Beware Heritable Beliefs

Some of the differences in our beliefs seem to be heritable.   "The Heritability of Attitutes: A Study of Twins," in Journal of Personality and Social Psychology in 2001, asked 339 twin pairs for their attitudes on 30 topics.  These attitudes had seven common factors, four of which moderate categories of beliefs:   

  • Life: voluntary euthanasia, abortion on demand, birth control, and organized religion.
  • Intellect: books, chess, education, and capitalism
  • Equality: open-door immigration, distinct gender roles, racial discrimination, and getting along with others
  • Punishment: death penalty for murder,  and castration for sex crimes

Genetic differences explained most of differences in attitudes to life and equality (66% and 55% of the variance respectively), but none (0%) of the attitudes to intellect and punishment.   

Thus your differing attitudes on  abortion, birth control, immigrants, gender roles, and race are mostly due to your genes, while your attitudes toward education, capitalism and punishment are due to your life experiences.

Is there a plausible story whereby those with genes encouraging your sort of beliefs on life and equality tend to have more accurate beliefs?   A correlation between IQ and such attitudes might be one such story, but I know of no data supporting this.

If you can’t find such a story, you should admit that the process that produced these beliefs of yours was random and uncorrelated with the truth on those subjects.   You should thus reject those beliefs as biased.   For more on the subject of rejecting arbitrary belief origins see my Theory and Decision paper "Uncommon Priors Require Origin Disputes.

GD Star Rating
loading...
Tagged as: , , ,
Trackback URL:
  • Guy Kahane

    “If you can’t find such a story, you should admit that the process that produced these beliefs of yours was random and uncorrelated with the truth on those subjects.”

    Well I’d say we need to have reason to believe that there *is* such a story, whether or not we can actually tell it.

    Of course, even if one’s attitudes to chess have no genetic origins, this hardly shows they’re not the result of random process. And many hold that it makes little sense to think of our normative beliefs (e.g. beliefs about equality) as tracking truth in any kind of relevant sense. People with such a view can deny that causal origins matter much in this area.

  • Andrew Edwards

    While my beliefs may well be more a product of extrinsic factors rather than some particularly brilliant reasoning on my part, I’m highly skeptical of the claim that genetics is the best explanation.

    The evidence from my life is that my attitudes towards immigrants, gender roles, and race are completely different from anyone else in my near family (on either side) except my brother. I’m 27 so presumably this isn’t simply rebellion.

    So what is different? My brother and I are a generation younger, which the evidence suggests does impact beliefs. We’re also more educated than most of my family (Master’s degrees for both of us, the first and second people in our family to do so). And we live in an urban area, while most of my family lives in more rural areas.

    All of those factors correlate well in social survey data to more liberal beliefs about things like gender roles and immigration. In the case of my life, in other words, those attitudes seem to be produced by environmental, rather than genetic factors.

    That explanation squares well with social survey evidence suggesting that certain attitudes have strong skews (e.g. towards the more educated, towards the more urban) that are much more congruent with a shared social experience (going to university, walking through Chinatown) than a shared genetic pool.

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Andrew, yes, some attitudes do correlate better with experiences than genes, but others correlate better with genes than experiences. The point here is to note this data on which is which.

    Guy, yes, we’d only need reason to believe there was a good story, and while genetic variation is a nice clear example of a random belief cause, it is hardly the only random cause.

    Regarding normative beliefs, beliefs that are framed as being about personal values or preferences need not track a common truth, though they should track a personal truth. And normative beliefs are usually framed as being about common moral truth, which does not depend on the person who believes.

  • http://pdf23ds.net pdf23ds

    Andrew,

    You anecdote, even if it were valid evidence, wouldn’t be strong evidence against a correlation of .66 or .55 between beliefs and genes, if the other factors involved, such as age and education, are also highly correlated (in the area of .33, say). More than one independent factor can have a strong correlation with a variable.

    Excuse me if I’m abusing the terminology.

  • Paul Gowder

    Robin: Is it really rational to reject beliefs resulting from a process (or a prior) uncorrelated with truth, in the absence of an available process or prior more correlated with truth with which to replace it?

    If you know that your opinion on, say, abortion-on-demand is probably genetic, it surely doesn’t compel you to change your opinion to the opposite view, which would also be probably genetic. It seems to suggest either that you attempt to apply a new process, better correlated with truth, to determine your opinion, or that you dump the opinion altogether.

    But suppose it’s not utility-maximizing to dump the opinion altogether? (Suppose, for example, that you want the experience of yourself as a responsible citizen who cares about public issues.) Then all that the dubious origins of your belief require you to do is look for a better origin for either your belief or the opposite. If that better origin is available, then the task is obvious: adopt it. But what does that mean? Nothing more than “hold beliefs supported by good arguments about true facts.” But that should be the task regardless of whether you believe your beliefs come from dubious processes or not: a rational agent is always open to facts and arguments that unsettle his/her beliefs.

    On the other hand, if the dubious beliefs are not replacable by better facts and arguments (such as if the abortion question depends in every argument on priors about physical ingegrity and humanity that have only intuitions rather than determinate arguments underneath), but having an opinion is superior to not having one, why not choose one at random?

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Chris (pdf23s), yes, it night not be utility maximizing to believe your best possible estimate of the truth, but that should be described as having preferences to be biased. I was careful to talk about belief variations from an average. If you have no good reason for your variation, you should just believe the average, at least if you want the most accurate beliefs possible.

  • Paul Gowder

    If there’s no reason to believe any of the processes are correlated with truth, then surely there’s no reason to believe the average is most likely to be the most accurate belief. There’s no central limit theorem for the accuracy of beliefs…

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Paul, take any estimate, add random error, it becomes less accurate. QED.

  • Paul Gowder

    Even if the estimate is wholly uncorrelated with reality? It’s not like we’re sampling from a population or talking about merely biased estimators here. If there’s no reason to believe that the average belief is related to anything other than the average genetic endowment, then isn’t it already independent and thus minimally accurate? I’m sure your probability chops are much, much, much stronger than mine (and I post this comment perfectly aware that I’m running the risk of humiliating display of ignorance), but I find it very hard to believe that the combination of two independent random events with a third random event that’s independent from the other two are better estimate of anything than one. For example, suppose you try and determine the gross national product by having a computer pick a random number between 1 and a quadrillion. That’s your first estimate. Now suppose you do that, and then randomly choose a number between 1 and 10 to multiply it by. That’s your random error. Is the second really less accurate than the first?

  • Paul Gowder

    (uh, hard to believe they’re a “worse” estimate, obviously)

  • Paul Gowder

    Uh, never mind. That last comment was just wrong — I rethought it. Sorry.

  • Paul Gowder

    Ah-ha, but I have another argument. (Yes, I’ve been thinking about this for the last two hours.) People who hold opinions that are widely varied from the mean, even without any present good reason to hold them, are providing a social benefit. They have an incentive to discover reasons and arguments in favor of their claim, increasing the general store of knowledge, and they also have an incentive to challenge the received (general) opinion, encouraging people who believe in that opinion to develop reasons and arguments in its defense. These benefits would go away if people followed your advice and consciously adopted mean opinions where they believed their original opinion had bad priors. Thus individually rational behavior would lead to a socially irrational result.

    (John Stuart Mill, of course, gets credit for this whole line of argument.)

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Paul, yes, it is possible, though hardly obvious, that bias produces social benefit. Hopefully we will discuss that issue in this forum in more detail. But avoidable error it is, make no doubt about that.

  • Paul Gowder

    Well, acting on your definition of bias as costlessly or near-costlessly avoidable error, I’m not so sure. Even if we limit ourselves to individual cost rather than social cost, it seems like there’s a rather large individual cost to the decison rule you propose.

    To highlight that, let me state it in the starkest form.

    Hanson’s Rule: If, based on aggregate evidence, it is more likely than not that the deviation from the mean of my belief on issue N is attributable to non-reasoned factors*, then I should modify my belief on N to the mean.

    But some of the things we most value in life are beliefs based on non-reasoned factors. For example, someone whose belief on the dimension of organized religion deviates from the mean because that person happens to be a Jew would suffer a massive personal cost to conforming their view to the mean. (I think you’ve really got to get the mode here. It’s hard to imagine what the mean of religion would be. The mode would be what — Hinduism?)

    *It’s easy to apply this rule in a way that creates a fallacy of accident, too. Would you demand that Richard Dawkins abandon his decidedly non-mean views about organized religion because, statistically, they’re more likely to be rooted in non-reasoned sources than the mean? Obviously not, because Dawkins’s individual views about organized religion are tracable to things like the scientific method that we do believe are correlated with truth, even if on average outliers on that dimension are not so correlated. So far so good, *we* can distinguish Dawkins… but how is Dawkins to distinguish himself from the average non-rational-outlier on your test? He can’t rely on his own belief that his views are based on rational priors, because everyone has that belief, it’s an obvious bias. Why isn’t it rational for Dawkins personally to change his beliefs?

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Paul, I didn’t intend a tipping point rule, wherein you abandon all your differences if more than half of the cause is random. Rather you should use the genetic data to help estimate the random component and then abandon that.

    Also, I was trying to distinguish a resource cost, such as time or money, from a value-induced cost of avoiding error. If you gain value from your bias, such as peer approval, I still want it to be called bias, but I also don’t want to call it bias if you just didn’t have the time to get or process the relevant info that would have reduced your error. But, yes, I’m not entirely sure that this line can be coherently drawn.

  • Andrew Edwards

    I guess to my view:

    a) This granular data seems to clash with aggregate data showing that certain populations without a shared genetic background tend to share certain sets of beliefs about the exact things identified here as “genetically caused” (i.e. urban populations tend to be far more pro-immigrant). Conflicting data is the most interesting kind.

    b) If social factors can cause beliefs (pdf23ds’s .33 correlation) even though there is a genetic component, then without further investigation it would not be reasonable to conclude that my beliefs about immigration are a product of my genetics. Further this possibility of ‘outweighing’ genetics with socialization could suggest that it is possible to come to reasonable beliefs about immigration even if your genes say otherwise – you can ‘beat’ your genes.

    c) Additionally, a belief caused by my genes could nevertheless be correct. I am highly programmed to care about the well-being of my children, for instance, which is probably also a correct behavior.

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Andrew, yes average genetic beliefs can be explained by selection, which might well tend to induce correct beliefs. The subject here is your random deviation from that genetic average. And the suggestion is to estimate and repudiate that random component of your beliefs, rather than to repudiate all components of your belief.

  • Andrew Edwards

    “The subject here is your random deviation from that genetic average. And the suggestion is to estimate and repudiate that random component of your beliefs, rather than to repudiate all components of your belief.”

    Hmm. I was thinking that the idea was that we should recognize that many of our beliefs are held for reasons other than their truth-quality. For instance, they may be held for genetic reasons, social reasons, or ego reasons. This should cause us to re-examine our beliefs and be sure that we’re testing them on truthfulness, rather than on intuition.

    In other words, I don’t know why I would necessarily accept as true a belief that was consistent with the genetic average. Even average genetics, cleaned for random variations from the norm, could cause wrong beliefs.

    As Nietzsche put it, there are certain assumptions without which our lives would be unliveable. But this does not mean that those assumptions are true. The conditions of life, in his words, could contain error.

  • John Jenkins

    asked 339 twins for their attitudes on 30 topics.

    Where does that number come from? It can’t be right (it’s odd) and the paper says 672 (336 pairs) on p. 848.

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    John, thanks for noticing; I corrected “twins” to “twin pairs.”

  • http://www.stat.columbia.edu/~cook/movabletype/archives/2006/11/making_fine_dis.html Statistical Modeling, Causal Inference, and Social Science

    Making fine distinctions in understanding hereditability of attitudes

    Tyler Cowen points to Robin Hansen who points to this paper by Olson, Vernon, Harris, and Jang, “The heritability of attitudes: a study of twins”. Robin writes, summarizing the paper, your differing attitudes on abortion, birth control, immigrants, gen…

  • michael vassar

    What beliefs are correlated with the belief in the result of this study?

    With the propensity to debias beliefs?

    What reference classes should be used?

    By the way, 0% genetic influence on attitudes towards intellect/books/chess/etc strikes me as Radically unlikely given genetic influences on IQ and on five factor “openness”, the latter of which correlates at least .3 with IQ and basically *is* attitude towards intellect/books/chess/etc.

    Lots of data suggests the religious belief is negatively g loaded.
    g loading may sometimes be difficult to define simply if relationships are non-linear, so g-type effects may be more pervasive than typically stated.

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Michael, your first two questions are interesting; I hope they will be answered someday. A net zero correlation could result from several non-zero correlated influence pathways that cancel each other.

  • http://profile.typekey.com/halfinney/ Hal Finney

    Seems like there are two questions here. The first is, is there any reason to believe that genetically-determined beliefs on a given topic are likely to be correct? In other words, if our genes largely determine our beliefs on abortion, is there a plausible mechanism whereby selection effects would tend to amplify genes which lead to “correct” beliefs in that area? It’s hard to credit that this is the case, because abortion is such a specific question and probably not one which has been relevant for survival far back into history. So we would not have very good grounds to conclude that our beliefs in this area are correct. They are more likely to be what Gould called a “spandrel”, an accidental byproduct of genes which were selected for other purposes.

    The second question arises for genes where we can in fact come up with a plausible story along these lines. Consider for example attitudes towards fire. Fire has been around for a long time, it is dangerous and powerful, so it might be plausible that there is effective selection for genes which affect attitudes and behavior towards fire. Therefore a priori one might assume that his own attitudes towards fire are “correct” even if genetically determined. However, suppose one then discovers that his own attitudes towards fire are very different from those of society as a whole. Wouldn’t it then be reasonable to conclude that it is more likely that society’s attitudes are the correct ones, and that ones own attitudes are due to some genetic mistake or malfunction, since members of society have been subject to the same genetic selection pressures but are far more numerous than onesself?

    Therefore it seems that whether or not one can come up with a mechanism for how a genetically based belief can be reasonably attributed to selection pressure and presumed to be correct, either way you can’t justify ignoring the social consensus in favor of your genetically-determined belief.

  • http://www.halfsigma.com/2006/11/why_we_believe_.html Half Sigma

    Why we believe what we believe

    I found a comment on someone else’s blog which I thought was pretty interesting. The post was about how a study of identical twins showed that some political and social attitudes had a high degree of hereditability. The commenter, who

  • http://www.intheagora.com/archives/2006/12/quick_links_3.html In the Agora

    Quick Links

    Here are some pieces so interesting I’m at a loss for further commentary:Insurers are starting to account for the costs of global climate change.The two-dollar bill is making a big comeback.Barak Obama is the real deal. Click and listen why.Some…

  • Pingback: Quick Links - In The Agora