gwern comments on We Change Our Minds Less Often Than We Think - Less Wrong

39 Post author: Eliezer_Yudkowsky 03 October 2007 06:14PM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (115)

Sort By: Old

You are viewing a single comment's thread. Show more comments above.

Comment author: gwern 10 January 2012 07:46:30PM *  2 points [-]

instead I mean that when the risk is death of oneself or others then the risk is so high as to outweigh most, if not all rewards.

And how does one judge when the reward is outweighed? Hm, that wouldn't be subjective would it...?

If one gets better at balancing those things during development, ie. growing up, that reveals one has a lack in balancing those things, which are to do with judgement - and poor judgement is one form of overconfidence.

Never reason from a price change. Are teenagers overconfident about how much daytime sleep they need, and their circadian rhythm shifts due to them getting "better at balancing impulse, desire, goals, self-interest, rules, ethics, and even altruism, generating behavior that is more complex and, sometimes at least, more sensible"?

Dobbs continues "If offered an extra reward, however, teens showed they could push those executive regions to work harder, improving their scores."

This draws out a reason why they underestimate risk - because if the reward isn't "extra" or perceived as higher to the teen, then they likely won't push those executive regions to work harder, as is revealed in the articles previous paragraph, which I haven't quoted.

This practically demonstrates the opposite: that the reward is the important part! What rewards do experimenters usually offer? Pretty lousy ones. Why is it surprising that teens might not work as hard as conscientious saps - I mean, mature adults. (I am reminded of how money rewards can improve IQ test performance by half a standard deviation or so.) This is like the usual criticism of the PISA test scores: of course Americans will underperform, since not a single one of the test-takers cares about what score they get. Incentives matter, which is what I've been saying all along.

This can actually be interpreted either way, Steinberg chooses a higher reward, whilst one can just as reasonably choose an underestimation of risk – a rationale for which is alluded to in article is that social feelings/thoughts are more sensitive for the teenagers, hence the brain is more focussed on social cognition than risk estimation (as the risk estimation processes require more effort), hence less brain cycles on risk estimation – hence risk underestimation.

Wow. So your explanation for a clear-cut reward link is... they get distracted and can't estimate risk as accurately.