Track Records

More evidence that a huge way to improve your accuracy is track records: simply write down your forecasts and check their accuracy later:

In the context of a Super Bowl loss (Study 1), a presidential election (Studies 2 and 3), an important purchase (Study 4), and the consumption of candies (Study 5), individuals mispredicted their affective reactions to these experiences and subsequently misremembered their predictions as more accurate than they actually had been. … This recall error results from people’s tendency to anchor on their current affective state when trying to recall their affective forecasts. Further, those who showed larger recall errors were less likely to learn to adjust their subsequent forecasts and reminding people of their actual forecasts enhanced learning. These results suggest that a failure to accurately recall one’s past predictions contributes to the perpetuation of forecasting errors. (more)

What we need are techs to make it very easy for record our forecasts as we go about our lives, and to review them later for accuracy.

GD Star Rating
loading...
Tagged as: ,
Trackback URL:
  • gwern

    One wonders if a site like http://predictionbook.com/ is easy enough to use?

    I enjoy using it, anyway. Predicting is a lot of fun if the predictions are interesting.

  • Randall Randall

    One thing that would make that easy, and which is coming on quickly anyway, is lifecasting. The richer a person’s lifestream, the more likely that predictions will be incidentally captured in it. Of course, this doesn’t help if they never review old predictions.

  • j r

    Something like this would be interesting because it would tell us something about forecasting as separate skill apart from the particulars of any given forecast. In other words, are people who know certain topics well better forecasters than people who know how to forecast well?

    • Jordan

      In my estimation, the best forecasters seem to be the ones that abstain from forecast until they know that certain topic very well.

      Put another way, good forecasting may simply be a matter of good self-appraisal of how confident one can be in one’s current knowledge, and knowing when not to make a prediction at all is just as important as knowing what prediction to make.

  • cournot

    Surely the tradeoff here is whether ex post accounting of your accuracy (and the ability to modify future predictions) is as valuable as being somewhat optimistic. There is some evidence that optimism (even slightly unrealistic optimism) may be better at motivating winners than excessive realism. Or at least that’s how I read work that indicates that depressed people are often more perceptive about their reality than successful people whose world views tend to be rosier than warranted.

    • Jordan

      This seems to have some real truth to it, thanks to self-fulfilling prophecies. If we tend to act upon what we see as foregone conclusions, and thus make them so, it is in our benefit to see success and happiness as a foregone conclusion in cases where it’s merely plausible.

      I’ve struggled with depression for quite a while, and one therapist I saw said (in so many words) that depression seems to be over-represented in intelligent, intellectual people. Perhaps there’s a very real harm to overcoming bias that we don’t properly concern ourselves with, since our mind, though fine with the vague truth, was never equipped to deal with the world bias-free, the same way that our body, though fine with tripping and falling, was never equipped to deal with a high-speed car accident. In this young era of candles in the dark, perhaps the early adopters of freedom from bias are essentially cruising around in an era before the invention of seat belts and air bags.

  • Jordan

    From personal experience with under-estimating myself and subsequently recording what turned out to be an excellent track record for prediction, I expect to see the same thing as the “above average effect”, wherein the top predictors will actually tend to under-estimate themselves.

    That is to say, I expect one’s metacognitive abilities to assess one’s own predictive powers tp follow exactly the same statistical patterns as one’s metacognitive abilities to assess one’s own abilities of any kind.

    Track records are to predictions what other records are to performance, and these prediction biases, I believe, are just a specific case of the more general above-average effect.

    This may seem obvious once you actually read it, but I didn’t see it explicitly mentioned yet.

  • Pingback: Overcoming Bias : Firm Inefficiency