289 — Rational Temporal Predictions Can Underlie Apparent Failures to Delay Gratification

I wrote recently about a recent revisit of the marshmallow test, and — as I often find about recent brain news, Adam Calhoun had Words:

On the recent 'failed replication' of the Marshmallow Test: control for income & effect of waiting disappears

I'm surprised nobody mentioned this paper (McGuire and Kable 2013) showing failure to wait can be rational if you don't trust your environment
https://t.co/g9CPpsHOb2 pic.twitter.com/hysgXJM7BP

— Adam J Calhoun (@neuroecology) June 5, 2018

So of course, that’s today’s paper.

The first thing that becomes immediately apparent is that children do not behave the same in this study as in the original marshmallow test study; nor do they behave like we would expect rational agents — ideal, logical economists — to behave.

A rational agent would look at a small reward, and compare it with a larger reward plus the cost of waiting or the larger reward…and immediately made a decision whether to take the small reward or wait for the larger one.

But that’s not what kids do; kids don’t know economics any more than dogs know trigonometry.

Instead, the children opted for — or against — the wait for the larger candy reward based upon if they believed that the test administrator would return in a timely manner; if the children believed this, it was worth the wait. If they didn’t, or more importantly, if they began the test believing this but their belief wavered, they would “abort” the test and opt for the smaller reward.

This is an economic paradox; in short, it’s analogous to assuming that a wait-time INCREASES the longer you wait (a sort of German Tank Problem). And this is why there is so much interest in the world of delayed gratification — the same mechanism that told these four-year-olds to abort their wait before the test administrator returned with the larger reward, and the same mechanism that makes it so hard to quit smoking.