Self-Enhancement in a Small World

My conference is cooler than your conference.

Posted Sep 18, 2016

J. Krueger
Source: J. Krueger

You ever read any Nietzsche? Nietzsche says there's two kinds of people in the world: people who are destined for greatness like Walt Disney... and Hitler. Then there's the rest of us, he called us "the bungled and the botched." We get teased. We sometimes get close to greatness, but we never get there. We're the expendable masses. We get pushed in front of trains, take poison aspirin... get gunned down in Dairy Queens.

            ~Robyn Williams in the Fisher King

In Small World, David Lodge (1985) describes the adventures of a cast of professors of English and literary criticism on the conference circuit. They contrive to see the world on someone else’s nickel and they bang out papers to justify the expenditures. Literary criticism comes across as a parasitic discipline that does not have a ‘there’ (I am sympathetic to this assessment). Sure enough, though, many of the professors feel strongly that their own papers are superior to those of their colleagues and they all hope to get the UNESCO Chair of Criticism, which would allow them to do more of the same. The conferences are boring, the papers soporific. What makes conferencing worthwhile is the opportunity to socialize, gossip, tack on a few days of sightseeing, and maybe get lucky. Reliably, there is also the reinforced belief that one’s own work is better than one’s neighbor’s work.

For social psychologists it is easy to look down on professors who not only have no empirical data but who don’t even exist. Surely, our approach to conferencing meets a higher standard. Yet, our being entertained by a novel such as Small World suggests that it brings out a bit of self-recognition – if only by way of caricature. Its theme rings true. We do the things the fictional critics do, and we feel more keenly than they do that the scholarly value of conferences is mainly mythical. As psychologists, we know that listening to a lecture is the worst way to learn known to man, but here we are.

Today, we explore the phenomenon of self-enhancement, which is one of the oldest and most robust phenomena in social cognition. Many people self-enhance in many circumstances. Exceptions and moderator variables have been noted, but there is little risk that we will one day read in the morning paper that the self-enhancement effect was felled by the blunt axe of replication science.

There’s a host of ways in which self-enhancement can be studied, and the most prominent is to look for the Better-Than-Average Effect as the footprint of the self-enhancing mind. Relevant data go back to mythical time. The most dramatic demonstration in living memory is Ola Svenson’s (1981) Acta Psychologica paper  titled “Are we all less risky and more skillful than our fellow drivers?” This is a bit of a rhetorical question because you would not publish a paper where the answer is “Er, no.” When the median percentile rating for the self is about 80%, something seems amiss. At least some of those who claim to be better than average drivers must be wrong. But who? And how do the drivers come up with these numbers?

Three process hypotheses

Let’s begin with the second question. There are three types of answer: motivation, ability, and ecology. The motivational hypothesis says that people have a desire to feel good about themselves and that nudging their self-other comparisons toward a positive result is one way to satisfy this desire. There is a body of creative work dedicated to securing evidence for the motivational hypothesis – notably from Mark Alicke’s lab (e.g., Alicke & Govorun, 2005) – and I believe there is enough data to make the case. Making the case must have been difficult because in its naïve form the motivational hypothesis flirts with tautology. To put it bluntly, we would not want to say that people self-enhance because they want to and accept evidence of self-enhancement as corroborating evidence for the process hypothesized to generate this very result. In other words, a finding cannot do double duty as evidence for the existence of a phenomenon and evidence for a particular process that generated this phenomenon.

The other difficulty is that there must be a limit to the ability to satisfy a desire by mere self-manipulation. You cannot quench your thirst by imagining to have a drink. If anything, a hallucinated satisfaction will make the desire stronger. If there is a need to be better than average, the mere imagination that one is better than average is a short-term palliative at best. If we tried to solve this question by retreating to the claim that people merely want to feel they are better than average and then satisfy this need by thinking they are better than average, we are back in the land of the tautologues.

The ability hypothesis says that a lot of people are just too dim to process information properly. They have trouble with self-insight. Their sense of superiority is a cognitive-perceptual illusion produced by the untrained infantile “System 1,” which runs the intuitive, unreflective inference machine in what used to be called the limbic system. According to the so-called unskilled-and-unaware hypothesis it is the low performers who are particularly impaired (Kruger & Dunning, 1999). Not only do they perform poorly (e.g., when driving), but they also lack insight into their lack of skill. This is a tricky hypothesis to test and David Dunning and I have argued about how it might be done (Krueger & Mueller, 2002). Here I just want to say this: If low performers overestimate their own performance, this is consistent with the idea that they are not smart enough to recognize their own low performance, but it is not a conclusive test of the hypothesis that unawareness of own low performance causes self-overestimation. High performers do not overestimate their own performance. The claim that they have greater self-insight independent of their performance outcome implies that these high performers would recognize their own performance if they performed poorly. But they didn’t. Therefore the claim that poor performers overestimate their own performance because they are unaware of their own poor performance also flirts with tautology.

The ability hypothesis also shares the second difficulty with the motivation hypothesis. How bad can it get before self-enhancers face social and material challenges that are so costly that they interfere with reproductive success?

The ecological hypothesis says that some clues regarding the origins of self-enhancement lie outside of the individual, in the information that is available and how this information interacts with evolved capacities to process it. Consider Svenson’s finding that most of us think we are better drivers than others. If we plotted all acts of driving and the events they produce against an axis of skill or goodness, we’d see a left-skewed distribution, where very bad acts and events lie to the left end of the scale in the thin tail of the distribution. Most acts and events on the road are good or good enough to find their way to the right half of the scale. There isn’t much of a right tail because there are fewer ways in which drivers can excel than ways in which they can mess up. A left-skewed distribution is a happy distribution, as Allen Parducci (1968), taught us with his range-frequency theory.

Consider Derik who is asked how well he drives relative to others. Derik begins by sampling from the left-skewed distribution as represented by his own experience. His method of sampling may be selective and biased, but it does not need to be (Fiedler & Juslin, 2005). Suppose he samples randomly. If so, most of his samples will yield a mean above the midpoint of the scale. Indeed most samples of small to medium size will overestimate the mean a little, while a few will underestimate it by a lot. Next, Derik samples the acts and events involving other drivers. This yields two scenarios of interest: In one scenario, Derik does not come up with much of a sample at all and being forced to make an estimate for others he goes to the midpoint of the scale. Then, comparing his self-estimate to the other-estimate, he concludes that he is a better than average driver. In the other scenario, he has a very large sample of others, and he puts his estimate for their average skill correctly a bit above the midpoint of the scale. Then he compares this estimate with his self-estimate. Since the latter is based on small samples most of which will overestimate the average due to distributional skew, Derik again concludes that he is a better than average sample. According to this view of unbiased ecological sampling, there will be self-enhancement if the underlying distribution is negatively skewed and if there is a difference in sample size (in whatever direction) between self- and other- related information.

Three methods of measurement

Much as versions of the three basic process hypotheses continue to float in the literature, so do different types of measure to index self-enhancement.

Social comparison measures im- or explicitly incorporate both estimates of self and estimates of others. Svenson and Dunning, for example, used percentile estimates, which require a respondent to judge the self, judge others, assess the difference, and scale it along a single dimension. An alternative is to collect self- and other-judgments from the respondents and to do the subtracting and rescaling for them.

Social reality (or self-insight) measures relate self-judgments to judgments made by an aggregate of observers or other external information, such as test scores. The researchers then compute difference scores or regression residuals as measures of self-enhancement vs. –effacement.

The problems with these measures are (or should be) well known. Adding the two measures up into a composite does not help much either (Krueger & Wright, 2011). The social-relations index suggested by Kwan, John, Robins, Bond, and Kenny (2004) was meant to capture an interaction between social comparison and social reality. This was attempted by subtracting a target effect (how the person is seen by others) and a perceiver effect (how the person sees others) from the person’s self-judgment. In analysis of variance one finds an interaction by subtracting the main effects. In a 2x2 design, an interaction presents itself as the cross-over of two lines connecting 2 pairs of points. But if there is only one point because there is only one self-judgment per person, no such interaction can be seen. The dual residual may be part of an interaction or it may be measurement error. We will never know.

Kwan and colleagues tried to combine the social comparison with the social reality approach and this is a good idea. Last year, Patrick Heck and I picked it up and suggested a decision-theoretic approach to the measurement of self-enhancement (Heck & Krueger, 2015). The rationale is simple. Measure participants’ performance on a test, ask them how well they think they did and ask them how well they think the average person did. Once you score the test, you can group participants into the four classes of those who think they did better than average and did (H for ‘hit’), those who think they did better than average but did not (FA for ‘false alarm’), those whose think they did not do better than average but did (M for ‘miss’), and those who think they did not do better than average and indeed did not (CR for ‘correct rejections’). The results of our studies are not as interesting (we think) as the opportunities this simple decision-theoretical method provides for future research.

A key advantage of the decision-theoretic measure is that it offers an answer to the first question, which was ‘Who among the self-enhancers is actually wrong?’ The decision-theoretic measure separates error from bias, which is a critical thing to do, and everyone who has studied classification knows it.

Self-enhancement and confidence

In ongoing work, we are looking at the associations between confidence in one’s judgments and true performance scores. Preliminary results show that when self-enhancers (S > O) are more confident that their relative self-assessment is correct than are self-effacers (S < O). Moreover, those who truly score better than average are more confident than those who do not. If confidence ratings are related to true scores beyond the shared association with S, then combining self-judgments with confidence judgments will be a way to predict people’s true scores more accurately. The inclusion of a confidence measure will provide a more fine-grained look at those respondents who commit a self-enhancement error (FA), isolating those who commit this error with full confidence as those who might need the most help.  


Alicke, M. D., & Govorun, O. (2005). The better-than-average effect. In M. D. Alicke, D. Dunning & J. Krueger (Eds.), The self in social judgment (pp. 85-106). New York: Psychology Press.

Fiedler, K., & Juslin, P. (2005). Information sampling and adaptive cognition. New York, NY: Cambridge University Press.

Heck, P. R., & Krueger, J. I. (2015). Self-enhancement diminished. Journal of Experimental Psychology: General, 144, 1003-1020.

Krueger, J., & Mueller, R. A. (2002). Unskilled, unaware, or both? The contribution of social-perceptual skills and statistical regression to self-enhancement biases. Journal of Personality and Social Psychology, 82, 180-188.  

Krueger, J. I., & Wright, J. C. (2011). Measurement of self-enhancement (and self-protection). In M. D. Alicke & C. Sedikides (Eds.), Handbook of self-enhancement and self-protection (pp. 472-494). New York, NY: Guilford.

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, 1121–1134.

Kwan, V. S. Y., John, O. P., Kenny, D. A., Bond, M. H., & Robins, R. W. (2004). Reconceptualizing individual differences in self-enhancement bias: An interpersonal approach. Psychological Review, 111, 94-110.

Lodge, D. (1985). Small world. New York, NY: MacMillan.  

Parducci, A. (1968). The relativism of absolute judgments. Scientific American, 219, 84-90.

Svenson, O. (1981). Are we all less risky and more skillful than our fellow drivers? Acta Psychologica, 47, 143-148.