More power to personality

Representing personality trait-outcome associations faces two seemingly contradictory challenges. The associations are often unspecific, with similar configurations of desirable (or undesirable) trait levels tracking outcomes that differ in nature and only share a positive (or negative) valence. And yet the associations are often driven by some specific outcome-relevant aspects of the traits. I will describe a way of representing personality-outcome associations that addresses both of these challenges. The approach, based on building predictive models using the smallest personality units (items, also called nuances), generally increases the out-of-sample predictive power of personality ratings for the distinctive aspects of outcomes. Perhaps most interestingly, the approach also allows the co-variations of different outcomes to be accounted for by shared personality profiles. I will illustrate the approach by predicting Body Mass Index and several related outcomes from personality ratings and quantifying the extents to which they overlap in their personality profiles (N = 3,561).