ausblenden:
Schlagwörter:
-
Zusammenfassung:
Learning to predict one’s causal impact on the world by correctly assigning credit or blame for
past outcomes to one’s previous actions is essential for decision-making. This is challenging,
since the extent to which outcomes are due to one’s own actions, rather than other people’s,
or environmental circumstances is often unknown, and must be inferred. Attribution, and its
implications for the maintenance of beliefs about one’s abilities and effects, has been
suggested as going awry in various psychiatric disorders. In particular, “negative attributional
style” is correlated with vulnerability to depression, but of questionable reliability as predictor
of depression onset [1]. Bentall [2] suggested that this is because attributional style is not
fixed, but interacts with the dynamics of beliefs about the self. Detecting such interactions
would require extended time series of both attribution and self-beliefs - something that is currently lacking. We therefore designed and administered a novel task to quantify the
relationships between attributions and beliefs.
Subjects repeatedly played a game of skill or watched 'another subject' do so (actually their
own previous trials), whilst making attributions about outcomes and estimating how skilled
they/the 'other' are. We investigated the effect of attributions on skill estimates and the effect
of reported skill on attributions – in both cases for beliefs about the self versus the ‘other’. We
also measured reaction times.
We find that participants used outcomes to update their estimates of their own and ‘other’s
skill, and they did so differently for losses attributed internally vs externally (KS test p self,
‘other’ = 0, Hedges corrected d self = 0.19, ‘other’ = 0.35 ), but not for wins. A comparison
between different models of belief updating favors accounts with different learning rates for
internal vs external attributions. Conversely, we find that for both self and ‘other’, subjects are
more likely to attribute wins, and less likely to attribute losses, internally in the case of high
estimated skill (bottom vs top quartile skill responses repeated measures t-test self p = 0.03,
Hedges d = 0.36, ‘other’ p= 0, Hedges d= 1.05 ) and losses (self p = 0.003, d=0.47, ‘other’
p=0 , d=1.09).
We also find differences in the RTs for reporting skill: subjects do so significantly faster after
wins vs losses (repeated measures t test p = 0.013, Hedges corrected d= 0.25) and after
outcomes attributed internally vs externally (repeated measures t test p =0.0066 , Hedges
corrected d=0.26 ). They are also significantly faster in reporting their own vs the ‘other’s
skill(repeated measures t test p =0.0035, Hedges corrected d= 0.54).
Our task quantifies changes in, and interactions between, attribution propensities and beliefs
about the self/’other’. We found differences in such interactions when processing wins vs
losses, consistent with positive belief maintenance plus sensitivity to sources of learning
about negative avoidable outcomes. We find similar mechanisms for self and other, but
stronger effects for the latter, which might stem from greater complexity in the more
emotionally salient self condition.
