Originally posted on CFAR’s blog.

Our predictions fail in predictably bad ways. Picture the last time you were getting ready to leave the house and you were asked how much longer you would take. And when was the last time you actually left by then? You might be better at this than I am, but I have struggled with forecasting how long it will take. My dad used to account for this, and pretended we needed to leave 15 minutes before we actually did. Without knowing, my dad was accounting for Hofstadter’s Law:

It always takes longer than you expect, even when you take into account Hofstadter’s Law.

— Douglas Hofstadter, Gödel, Escher, Bach: An Eternal Golden Braid

What does it matter if we haven’t left on time? That might well be sometimes inconsequential. But leaving on time is easy compared to plenty of other predictions you need to make, that do matter. How much will I enjoy my new job? Should I move to sunny California? Should I buy that lottery ticket?

Psychological science has made a lot of progress in determining where we systematically make inaccurate predictions. So where do we stumble?

1. Simulations are unrepresentative
We tend extrapolate from tiny bits of data that are unrepresentative: We characterize an experience by its worst elements, and predict on that basis. For example, when people are asked to imagine missing a train in the future, they tend to remember their worst train-missing experience rather than their typical train-missing experience. They picture missing their train before that important meeting they had and being lambasted by their boss, not every other time they just caught the next one in a few minutes. The converse is also true: We idealize ‘good’ experiences by predicting, for instance, that moving to sunny California would change our lives, when, really, weather is only a small part of what makes us happy.
Everyone remembers their best day, their worst day, and their yesterday.

2. Simulations are essentialized
We often only remember what we see as the essential pieces of an experience, not everything surrounding it. Participants asked about a recent theater experience, they only remembering the theater as being what’s on stage, not the cold, the parking, the noise, and the interlude. Because accurate predictions are contingent on accurate memories, it’s hard to get the final step right if the first one is in the wrong direction. Simulations of events omit inessential features, and people tend to predict that good events will be better and bad events will be worse than they actually turn out to be.

3. Simulations are abbreviated
Only thinking about a small portion of the experience can also get us into trouble. How happy would you be if you were to win the lottery? Ecstatic, right? What about 6 months down the track? Still incredibly happy, I’m guessing? When Daniel Gilbert took a sample of people who had this happen to them, he found their happiness to be back at baseline within a few months, with a minor blip in between [1]. So why is this? Turns out when you’re predicting, you only think of day 1, not day 277 and we humans are remarkably good at adapting to our circumstances.

4. Simulations are decontextualized
When the context in which you make your prediction differs from the context of the event, problems arise. For instance, Gilbert, Gill, and Wilson [2] found that hungry people predict they’d like spaghetti for breakfast, while sated people predict they’d dislike it for dinner. As you might have guessed, and running thematically with the rest of this article, they weren’t very good at getting it right. Hungry participants overrated their enjoyment of a spaghetti breakfast, and full participants underrated their enjoyment of a spaghetti dinner.

So, what can we do about them?
Daniel Gilbert and Timothy Wilson, academic psychologists who research this area, recommend we pretend the event is happening tomorrow and thinking through events hour per hour. So a nebulous desire of moving to California turns into thinking what will be different in every hour of waking up, getting ready, commuting, working, eating etc. And the difference might not amount to that much. This, he reasons, assists in making our predictions more concrete. Similarly, on a smaller scale, we can break everything down into its constituent steps. Think through each step of what you’ll be doing, and try to avoid relying on your brain’s associations with whole events. In an attempt to make decisions when you might be feeling differently, consider specific factors that might be altering your decision and account for them, or try to defer decision to when you’re in a similar state. We should try, Gilbert recommends, shopping when you’re not feeling so hungry, or pre-committing to doing that study before you change your mind, for instance.

Can you predict how successful you’ll be able to influence your predictions?

Notes
[1]. Gilbert, D. T., et al. (2002). “The future is now: Temporal correction in affective forecasting.” Organizational Behavior and Human Decision Processes 88(1): 430-444. http://www.sciencedirect.com/science/article/pii/S0749597801929826

[2] Gilbert, D. T. and T. D. Wilson (2007). “Prospection: Experiencing the future.” Science 317(5843): 1351-1354. http://www.sciencemag.org/content/317/5843/1351.full