self- vs other-prediction of human behaviour

[kaberett]

A remarked to me the other day that People Are Terrible At Predicting Their Own Future Behaviour, But Pretty Good At Predicting Their Friends', which... he knew... because he read it online somewhere at some point.

I have therefore spent a little while this evening trying to track down sources for this, because it possibly Explains A Lot about things I find frustrating about human interaction, and also possibly gives me something solid to consider when trying to work out if, as I am often told, I really am atypically self-aware & self-reflective.

The first thing I turned up was a 1996 paper asserting that, based on two studies, self-prediction of future behaviour was more accurate than predictions made by "knowledgeable others" (here being mothers or peers), especially when predicting negative outcomes. Hmm, I thought.

But then! I found an article that is probably roughly what he was talking about, being a pop-psych review of the work of David Dunning (as in the Dunning-Kruger effect, which turned out to be a bit more complicated than that) on, well, self-assessment (and the extent to which it's a learnable skill). Dunning himself wrote up an overview for the BPS, which I will read in more detail at some point when it's not Definitely Getting On For My Bedtime, but -- having hunted down the links I'm leaving them here in case any of you lot feel moved to weigh in.

Comments

[recessional]

I think that like most things it'd depend hardcore on what predictions of what future behaviour we're talking about, and the specific details of the setup.

I think people are often crappy at admitting to themselves and to others aspects of their personality/behavioural patterns (and thus future likely behaviours) that they find embarrassing, shameful, or otherwise inadequate, particularly to listeners or whatever whom they have a reason not to want to perceive them in ways they find negative.

But what people do or don't feel ashamed/awkward admitting, or care about whether or not it's true about themselves, and how much, and why, is going to be pretty scattershot.

So I think it would be very complicated to control for all the variables prompting specific contexts of what people actually know vs what they'll self-report vs whatever.

And I think most of that landscape is significantly simplified when we're talking about OTHER people, so.

[jedusor]

haven't read the linked articles but this sounds relevant to the Fundamental Attribution Error, my favorite concept in social psych: people attribute their own behavior to circumstance and other people's behavior to disposition.