Predictably Lacking

Due to a particularly engaging high school teacher, my undergraduate minor was in economics. Upon taking a number of classes in economics at the college level, I realized that most of the assumptions made by economists about how people should be expected to behave were about as useful for understanding human behavior as most of my undergraduate psychology classes; that is to say not very. It was through Dan Ariely’s books that I was initially exposed to behavioral economics; a field which seemed to take a stand against the nonsensical assumptions of traditional economics. Happy as I was to see that first step, my enthusiasm was dampened somewhat by the fact that behavioral economics was not evolutionary economics. Economists, behavioral or otherwise, were still dealing with the human mind, and they lacked a good theory for understanding how and why the mind works. On a related note, I just finished Dan Ariely’s latest offering, The (Honest) Truth About Dishonesty: How We Lie to Everyone – Especially Ourselves. (2012).

There’s only room enough in my life for one book with parentheses in the title, and this is that book.

Due to a miscommunication with Amazon, I actually ended up getting my copy of this book for free, and, beyond simply saving money, I’m quite happy I did for a simple reason: I don’t think Dan’s new book is really worth spending the money on, (the book jacket suggested a retail of $27) especially if you’ve already read his first two offerings. In the (ostensibly selfless) interests of saving others time and money, here’s the main finding of the research presented in the book: given the opportunity to cheat, most people will cheat to some (relatively small) degree, with very few will going all out and cheating as much as possible. Of course, the precise degree to which people cheat is flexible, and various contexts make it more or less likely that people will cheat. This might suggest that there are certain parts of the mind monitoring various environmental cues in an attempt to determine when cheating would be profitable, and to what extent one should cheat.

The research that Dan reviews cuts against what he calls the “Simple Model of Rational Crime”, in which people consciously think through the costs and benefits when it comes to deciding whether or not to commit a crime (or act immorally, more generally). Standard economic assumptions don’t seem to pan out well, and anyone familiar with Dan’s previous work will already know that. Unfortunately, Ariely replaces that simple model with his own – arguably simpler – model that goes like this:

In a nutshell, the central thesis is that our behavior is driven by two opposing motivations. On the one hand, we want to view ourselves as honest, honorable people. We want to be able to look at ourselves in the mirror and feel good about ourselves (psychologists call this ego motivation). On the other hand, we want to benefit from cheating as get as much money as possible. (p.27)

Right from the start, Ariely’s central thesis is deeply flawed. As others have pointed out (Kurzban, 2010), “feeling good” about ourselves is not a plausible function for any part of our psychology. Evolution is (metaphorically) blind to what organisms feel; it can only see what organisms do. An organism that feels terrible but does useful things would win out against an organism that feels great but doesn’t do useful things every single time. A quick example should demonstrate why. Let’s say feeling good is actually important, in and of itself. There are two organisms presented with potential benefit from cheating: the first organism cheats, but only cheats a little bit in order to maintain its positive sense that it’s an honest individual; after all, it didn’t cheat that much, and it wasn’t doing any real harm, so it’s probably still a morally upstanding creature. The second organism cheats as much as it can and feels pretty good about its cheating; it doesn’t try to feel good by justifying its behavior, it just feels good about what it does generally.

You know what else feels good? Getting a perfect score.

That example should make the problem with Ariely’s central thesis stand out in stark relief: why should an organism care about seeing itself as a morally upstanding creature, and why should seeing itself as such hinge on its perception of its own integrity? By focusing on a conscience-centric model without making the function of such a perspective clear, Ariely misses the mark. As DeScioli & Kurzban (2009) suggest, we cannot understand the function of conscience without first examining condemnation. In a world where others judge our actions, and those judgments cause those others to behave in certain ways towards us, conscience can serve as a defense mechanism. Rather than risking costly punishment and social sanction from behaving in a manner others perceive as immoral, potentially detrimental actions can be avoided in the first place.

Now one might counter that, in the experiments Ariely reports on, there was no risk of subjects being caught or punished, and further that the subjects knew this; since any fear of punishment should have been, effectively removed, concerns for condemnation can’t explain these results. However, to do so would be to make the basic error of failing to understand the difference between adapted and adaptive. Just because someone might consciously report that they understand there was no real risk, it doesn’t mean other modules in their brain came to the same conclusion.

One final point I’d like to touch on is the chapter concerning self-control. Not to rely too heavily on Kurzban (2010) here, but self-control is not like a muscle, and thinking of it as such leads one to an incorrect model of the mind. (For references, see here, here, and here). Since an incorrect model of the mind seems to be the central thesis of the book, it’s at least consistent in that regard. There are, no doubt, some interesting things to be learned from the research in Dan’s book. However, you’ll need to figure them out, more or less, on your own.

References: Ariely, D. (2012). The honest truth about dishonesty: How we lie to everyone else – especially ourselves. New York, NY: HarperCollins

DeScioli P, & Kurzban R (2009). Mysteries of morality. Cognition, 112 (2), 281-99 PMID: 19505683

Kurzban, R. (2010). Why everyone (else) is a hypocrite: Evolution and the modular mind. Princeton, NJ: Princeton University Press.

Assumed Guilty Until Proven Innocent

Motley Crue is a band that’s famous for a lot of reasons, their music least of all. Given their reputation, it was a little strange to see them doing what celebrities do best: selling out by endorsing Kia. At least I assume they were selling out. When I first saw the commercial, I doubted that Motley Crue just happened to really love Kia cars and had offered to appear in one of their commercials, letting it feature one of their many songs about overdosing. No; instead, my immediate reaction to the commercial was that Motley Crue probably didn’t care one way or another when it came to Kia, but since the company likely ponied up a boat-load of cash, Motley Crue agreed to, essentially, make a fake and implicit recommendation on the car company’s behalf. (Like Wayne’s World, but without the irony)

What’s curious about that reaction is that I have no way of knowing whether or not it’s correct; I’ve never talked to any of the band members personally, and I have no idea what the terms of that commercial were. Despite this, I feel, quite strongly, that my instincts on the matter were accurate. More curious still, seeing the commercial actually lowered my opinion of the band. I’m going to say a little more about what I think this reaction reflects later, but first I’d like to review a study with some very interesting results (and the usual set of theoretical shortcomings).

I’m not being paid to say it’s interesting, but I’ll scratch that last bit if the price is right.

The paper, by Inbar et al (2012), examined the question of whether intentionality and causality are necessary components when it comes to attributions of blameworthiness. As it turns out, people appear quite willing to (partially) blame others for outcomes that they had no control over – in this case, natural disasters – so long as said others might only have desired it to happen.

In the first of four experiments, the subjects in one condition read about how a man at a large financial firm was investing in “catastrophe bonds”, which would be worth a good deal of money if an earthquake struck a third world country within a two year period. Alternatively, they read about man investing in the same product, except this time the investment would pay out if an earthquake didn’t hit the country. In both cases, the investment ends up paying off. When subjects were asked about how morally wrong such actions are, and how morally blameworthy the investor was, the investor was rated as being more morally wrong and blameworthy in the condition where he benefited from harm, controlling for how much the subjects liked him personally.

The second experiment expanded on this idea. This time, the researchers varied the outcome of the investment: now, the investments didn’t always work out in the investor’s favor. Some of the people who were betting on the bad outcome actually didn’t profit because the good outcome obtained, and vice versa. The question being asked here was whether or not these judgments of moral wrongness and blameworthiness were contingent on profiting from a bad outcome or just being in the position to potentially benefit. As it turns out, actually benefiting wasn’t required: the results showed that the investor simply desiring the harmful outcome (that one didn’t cause, directly or indirectly) was enough to trigger these moral judgments. This pattern of results neatly mirrors judgments of harm – where attempted but failed harm is rated as being just about as bad as the completed and intended variety.

The third experiment sought to examine whether the benefits being contingent on harm – and harm specifically – mattered. In this case, an investor takes out that same catastrophe bond, but there are other investments in place, such that the firm will make the same amount of money whether or not there’s a natural disaster. In other words, now the investor has no specific reason to desire the natural disaster. In this case, subjects now felt the investor wasn’t morally in the wrong or blameworthy. So long as the investor wasn’t seen as wanting the negative outcome specifically, subjects didn’t seem to care about his doing the same thing. It just wasn’t wrong anymore.

“I’ve got some good news and some bad news…no, wait; that bad news is for you. I’m still rich.”

The final experiment in this study looked at whether or not selling that catastrophe bonds off would be morally exculpatory. As it turned out, it was: while the people who bought the bonds in the first place were not judged as nice people, subsequently selling the bonds the next day to pay off an unexpected offense reduced their blameworthiness. It was only when someone was currently in a position to benefit from harm that they were seen as more morally blameworthy.

So how might we put this pattern of results into a functional context?. Inbar et al (2012) note that moral judgments typically involve making a judgment about an actor’s character (or personality, if you prefer). While they don’t spell it out, what I think they’re referring to is the fact that people have to overcome an adaptive hurdle when engaging socially with others: they need to figure out which people in their social world to invest their scarce resources in. In order to successfully deal with this issue, one needs to make some (at least semi-accurate) predictions concerning the likely future behavior of others. If one sends the message that their interests are not your interests – such as by their profiting if you lose – there’s probably a good chance that they aren’t going to benefit you in the long term, at least relative to someone who sends the opposite signal.

However, one other aspect that Inbar et al (2012) don’t deal with brings us back to my feelings about Motley Crue. When deciding whether or not to blame someone, the decision needs to be made, typically, in the absence of absolute certainty regarding guilt. In my case, I made a judgment based on zero information, other than my background assumptions about the likely motives of celebrities and advertisers: I judged the band’s message as disingenuous, suggesting they would happily alter their allegiances if the price was right; they were fair-weather friends, who aren’t the best investments. In another case, let’s say that a dead body turns up, and they’ve clearly been murdered. The only witness to this murder was the killer, and whoever it is doesn’t feel like admitting it. When it comes time for the friends of the deceased to start making accusations, who’s going to seem like a better candidate: a stranger, or the burly fellow the dead person was fighting with recently? Those who desired to harm others tended to, historically, have the ability to translate those desires into actions, and, as such, make good candidates for blame.

“I really just don’t see how he could have been responsible for the recent claw attacks”

Now in the current study there was no way the actor in question could have been the cause of the natural disaster, but our psychology is, most likely, not built for dealing with abstract cases like that. While subjects may report that, no, that man was not directly responsible, some modules that are looking for possible candidates to blame are still hard at work behind the scenes, checking for those malicious desires; considering who would benefit from the act, and so on (“It just so happened that I gained substantially from your loss, which I was hoping for,” doesn’t make the most convincing tale). In much the same way, pornography can still arouse people, even though the porn offers no reliable increase in fitness and “the person” “knows” that. What I feel this study is examining, albeit not explicitly, are the input conditions for certain modules that deal in the uncertain and fuzzy domain of morality.

(As an aside, I can’t help but wonder whether the people in the stories – investment firms and third world countries – helped the researchers find the results they were looking for. It seems likely that some modules dealing with determining plausible perpetrators might tap into some variable like relative power or status in their calcuations, but that’s a post for another day.)

References: Inbar, Y., Pizarro, D., & Cushman, F. (2012). Benefiting From Misfortune: When Harmless Actions Are Judged to Be Morally Blameworthy Personality and Social Psychology Bulletin, 38 (1), 52-62 DOI: 10.1177/0146167211430232

Intentional Or Not, Incest Is Still Gross (And Wrong)

For a moment, let’s try to imagine a world that isn’t our own. In this world, the intentions behind an act are completely disregarded when it comes to judging that act morally; the only thing that matters is the outcome. In this world, a man who trips and falls down the stairs, accidentally hitting another man on the way down, is just as wrong as the man who walks up to another and angrily punches him right in the face. In another case, a sniper tries to assassinate the president of the country, but since he misses by an inch no one seems to care.

Such a world would be a strange place to us, yet our sense of disgust seems to resemble the psychology of that world to some degree. While intent doesn’t stop mattering altogether when it comes to disgust, it would seem to matter in a different way than is typically envisioned when it comes to the domain of physical harm.

Sure, it may look disgusting – morally or otherwise – but who doesn’t love Red Velvet?

A recent paper by Young & Saxe (2011) set out to examine the role that intentions placed in the contexts of a more physical harm – poisoning – relative to their role in contexts that elicited disgust – the ever popular case of sibling incest. Subjects read stories in which incest was committed or a friend served another friend peanuts despite knowing about their friend’s peanut allergy; for these stories there was a bad intent and a bad outcome. When both acts were committed intentionally, harm tended to be rated as slightly more morally wrong than incest (6.68 vs 6.03, out of 7). However, the story changed when both acts were committed by accident – when there was still a bad outcome, but only neutral intentions. While the harm condition was now rated as not very wrong, the incest condition was still rated as fairly wrong (2.05 vs 4.24, out of 7).

Another study basically replicated the results of the first, but with one addition: there was now an attempt condition in which an actor intends to commit an act (either harm someone or commit incest), but fails to do so. While the intentional condition (bad intent and bad outcome) was rated as the worst for both incest and harm, and the accidental condition (neutral intent and bad outcome) saw incest rated as worse than harm, the attempt condition showed a different pattern of results: while attempted harm was rated to be just as bad as intentional harm (6.0 and 6.5, respectively), attempted incest was rated more leniently than intentional incest (4.2 and 6.4). In other words, moral judgments of incest were more outcome dependent, relative to moral judgments of harm.

One final study on the topic looked at two different kinds of failed attempts concerning incest and harm: the ‘true belief but failed act’ and the ‘false belief but completed act’. The former involved (in the case of incest) two siblings correctly believe they’re siblings and attempt to engage in intercourse but are interrupted before they complete the act. The latter involved two people who incorrectly believe they’re siblings and actually engage in intercourse. The harm contexts were again outcome independent: whether the harm was completed or not didn’t matter. However, the incest contexts told a different story: the ‘true belief but failed act’ condition  was rated as being more immoral than the ‘false belief but completed act’ condition (5.65 vs 4.2). This means subjects were likely rating the act relative to how close it approximated actual incest, and the subjects apparently felt an unconsummated attempt at real incest looked more like incest than a consummated act where the two were just mistaken about their being siblings.

And I think we can all relate to that kind of disappointment…

A further two studies in the paper sought to examine two potential ways to account for this effect. In one case, subjects rated the two stories with respect to how emotionally upsetting they were, how much control over the situation and knowledge of the situation the actors had, and the extent to which the agents were acting intentionally. In no case were there any significant differences, whether concerning disgust or harm, or whether the act was intentional or accidental. The subjects seemed to be assessing the two stories in the same fashion. The second study sought to examine whether subjects were using moral judgments to express the disgust they felt about the story, rather than actually judging the act to be immoral. However, while subjects rated intentional incest as worse than accidental incest, they rated both to be equally as disgusting. Accordingly, it seems unlikely that people were simply using the morality scale as a proxy for their disgust.

It is my great displeasure to have to make this criticism of a paper again, but here goes: while the results are interesting,Young & Saxe (2012) really could have used some theory in this paper. Here’s their stated rationale for the current research:

Our hypothesis was initially motivated by an observation: in at least some cultures, people who commit purity violations accidentally or unknowingly are nevertheless considered impure and immoral.

Observing something is all well and good, but to research it, one should – in my opinion – have a better reason for doing so than just a hunch you’ll see an effect. The closest the authors come to a reasonable explanation of their findings – rather than just a restatement of them – is found in the discussion section, and it takes the form of a single sentence, again feeling like an afterthought, rather than a guiding principle:

…[R]ules against incest and taboo foods may have developed as a means for individuals to protect themselves, for their own good, from possible contamination.

Unfortunately, none of their research really speaks to that possibility. I’d like to quickly expand on that hypothesis, and then talk about a possible study that could have been done to examine it.

Finding an act disgusting is a legitimate reason to not engage it yourself. While that would explain why someone might not want to have sex with their parents or siblings, it would not explain why one would judge others as morally wrong for doing so. For instance, I might not feel inclined to eat insects, but I wouldn’t want someone else punished because they enjoyed doing so. However, within the realm of disgust, the threat of contamination looms large, and pathogens aren’t terribly picky about who they infect. If someone else does something that leads to their becoming infected, they are now a potential infection risk to anyone they interact with (depending on how the pathogen spreads). Accordingly, it’s often not enough to simply avoid engaging in a behavior yourself; one needs to avoid interacting with other infected agents as well. One way to successfully deter people from interacting with you just happens to be aggressive behavior. This might, to some extent, explain the link between disgust and moral judgments. It would also help explain the result that disgust judgments are outcome dependent: even if you didn’t intend to become infected with a pathogen, once you are infected you pose the same risk as someone who was infected more intentionally. So how might we go about testing such an idea?

One quick trip to the bookstore later…

While you can’t exactly assign people to a ‘commit incest’ condition, you could have confederates that do other potentially disgusting things, either intentionally or accidentally, or attempt to do them, but fail (in both cases of the false or true beliefs). Once the confederate does something ostensibly disgusting, you assign them a partner in one of two conditions: interacting at a distance, or interacting in close proximity. After all, if avoiding contamination is the goal, physical distance should have a similar effect, regardless of how it’s achieved. From there, you could compare the willingness of subjects to cooperate or punish the confederate, and check the effect of proximity on behavior. Presumably, if this account is correct, you’d expect people to behave less cooperatively and more selfishly when the confederate had successfully done something disgusting, but this effect would be somewhat moderated by physical distance: the closer the target of disgust is, the more aggressive we’d expect subjects to be.

One final point: the typical reaction to incest – that it’s morally wrong – is likely a byproduct of the disgust system, in this account. Incestuous acts are, to the best of my knowledge, no more likely to spread disease than non-incestuous intercourse. That people tend to find them personally rather disgusting might result in their being hooked onto the moral modules by proxy. So long as morally condemning those who engaged in acts like incest didn’t carry any reliable fitness costs, such a connection would not have been selected against.

References: Young, L., & Saxe, R. (2011). When ignorance is no excuse: Different roles for intent across moral domains Cognition, 120 (2), 202-214 DOI: 10.1016/j.cognition.2011.04.005

Does Infidelity Pay Off (For Sparrows)?

For some species, mating can be a touch more complicated than others. In species where males provide little more than their gametes, the goal of mating for females is simple: get the best gametes available. While the specifics as to how that’s accomplished vary substantially between species, the overall goal remains the same. Since genes are all the female is getting, she may as well get the best that she can. In contrast, for some other species males provide more than just their genes; they also provide some degree of investment, which can take the form of a one-time gift through upwards of decades of sustained investment. In these species, females need to work this additional variable into their mating calculus, as the two goals do not always overlap. The male who’s willing to provide the best investment might not also happen to have the best genes, and pursuing one might risk the other.

Accordingly, it’s long been assumed that extra-pair mating (cheating) is part of the female strategy to have her cake and eat it too. A female can initiate a pair-bond with a male willing to invest while simultaneously having affairs with genetically higher-quality males, leaving the unfortunate cuckold to invest in offspring he did not sire. Undertaking extra-pair matings, however, can be risky business, in that detection by the investing male might lead to a withdrawal of investment and, in certain cases, bodily harm.

Good luck to all you parents when it comes to weaving that tidbit into your birds and bees talk.

These risks would require that offspring sired through extra-pair mating to tend to actually be fitter than offspring sired by the within-pair male, in order to be selected for. Abandonment can entail some serious risks, so females would need some serious compensating gains to offset that fact. A new paper by Sardell et al (2012) sought to determine whether extra-pair offspring would in fact be ‘fitter’ than within-pair offspring in Melospiza melodia – the song sparrow – when fitness was measured by lifetime reproductive success in number of offspring hatched, the number that survived to enter the breeding population, and the number of grand-offspring eventually produced. The results? Data gathered across 17 years, representing 2,343 hatchlings and 854 broods found that extra-pair offspring seemed to actually be less fit than their within-pair half-siblings. Well, kind of… but not really.

Over the 17 years of data collection, roughly 28% of the offspring were classed as being extra-pair offspring, and only broods with mixed paternity was considered for the present study (i.e. there was at least 1 offspring from the resident male and also at least 1 offspring from an extra-pair male). This cut the sample size down to 471 hatchlings, representing 154 mixed paternity broods across 117 pair bonds. The first point I’d like to make is that a 28% non-paternity rate seems large, and, unless it’s the result of an epidemic of forced copulations (rape), that means these female sparrows are having a lot of affairs, presumably because some mating module in their brain is suggesting they do

Within the sample of sparrows, female extra-pair offspring (the ones who were sired by the non-resident male) averaged 5.4 fewer hatched offspring over their lives, relative to their within-pair half-siblings; for extra-pair males, the corresponding average was 1.5 fewer offspring. However, not all of those hatchlings live to eventually breed. Of the 99 that did, the females that were the result of  extra-pair mating, on average, had 6.4 fewer hatchlings of their own, relative to the within-pair females; the extra-pair males also had fewer hatchlings of their own, averaging 2.6 fewer. Thus, relative to their within-pair half-siblings, extra-pair offspring seemed to produce fewer offspring of their own, and, in turn, fewer grand-offspring. (I should note at this point that any potential reasons for why extra-pair young seemed to be having fewer hatchlings are left entirely unexamined. This strikes me as something of a rather important oversight)

Are we to conclude from this pattern of results (as this article from the Huffington post, as well as the authors of the current paper did) that extra-pair mating is not currently adaptive?

And is it time for those who support the “good genes” theory to start panicking?

I don’t think so, and here’s why: when it came to the number of recruited offspring – the hatchlings who eventually reached breeding age – extra-pair females ended up having 0.2 more of them, on average, while extra-pair males had 0.2 less of them, relative to their within-pair half-siblings. While that might seem like something of a wash, consider the previous finding: within-pair offspring were having more offspring overall. If within-pair offspring tended to have more hatchlings, but a roughly equal number reach the breeding pool, that means, proportionally, more of the within-pair offspring were dying before they reached maturity. (In fact, extra-pair offspring had a 5% advantage in the number of total hatchlings that ended up reaching maturity) Having more offspring doesn’t mean a whole lot if those offspring don’t survive and then go on to reproduce themselves, and many of the within-pair offspring were not surviving.

One big area this paper doesn’t deal with is why that mortality gap exists; merely that it does. This mortality gap might even be more surprising, given that the potential risk of abandonment might mean males were less likely to have been investing when they doubted their paternity, though the current paper doesn’t speak to that possibility one way or another. Two of the obvious potential suspects for this gap are predation and parasites. Extra-pair young may be better able to either avoid predators and/or defend against pathogens because of their genetic advantages, leading to them being more likely to survive to breeding age. Then there’s also a possibility of increased parental investment: if extra-pair hatchlings are in better condition, (perhaps due to said pathogen resistance or freedom from deleterious mutations) the parents may preferentially divert scarce resources to them, as they’re a safer wager against an uncertain future. Alternatively, extra-pair offspring might have commanded a higher mating value, and were able to secure a partner more able and/or willing to invest long term. There are many unexplored possibilities.

The heart of the matter here concerns whether the female sparrows who committed infidelity would have been better off had they not done so. From the current data, there is no way of determining that as there’s no random assignment to groups and no comparison to non-mixed paternity broods (though that latter issue comes with many confounds). So not only can the data not definitely determine whether the extra-pair mating was adaptive or not, but the data even suggests that extra-pair offspring are slightly more successful in reaching breeding age. That is precisely counter to the conclusions reached by Sardell et al (2012), who state:

Taken together, these results do not support the hypothesis that EPR [extra-pair reproduction] is under positive indirect selection in female song sparrows…and in fact suggest… [that] other components of selection now need to be invoked to explain the evolution and persistence of EPR.

Their data don’t seem to suggest anything of the sort. They haven’t even established current adaptive value, let alone anything about past selection pressures. Sardell et al ‘s (2012) interpretation  of this mountain of data seems to be biting off more than they can chew.

It was a good try at least…

One final thoroughly confusing point is that Sardell et al (2012) suggest that how many grand-hatchlings the extra-pair and within-pair young had mattered. The authors concede that, sure, in the first generation within-pair sparrows had more hatchlings, proportionately more of which died, actually leaving the extra-pair offspring as the more successful ones when it came to reaching the breeding pool. They then go on to say that:

However, since EPY [extra-pair young] had 30% fewer hatched grandoffspring than WPY [within-pair young], higher recruitment of offspring of EPY does not necessarily mean that EPY had higher LRS [lifetime reproductive success] measured to the next generation. (p.790)

The obvious problem here is that they’re measuring grandoffspring before the point when many of them would seem to die off, as they did in the previous generation. So, while number of hatched grandoffspring says nothing important, they seem to think it does this time around. It’s been known that counting babies is only of limited use in determining adaptive value (let alone past adaptive value), and I hope this paper will serve as a cautionary tale for why that’s the case.

References: Sardell, R., Arcese, P., Keller, L., & Reid, J. (2012). Are There Indirect Fitness Benefits of Female Extra-Pair Reproduction? Lifetime Reproductive Success of Within-Pair and Extra-Pair Offspring The American Naturalist, 179 (6), 779-793 DOI: 10.1086/665665

Know Thyself…And Everyone Else Too, Apparently

“Most of us have experienced frustration in our social or political debates. We feel that the “other side” just “doesn’t get” our point of view, and that agreement could be reached if only we could somehow make those views, and the basis for those views, clear to them… By contrast, we think we get their point of view; we simply reject it as invalid, so that little would be gained from hearing them expound those views in more detail.. .That is, the members of each group feel that they understand the other group better than vice versa – that they are the ones being misunderstood, misinterpreted, or stereotyped, and that it is the other group that stands in need of enlightenment” – Pronin et al (2001)

In the context of debates, perhaps no rhetorical trick has a prouder history than framing. Recently, a bill called the “Paycheck Fairness Act” was voted down in the US. Are we to conclude from that outcome that those who voted against it are in favor of unfair paychecks? On the one hand, no; not unless we’re also supposed to consider plumbing one of the most sexist professions in the US. We encounter a similar frame issue regarding the debates about access to abortion and birth control, or, as it’s known to some, part of “the war on women”. Such a term is most certainly clumsy when it comes to accounting for (a) the large portion of women who consider themselves “pro-life”, and (b) the large number of men who would also be affected negatively were access to contraceptives and abortions limited.

On the other hand, those who create and make use of those frames are trying to get us to conclude precisely that: vote for this bill or it means you favor discrimination; vote on this bill or it means you hate women (or hate freedom, or love terrorists, or hate Capitalism, and so on). The way these issues are framed, then, might tell us more about what intentions or mindset the side doing the framing attributes to their opponents, rather than about the issue itself.

“Our newest line of frames serve to accentuate your overly optimistic sense of self. Now on sale at Michaels stores everywhere!”

It would seem that the frame-generators often feel confident enough in their assessment abilities that they feel they can accurately infer the hidden motives and intentions that guide the behavior of others. In other words, they feel they know their opponent’s “true self” well, perhaps even better than their opponents know themselves. However, if said opponent was to suggest that the framer themselves had some hidden motive they were blind to (or hiding from others), the framer would tend to be incensed at the suggestion. As we all know, it’s always the other side that has bias; not us. This brings us nicely to a paper entitled, “You don’t know me, but I know you: The illusion of asymmetric insight”  by Pronin et al (2001). Across six experiments, the authors sought to examine the extent to which people (a) felt they “knew” others better than those others knew themselves, (b) felt they “knew” themselves better than others knew them, and (c) the discrepancy between those two ratings.

Before I begin to review the paper, I’d like to commend these authors for coming so very close to articulating a coherent rationale for their research. Towards the end of their paper, the authors do speculate about the possible functions (and even potential costs) of the behavior in question. Unfortunately, my enthusiasm for that section is dampened somewhat by their initial stated rationale: their everyday experiences, and hunches that they might find some effect. The most useful part seemed more like an afterthought than a guiding force for this research, and I’ll return to that point later.

Now, onto the experiments. The first of the six experiments involved asking 125 undergraduates to rate how knowledgeable they were about a close friend of their choice, how knowledgeable that close friend was about the subject themselves, and to what extent the subject felt their “true self” and their friend’s “true self” was observable to others. On average, the subjects suggested they knew and understood more about their close friend than that close friend knew or understood about them. Further, subjects felt their “true self” was less observable than others “true selves” were (as indicated by circling pictures of partially submerged icebergs). The difference was small (about 4% in all cases), but in the predicted direction. While 24 of the subjects indicated that their close friend knew them better than the reverse, twice as many – 48 – indicated the opposite pattern.

The second study posed similar questions to roommate pairs in order to attempt and deal with certain sources of possible bias. Again, each roommate tended to think they knew their other roommate better than the reverse, and, again, the effect was similar in size. A larger effect (about 8%) was found with respect to self-knowledge: roommates tended to think they knew themselves better than their other roommate knew themselves. While the subjects in this sample did seem to feel that their roommate knew themselves better than the subject did (to the tune of about 10%), when it came to the reverse discrepancy – how well the subject knew themselves, relative to how well their roommate knew them – the size of the gap doubled.

Their entire friendship is based on each one thinking the other isn’t quite a smart.

While there’s more to the paper, it’s much the same as the first two studies (the perception that we can perceive others “true self” better than others can perceive ours), so I’m going to skip ahead to the final study that examined intergroup knowledge; the section the initial quote was drawn from. Eighty subjects were asked about whether they were liberal or conservative, pro-life or pro-choice, and whether they were men or women. Once their group membership was established, subjects were asked at a later date about whether their group had more knowledge or understanding about the other group in question. It’s worth noting there was some agreement here: both men and women agreed that women understood men more than the reverse. At least the genders were on the same page there (more or less; the women did still feel this gap was 2.5 times wider than the men thought it was). Liberals claimed to have greater knowledge of conservatives than they thought conservatives had about them (about a 7% gap), whereas conservatives felt they had about as much knowledge of liberals as liberals did about them. When it came to the abortion issue, pro-choicers felt they knew pro-lifers better than the reverse (11%), whereas pro-lifers thought they understood the pro-choicers more, but the gap wasn’t as wide (3%). However, since those two effects pulled in opposite directions, the gulf between the two groups actually loomed around 14% in that sample.

So where does all this leave us? As was the case in the last post, it would seem that people’s assessment abilities aren’t exactly wired for accuracy in some cases. This raises the inevitable question of precisely what they are wired for. The authors speculate that this pattern of assessment could:

“…lead us to talk when we would do well to listen and to be less patient than we ought to be when others express the conviction that they are the ones being misunderstood or judged unfairly. The same convictions can make us reluctant to take advice from others…but all too willing to give advice… But as previous research has pointed out, illusions can be also be helpful. The feeling that one knows another better than is really the case can increase readiness to trust, cooperate, or seek greater intimacy… This same illusion, however, can also increase one’s susceptibility to exploitation, or at least to unwise investments of time, resources, or affection.” (pp. 652-3)

This is the point I had initially commended the authors for, but, as I said, it feels like an afterthought. It feels that way because none of the six studies examined any of those contexts. There was no studies looking at argumentation, or cooperation, or trust, or exploitation. Those are all potential explanations that, frankly, these six studies should have been aimed at testing.

They’re like scientific hipsters; they didn’t test their hypotheses, but they did that ironically.

Had experiments been done to examine those contexts, the results might have been a bit more impressive. At the very least, it would allow us to begin to arbitrate between potential functions for this biased assessment of proportional understanding. For instance, one of the largest effects concerned a highly contentious issue: abortion. When arguing against someone – as the initial frame example showed – it would be beneficial to (perhaps incorrectly) believe you have a very strong grasp of your opponents motives if that confidence allows you to better persuade others. “The potential war on certain women” just doesn’t have the same ring to it as the “war on (all) women”; neither does the “There might be some active discrimination against women in regards to their salary in certain fields, but then again, there might not be…act”.

Claiming that you not only already understand your opponent’s position (which typically comes complete with some nefarious hidden motive), but also that your opponent doesn’t understand yours, is a rhetorical one-two punch. So long as you can effectively convey that message – as well as defend against such accusations when they’re directed your way – you’re in a very strong social position. When making such proclamations, being accurate is only useful insomuch as it would aid in being convincing.

References: Pronin, E., Kruger, J., Savtisky, K., & Ross, L. (2001). You don’t know me, but I know you: The illusion of asymmetric insight. Journal of Personality and Social Psychology, 81 (4), 639-656 DOI: 10.1037//0022-3514.81.4.639

I’m (Not) Sexy And I (Don’t) Know It

Imagine, for a moment, that someone you know tells you that they hate tomatoes, and how people who like tomatoes are seriously intellectually misguided. Later, you find that same person enjoying a BLT, going on about how much they really love the tasty red fruit between the lettuce and the bread. You might be rightly confused, and, if you’re someone who happens to like tomatoes, perhaps a bit put-off by their attitude about the whole thing.

On a related note, there are many critics of evolutionary psychology out there. Well, I say critics of evolutionary psychology, but what they’re actually critical of has little or nothing to do with anything found here. They just don’t know what they’re talking about. In fact, some of the ostensible critics of evolutionary psychology actually agree with many (or all) of the theoretical assumptions of field without even knowing it. Not only are these critics not experts about the topic they’re talking about (the politest possible way of putting it), but they are completely unaware of that fact. They display the same quality of self-assessment that is found in the recently iconic “I’m sexy and I know it”.

“Wiggle Wiggle Wiggle Wiggle.” -LMFAO

Not so recently, Kruger and Dunning (1999) examined self-generated judgments, asking people to rate their performance on various task of logic, humor, and grammar, and then compared those judgments to people’s actual performance. At this point, I doubt that I’m spoiling anything by telling you that people aren’t very good at delivering accurate self-assessments in certain contexts, humor, logical, and grammatical abilities being three of them.

The first of the four studies in this paper concerned humor. Thirty ostensible jokes were rated by a panel of professional comedians on a scale of 1 to 11, yielding an average humor score for each item. These 30 items were then rated on the same scale by participants; those who assessed the jokes with a similar score to the comedians were said to have performed well, and the further the scores deviated, the worse their performance was rated as being. Subjects then rated their performance in assessing the quality of jokes, relative to their peers. On average, people in the sample rated themselves to have performed in the 66th percentile; moderately above average. Among those who scored in the lowest quarter of the distribution when it came to assessment – the worst performers – they still rated themselves, on average, to be in the 58th percentile; mildly above average in their ability.

The second study looked at performance on tests of logical reasoning. The logical domain has an advantage over humor in that there are objective answers to the questions being examined, rather than subjective ones. Forty-five subjects completed a 20-item test of logical reasoning and were then asked to estimate both their logical reasoning abilities, relative to their peers, and also how many of the 20 questions they got right. Again, on average, the subjects placed their performance in the 66th percentile. In this case, perceived performance on the task was not significantly correlated to actual performance. Also again, the worst performers – the ones in the bottom forth of the sample – rated their logical reasoning ability in the 68th percentile and their performance in the 62nd. Whereas they thought they had answers 14.2 questions correctly, they had in fact only answer 9.5 right, on average. Lest I bore you with more repetition, a set of nearly identical results were also obtained for measures of grammatical ability in the third study.

“So many people wasted their time getting a degree they won’t use; completely unlike me!”

The forth study found that training can, to some extent, help participants become a bit more realistic in their self-assessments. In fact, those subjects that actually fell into the bottom quarter of all subjects had initially rated their performance as being in the 55th percentile. After training, when they realized some of the mistakes they had made, those estimates were revised…to be in the 44th percentile. So they went from rating themselves as slightly above average to slightly below average. They were now off by a mere 30 points, rather than 40, but that’s progress, I suppose.

Kruger and Dunning (1999) attempt to explain this pattern of results by suggesting that those who lack the skills to perform in certain domains tend to, as a result, lack the ability to judge competent from incompetent performance. The two, as the authors suggest, “are often the very same skills” (p.1121 – my emphasis). While that may sound like a plausible explanation at first glance, I find it lacking. As the authors note, is that this effect is not found universally across domains: my ability to recognize some badly performed karaoke is not dependent on my ability to produce good karaoke, and I won’t be attempting to dunk on Michael Jordan anytime soon even though I’m not good at basketball. As I’ve written before, these positively biased self-assessments (known as the ‘above-average effect’) would appear to be more prevalent in fuzzy domains. When skills (like whether you can successfully skate down three flights of stairs) or physical traits (like your height or eye color) are readily observable, we shouldn’t expect to see much in the way of biased self-assessment. This is because, outside of certain social contexts involving persuasion, being wrong about them tends to carry costs that aren’t going to be reliably offset by the benefits.

Further, it seemed to be the case that the accuracy of self ratings had more to do with chance than assessment proficiency in the present study (Burson et al., 2006). Most everyone seems to hover around a self-assessment set of point of mildly-to-moderately above-average. For those who perform at or around that level, those kind of assessments will tend to be accurate; for anyone performing above or below that point, you’ll find that self-assessments get less and less accurate. Since most people, by definition, perform below that point, you’ll see the worst misrepresenters there, which we do. However, it also works in the opposite direction; the best performers, for instance, consistently underestimated how competent they were, relative to others. If assessment abilities are supposed to be tied to production abilities, it’s unclear why that gap exists. Kruger and Dunning (1999) suggest that this is because of a false consensus effect among the highly knowledgable (i.e. “I assume other people know what I do”), which is more of a restatement of the original finding than an explanation of it.

Thanks again, field of psychology.

This non-explanation is similar to Kruger and Dunning’s (1999) suggestion that a lack of metacognitive abilities among the worst performers are responsible for their very poor assessments. They are bad at assessing their performance because they lack metacognitive abilities. What are metacognitive abilities, you ask? They’re the abilities required to accurately assess performance. So the fact that people are bad at assessing their performance is explained by the fact that those people are bad at assessing their performance. All this theoretical spinning is making me dizzy.

I feel these results can be better explained (by which I mean actually explained) by considering a persuasion framework. There are certain things that might make me better off if others believe them (i.e. I’m a desirable mate or a reliable friend). In the service of persuading others about them, it helps me to be strategically wrong about them myself (Kurzban, 2010). However, when it comes to things that can’t be persuaded (like gravity) or cases where persuasion seems implausible (like your ability to speak a foreign language you can’t actually speak), self-assessments should tend to head towards a more accurate assessment, provided relevant feedback information is present. The important thing to bear in mind here is that it’s not the accuracy of these self-assessments, per se, that matter; what matters is what those self-assessments ultimately end up leading an organism to do. Evolution is blind to what you feel but not blind to what you do. These self-assessments should be considered in light of their consequences, not their accuracy.

References: Burson KA, Larrick RP, & Klayman J (2006). Skilled or unskilled, but still unaware of it: how perceptions of difficulty drive miscalibration in relative comparisons. Journal of personality and social psychology, 90 (1), 60-77 PMID: 16448310

Kruger, J., & Dunning, D. (1999). Unskilled and unaware of it: How difficulties in recognizing one’s own incompetence lead to inflated self-assessments. Journal of Personality and Social Psychology, 77 (6), 1121-1134 DOI: 10.1037//0022-3514.77.6.1121

Kurzban, R. (2010). Why everyone (else) is a hypocriteEvolution and the modular mind. Princeton, NJ: Princeton University Press.