Sorry if the positive illusions are old hat, but I searched and couldn't find any mention of this peer prediction stuff! If nothing else, I think the findings provide a quick heuristic for getting more reliable predictions of your future behavior - just poll a nearby friend!

Peer predictions are often superior to self-predictions. People, when predicting their own future outcomes, tend to give far too much weight to their intentions, goals, plans, desires, etc., and far to little consideration to the way things have turned out for them in the past. As Henry Wadsworth Longfellow observed,

"We judge ourselves by what we feel capable of doing, while others judge us by what we have already done"

...and we are way less accurate for it! A recent study by Helzer and Dunning (2012) took Cornell undergraduates and had them each predict their next exam grade, and then had an anonymous peer predict it too, based solely on their score on the previous exam; despite the fact that the peer had such limited information (while the subjects have presumably perfect information about themselves), the peer predictions, based solely on the subjects' past performance, were much more accurate predictors of subjects' actual exam scores.

In another part of the study, participants were paired-up (remotely, anonymously) and rewarded for accurately predicting each other's scores. Peers were allowed to give just one piece of information to help their partner predict their score; further, they were allowed to request just one piece of information from their partner to aid them in predicting their partner's score. Across the board, participants would give information about their "aspiration level" (their own ideal "target" score) to the peer predicting them, but would be far less likely to ask for that information if they were trying to predict a peer; overwhelmingly, they would ask for information about the participant's past behavior (i.e., their score on the previous exam), finding this information to be more indicative of future performance. The authors note,

There are many reasons to use past behavior as an indicator of future action and achievement. The overarching reason is that past behavior is a product of a number of causal variables that sum up to produce it—and that suite of causal variables in the same proportion is likely to be in play for any future behavior in a similar context.

They go on to say, rather poetically I think, that they have observed "the triumph of hope over experience." People situate their representations of self more in what they strive to be rather than in who they have already been (or indeed, who they are), whereas they represent others more in terms of typical or average behavior (Williams, Gilovich, & Dunning, 2012).

I found a figure I want to include from another interesting article (Kruger & Dunning, 1999); it illustrates this "better than average effect" rather well. Depicted below is an graph summarizing the results of study #3 (perceived grammar ability and test performance as a function of actual test performance):

Along the abscissa, you've got reality: the quartiles represent scores on a test of grammatical ability. The vertical axis, with decile ticks, corresponds to the same peoples' self-predicted ability and test scores. Curiously, while no one is ready to admit mediocrity, neither is anyone readily forecasting perfection; the clear sweet spot is 65-70%. Those in the third quartile seem most accurate in their estimations while those the highest quartile often sold themselves short, underpredicting their actual achievement on average. Notice too that the widest reality/prediction gap is for those the lowest quartile.

New to LessWrong?

New Comment
16 comments, sorted by Click to highlight new comments since: Today at 8:29 PM

This doesn't really tell us a lot about how people predict others' success. The information has been intentionally limited to a very high degree. It's basically asking the test participants "This individual usually scores an 87. What do you expect her to score next time?" All of the interactions that could potentially create bias has been artificially stripped away by the experiment.

This means that participants are forced by the experimental setup to use Outside View, when they could easily be fooled into taking the Inside View and being swayed by perceptions of the student's diligence, charisma, etc. The subject would probably be more optimistic than average about themselves, but the others' predictions might not be nearly as accurate if you gave them more interaction with the subject.

In baseball prediction, it has been demonstrated that a simple weighted average with an age factor is nearly the best predictor of future performance. Watching the games and getting to know the players in most cases makes prediction worse. [I can't easily find a citation for this, but I think it came originally from articles at]

This really just leaves us with "use outside view to predict performance," which is useful but not necessarily novel.

Depicted below is the general relationship between the way we actually perform and the way we predict we will perform:

It looks to me like this is a copy of the graph summarizing the results of one of the four studies (specifically, the third) in the cited paper. It is not accurate to describe it as "the general relationship". The other three studies don't show the same curve, although they all do show the bottom quartile predicting their ability as above the median but below the top quartile.

One thing I've always found suspicious about this paper: Why report quartiles? The four studies had n=65, 45, 84, 140 respectively. Why choose to bin this large number of participants into only four bins? That seems unusually low resolution. Why quartiles and not quintiles or deciles?

You're exactly right, sorry. I'll keep the picture because I think it suffices to illustrate the trend, but I'll update my description for clarity. Here are the other summary graphs for studies 1, 2, and 4

ETA: Strangely apropos this post, David Dunning is doing a Reddit AMA right now; I should go ask him why he and Kruger (1999) chose to report quartiles!

If you tell me you are going to do better than the outside view says you'd better have a seriously convincing argument.


People, when predicting their own future outcomes, tend to give far too much weight to their intentions, goals, plans, desires, etc., and far to little consideration to the way things have turned out for them in the past.

I'm often curious about why we evolved in such a haphazard way. Sure, evolution gets stuck on local optima all the time, but this seems to imply that realistically modelling your peers and unrealistically modelling yourself has some sort of fitness payoff.

Generally speaking, it's more useful to believe that you can improve than to believe that improvement is impossible. If toddlers predicted their own walking ability based on past performance only, they would despair of ever walking.

People have a lot of their stated beliefs for signaling purposes and don't use them for guiding their own actions. I think you might get different outcomes if you ask students to bet that they will achieve a certain mark. Actually paving to put money on the line will make student less willing to overpredict their own performance.

has some sort of fitness payoff.

You impress potential mates.

Why don't potential mates ignore your self-estimates and go with the outside view like all other observers?

There are two reasons why having an inflated view of oneself is useful (there are also, of course, reasons why it's not useful, too).

First is that bragging is very widespread technique for winning mates, both in humans and animals. If you can't brag convincingly -- and internal belief certainly helps -- you're falling behind in the competition.

Second is that overestimation of your own abilities leads to feelings of confidence and self-assurance which are useful feelings to have in general (subject to reasonable limits, of course) and, in particular, are very valuable in giving out the right signals to potential mates. In the latter case, though, I think there's a noticeable difference between the sexes -- being self-assured is much more valuable for (straight) males then for (straight) females.

PSA: coincidentally, David Dunning (an author in every study I mention above) is currently doing a Reddit AMA. I did not plan this, but if you have any questions for him, he's all ears!

Don't you mean "Others' predictions of your performance are usually more accurate"?

And your last paper and graph are rather well-known by now.

Ah, yes that's much better isn't it. Am I allowed to change this? Sorry for being such a flagrant newcomer; it seems like I really need to tighten up my language.

EDIT: I've given it some thought and I think it has something to do with being active on Reddit, where there's lots of incentive to sensationalize your posts. I will be mindful of this going forward.


Sorry for being such a flagrant newcomer

Your concern for providing quality content is more valuable than any social hierarchy nonsense. Thanks for your contribution.

In general the standard is to edit to remove errors like that to the extend that the edit doesn't invalidate existing comments. Often simply posting: "Thanks for pointing out the error, I fixed it." allows any further reader to see that the the person pointing out an error was making a valid point while at the same time correcting the error.

Am I allowed to change this?

I am unsure of the actual mechanics of doing that, but there are no prohibitions against edits.

it seems like I really need to tighten up my language.

It would probably be for the best :-) LW is full of people who are not above shredding sloppy language for minor amusement value (and, of course, to save the world).