Verified by 
He who knows others is learned; he who knows himself is wise. ~ Lao-Tzu
The challenge of self-knowledge is a perennial concern in Western thought (and elsewhere; cf. Lao-Tzu); it is intimately connected with the human capacity for self-awareness and self-reflection. A naïve response to the challenge is to say that we know ourselves perfectly well, thank you very much. A more sophisticated view grants that we ourselves, as objects of our own perception, are unlike simple physical objects that can be described by a small set of unchanging features. Humans are dynamical systems, in which myriad features interact with an equally multivariate and dynamical environment. Yet, the accuracy of some of our self-beliefs can be assessed, and it is not bad. Within limits, we can reliably and validly describe who and what we are.
Many social psychologists think there is more to the story of inaccuracy than simple lack of information, lack of insight, or random error. The mainstream of the field, with a nod to Dr. Freud, assumes that unconscious defenses and distortions systematically undermine the accuracy of self-perception. According to this view, we can’t look at ourselves in the cruel light of objectivity. If we saw ourselves as we really are, we’d be too shocked to function. As Mark Twain put it, "Know thyself—and then thou wilt despise thyself.” So we protect ourselves with perceptual blind spots and distort what we see.
This raises an intriguing possibility. If the defenses and blind spots are a function of the ego, then these defenses should not matter to the perceptions of the alter. Another person, who does not have a stake in our ego, can perceive and judge us more objectively than we ourselves can. Well, that’s the hypothesis. For decades, though, the empirical result has been that self-ratings and other-ratings are about equally valid, and if there’s a difference, self-ratings tend to have the advantage. The explanation is that despite the blind spots and defenses, the self still has a lot more information at its disposal than observers do.
A New Look
Simine Vazire tried to breathe new life into this area of research by proposing that either the self or the observer can be more accurate; it all depends on the type of trait being judged. In an article in the Current Directions in Psychological Science (CDPS), Vazire & Carlson (2011) suggest selves do better when the traits are hard to observe from the outside and neither extremely desirable nor undesirable (e.g., anxiety). The hard-to-observe aspect implies that selves, but not observers, have access to privileged internal information; the lack of extreme evaluation implies selves are not overly motivated to defend or distort. More provocatively, Vazire & Carlson suggest that observers (friends) do better when traits are hard to observe but highly evaluative (e.g., intelligence). The hard-to-observe aspect should put observers again at a disadvantage, and so the trait’s extreme evaluativeness seems to matter all the more. For traits that are easy to observe and low in evaluation, there is no difference, and for traits that are easy to observe and highly evaluative, there are no data.
Here’s where the story could end. When observers are less accurate than selves, they are so because they have less information. When selves are less accurate than observers, they are so because they defend and distort. But the story does not end here. The original data are reported in an article in the Journal of Personality and Social Psychology (Vazire, 2010). Here we see two graphs instead of one. The graph reproduced for CDPS shows accuracy correlations, which use aggregated judgments for observers but use individuals’ self-judgments. If one were to compare the accuracy attained by a single self with the accuracy attained by a single friend, the self’s advantage increases for hard-to-observe/low-evaluation traits, and its disadvantage for hard-to-observe/high-evaluation traits disappears. And this is what the second graph in the JPSP shows. Comparing a single self with a single friend is comparing apples with apples; comparing a single self with aggregated friends is comparing apples with oranges. Remember, the question of interest is whether the self-perspective is different (better or worse) than the observer’s perspective. A proper test requires that other variables be held constant. If the number of raters or ratings is varied for one perspective but not for then other, there is a confound, and confounds are the pallbearers of validity.
Vazire knows this because it is textbook statistics. If each of a set of test items (observers) has non-zero validity (is correlated with a criterion), and if these items are imperfectly correlated with one another, then combining (averaging) them increases the correlation with the criterion. In a footnote (#6 on p. 290), Vazire acknowledges and dismisses the issue in one breadth.
“Although disaggregation makes it easier to compare the magnitude of the effect sizes, it should be noted that the increased reliability from aggregation is not merely a statistical artifact. It is a fact of life that you can obtain impressions of a person from multiple peers, but you can ask only one person for a self-rating. Thus, the increased reliability of aggregated peer-ratings is not artificial—in real-life contexts the fact that multiple peers can be turned to for information about a target is a real and important advantage.”
An artifact does not become a fact because it is possible to create it in a real life. If multiple observers are more accurate in their perceptions of you after their perceptions are averaged, it is not their perspective that is superior, but simply the fact there are many of them and only one of you. Their group advantage has absolutely no bearing on any perceptual defenses or distortions you might have. Incidentally, it is not true that multiple ratings or perspectives cannot be extracted from a single self. Imagine yourself rating yourself multiple times (e.g., once a day for a week), and then averaging the result. That way you too could capitalize on the error-reducing capability of statistical aggregation. Better yet, rate your self as you see yourself in diverse regions of your life space. You at home, where the observer is a co-habitant; you at the office, where the observer is a colleague; you at the gym, etc.
The defense-and-distortion hypothesis is not doing well. Not only do the friends in Vazire’s study individually fail to achieve greater accuracy than the selves, they also fail to see the selves less positively. If the pre-dominant motivational interest of defense and distortion is to make the self look good, than that should be the case. But it is not. Vazire & Carlson note that often close others judge us even more positively than we judge ourselves without being less accurate. This is proof that positivity bias and accuracy are two different things.
The perennial recital of the mantra that bias (defense and distortion) begets inaccuracy is a repetition compulsion on the part of the researchers. It is a blind spot that allows them to maintain their self-image of being disinterested detectives of the truth.
Vazire, S. (2010). Who knows what about a person? The self-other knowledge asymmetry (SOKA) model. Journal of Personality and Social Psychology, 98, 281-300.
Vazire, S., & Carlson, E. N. (2011). Others sometimes know us better than we know ourselves. Current Directions in Psychological Science, 20, 104-108.
