive benefit, nor even its modernized corollary of an individual nation’s competitive interest, is defined as my criterion of optimality. What if my stated moral philosophy is personhood ethics? Then even a grand ancestral victory – the enemy reduced to bloody bones, his boy-children slaughtered and women raped – is repugnant according to my stated moral philosophy, which is what determines how persuasive the argument deserves to be. Yet the biases go on operating anyway, distorting my beliefs away from veridicality – thus the link between bias and nonoptimality is very direct, and does not pass through any sort of evolutionary or even consequentialist considerations.

    This is what I would mean when I state that “Bias X makes argument Y more persuasive than it deserves to be.” I would mean that bias X makes argument Y more persuasive than it would appear if we had perfectly veridical belief. This is not quite tautological but it is certainly a great deal more direct, as a line of reasoning, than considering who “wins” in a competition between nations. (Hedonically, usually no one wins in a war – genes may win but their vehicles will pay quite a price in unhappiness for it.) Judging by the way that K&R use the term, I suspect that their intuitive apprehension of it is substantially closer to mine than to yours. The biases cause our appreciation of the facts to depart from factuality, therefore such biases automatically stand convicted.

    Note how this squares nicely with the attitude of the existing heuristics-and-biases literature, wherein a bias stands convicted once it is shown not to be Bayesian or to lead to views incorrect under a correspondence theory of truth; long-term life outcomes are not part of the study.

  • Eliezer, I understood before your claim that K&R’s disagreement with me is due to talking about differing goals, but I didn’t see any text evidence supporting that. Perhaps others can weigh in here.

  • Bruce K Britton

    A fallacy is a type of arguement that resembles good reasoning, but that we should not find persuasive.

    In evaluating Robin’s or Kahneman’s claims and arguements, what we should find persuasive is the evidence for the claim, and the truth of the premises of the arguement and the validity of the reasoning.

    We should not find it persuasive one way or the other whether Robin or Kahneman is naive, or of the wrong generation (too old like me and Kahneman, or too young like Renshon) or of a different theoretical orientation or school of psychology or economics, or have an advanced degree or not, or have written on the subject or not, or respect us or not, or are politically to our taste or not. These are irrelevant to whether the evidence for the claim is present and convincing and whether the arguements have true premises and are validly reasoned.

    With regard to Robin’s claim that ‘we have evolved most biases on purpose,’ the evidence is not stated. If the evidence had been stated or pointed to, Kahneman possibly could have evaluated the claim based on the evidence.

    Instead Kahneman seems to be evaluating the arguement — ‘ evolutionary arguements are sometimes used rather naively to defend the conclusion that the mind is perfect.’ Kahneman specifically criticizes the validity of the reasoning — ‘Assume for the sake of arguement that,,, we still would not see why Hanson is confident that…’

    This is what it looks like to me.

    On Robin’s claim itself, as to the evidence for evolution of particular things, the best evidence is the genes themselves, as for example in recent work showing human evolution in the last 10,000 years in such genes as those governing lactose tolerance since humans started drinking milk, alcohol metabolism since humans started drinking alcohol, and in 10 or so other gene complexes. Next best are bones, tools and other artifacts, and things like cave paintings, rock carvings, marks on sticks, etc. But in my opinion none of these hold a candle to the genes themselves, now that we can read them.