Thursday, July 27, 2017

How Everyone Might Reasonably Believe They Are Much Better Than Average

In a classic study, Ola Svenson (1981) found that about 80% of U.S. and Swedish college students rated themselves as being both safer and more skilled as drivers than other students in the room answering the same questionnaire. (See also Warner and Aberg 2014.) Similarly, most respondents tend to report being less susceptible to cognitive biases and sexist bias than their peers, as well as more honest and trustworthy -- and so on for a wide variety of positive traits: the "Better-Than-Average Effect".

The standard view is that this is largely irrational. Of course most people can't be justified in thinking that they are better than most people. The average person is just average! (Just between you and me, don't you kind of wish that all those dumb schmoes out there would have a more realistic appreciation of their incompetence and mediocrity? [note 1])

Particularly interesting are explanations of the Better-Than-Average Effect that appeal to people's idiosyncratic standards. What constitutes skillful driving? Person A might think that the best standard of driving skill is getting there quickly and assertively, while still being safe, while Person B might think skillful driving is more a matter of being calm, predictable, and within the law. Each person might then prefer the standard that best reflects their own manner of driving, and in that way justify viewing themselves as above average (e.g., Dunning et al. 1991; Chambers and Windschitl 2004).

In some cases, this seems likely to be just typical self-enhancement bias: Because you want to think well of yourself, in cases where the standards are ambiguous, you choose the standards that make you look good. Changing example, if you want to think of yourself as intelligent and you're good at math, you might choose to think of mathematical skill as central to intelligence, while if you're good at practical know-how in managing people, you might choose to think of intelligence more in terms of social skills.

But in other cases of the Better-Than-Average Effect, the causal story might be much more innocent. There may be no self-flattery or self-enhancement at all, except for the good kind of self-enhancement!

Consider the matter abstractly first. Kevin, Nicholas, and Ana [note 2] all value Trait A. However, as people will, they have different sets of evidence about what is most important to Trait A. Based on this differing evidence, Kevin thinks that Trait A is 70% Property 1, 15% Property 2, and 15% Property 3. Nicholas thinks Trait A is 15% Property 1, 70% Property 2, and 15% Property 3. Ana thinks that Trait A is 15% Property 1, 15% Property 2, and 70% Property 3. In light of these rational conclusions from differing evidence, Kevin, Nicholas, and Ana engage in different self-improvement programs, focused on maximizing, in themselves, Properties 1, 2, and 3 respectively. In this, they succeed. At the end of their training, Kevin has the most Property 1, Nicholas the most Property 2, and Ana the most Property 3. No important new evidence arrives in the meantime that requires them to change their views about what constitutes Trait A.

Now when they are asked which of them has the most of Trait A, all three reasonably conclude that they themselves have the most of Trait A -- all perfectly rationally and with no "self-enhancement" required! All of them can reasonably believe that they are better than average.

Real-life cases won't perfectly match that abstract example, of course, but many skills and traits might show some of that structure. Consider skill as a historian of philosophy. Some people, as a result of their training and experience, might reasonably come to view deep knowledge of the original language of the text as most important, while others might view deep knowledge the the historical context as most important, while others might view deep knowledge of the secondary literature as most important. Of course all three are important and interrelated, but historians reasonably disagree substantially in their comparative weighting of these types of knowledge -- and, I think, not always for self-serving or biased reasons. It's a difficult matter of judgment. Someone committed to the first view might then invest a lot of energy in mastering the details of the language, someone committed to the second view might invest a lot of energy in learning the broader historical context, and someone committed to the third view might invest a lot of energy in mastering a vast secondary literature. Along the way, they might not encounter evidence that requires them to change their visions of what makes for a good historian. Indeed, they might quite reasonably continue to be struck by the interpretative power they are gaining by close examination of language, historical context, or the secondary literature, respectively. Eventually, each of the three might very reasonably regard themselves as a much better historian of philosophy than the other two, without any irrationality, self-flattery, or self-enhancing bias.

I think this might be especially true in ethics. A conservative Christian, for example, might have a very different ethical vision than a liberal atheist. Each might then shape their behavior according to this vision. If both have reasonable ethical starting points, then at the end of the process, each person might reasonably regard themselves as morally better than the other, with no irrational self-enhancing bias. And of course, this generalizes across groups.

I find this to be a very convenient and appealing view of the Better-Than-Average Effect, quite comforting to my self-image. Of course, I would never accept it on those grounds! ;-)


Bird lover said...

Given that people normally spend the bulk of their energy trying to show themselves as better than others, it is no mystery that they think they are. In fact to have self awareness is to think one is better than others. To be aware of the self, one has to see one's self as different. This difference is valuative. A synonym for different is better.

Marco Devillers said...

I doubt people's behavior is driven by ethics. Pretty sure they just rationalize after the fact, sometimes before, and you can rationalize everything.

Robert Dawson said...

Another rational explanation in some cases: they are using a metric in which the distribution is skewed. There`s an old joke about having more than the average number of legs... because amputations are more common than extra limbs, that average is just under two.

Now, "average" is not a technical word in statistics. It can mean any of several measures of location - mean, median, mode, or others - and is used casually to mean some conflation of the three. The leg joke is, perhaps, funnier because "mode" colors our interpretation in some ways. For driving skill, which doesn't have a canonical numerical measure, "mean" is strictly meaningless. Nevertheless, I think we do expect the undefined "average" to have the property that everybody's individual behavior affects it - even though this is untrue for median or mode.

While there are some drivers who have taken special training (bodyguards, police, race drivers), when they are driving their private cars on the highway this normally doesn't show. To the bystander, all good drivers resemble each other: the visibly bad driver is the one on the stretcher, or the jerk who cut you off. So, if you look around you, you see a few visibly worse drivers and nobody who is visibly better: good driving is generally unremarkable. This, I'd argue, makes it reasonable (in the absence of data that you cannot get) to assess "average" driving skill as a little worse than your own.

Possibly it's a silly calculation to do, under the circumstances. But the experimenter did prompt these subjects to do it (another can of worms.)