PROJECT SUMMARY
There is an emerging consensus that using computational modeling to mathematically operationalize and
identify the drivers of behavior such as emotional response is critical to better account for individual
differences. The hope is that operationalizing the drivers of emotion (and other psychiatrically relevant
processes) will standardize how these psychiatrically relevant processes are defined and this will speed the
progress of mental health research. However, despite the promise for computational modeling to better
account for and parse the timecourse of emotional responses, there have thus far been few clinically relevant
insights. One reason for the lack of translation is that computational modeling of psychiatrically relevant
processes rarely employ ecologically meaingful paradigms. With few exceptions, there are virtually no studies
that focally measure emotional responses precisely timed to when personally relevant and meaningful events
occur. Alongside needing to measure emotional responses after personally meaningful events is the need to
measure the timecourse of such responses–which, in the case of personally meaningful events unfold over
hours, not on the timescale of seconds as is often assessed in the lab. In this proposal, we build on our initial
work using ecological momentary assessment (EMA) of positive and negative emotion in an unselected
undergraduate sample using exam grade feedback as a personally meaningful event; students in General
Chemistry care deeply about their grades in the course. We build computational models to predict the
timecourse of emotion and find that when we time-lock EMAs once individuals first see their exam grades that
both the grade prediction error (PE; the difference between the grade they report they think they will receive
[after taking the exam but before exam feedback]) and the grade itself are necessary to account for the
timecourse of the emotional response; further, the grade PE has a significantly larger effect on the timecourse
of the emotional response than the grade itself. This R21 proposal advances this work toward building a
fundamental, basic-science understanding of the drivers of emotion using valid, reliable, and
comprehensive computational models in convenience samples. We will (1) expand the set of predictors in
the model to improve our computational characterization of EMA-assessed emotional response to real-life
outcomes, including, prediction confidence, social comparison, and perceived control; (2) determine which
parameters from this computational model most strongly impact the PA and NA timecourse; and (3) test
whether individual model parameters are linked to depression and anxiety symptoms. This project will position
us for a follow-up R01 focused on how these mechanisms go awry in individuals suffering from affective
disorders.