Neurocomputational mechanisms of impaired social decision-making in borderline personality

Borderline Personality Disorder (BPD) is a debilitating condition characterized by affective instability,
interpersonal dysfunction, and impulsive and self-harming behaviors. Individuals with BPD often make
disadvantageous decisions in response to negative interpersonal events, yet little is known about decision
processes that precipitate maladaptive behaviors. The past two decades of neuroscience research has
provided overwhelming evidence that decision-making can be understood in terms of Pavlovian and goal-
directed computational systems that are implemented in specific cortical-striatal-limbic circuits. Supported by
our preliminary data, we propose that in BPD, the cingulo-opercular network’s role in goal-directed learning is
vulnerable to disruptions by social-emotional cues that exert Pavlovian influences on decision-making.
Although BPD has historically been diagnosed in adults, symptoms often emerge in adolescence and their
severity may peak in early adulthood. The maturation of the cingulo-opercular network from adolescence to
early adulthood underlies developmental improvements in the integration of motivationally salient cues with
goal-directed behavior. We will test the hypothesis that in BPD, both approach- and avoidance-related
Pavlovian computations dominate the cingulo-opercular network via the phylogenetically old pathway from the
central nucleus of the amygdala to the nucleus accumbens core, which underpins emotion-driven Pavlovian
responses. The proposed case-control study will characterize abnormalities in Pavlovian and goal-directed
decision-making in 49 young adults with BPD symptoms compared to 49 matched individuals with social
anxiety disorder and 49 healthy controls. Studying these processes in early adulthood is essential because
BPD symptoms change rapidly during this period, which may reflect neurodevelopmental maturation of
emotion- and decision-related circuits. At the behavioral level, we will characterize participants using a decision
battery and corresponding hierarchical Bayesian reinforcement learning (RL) models that span social and
nonsocial contexts (Aim 1). We will link decision signals, particularly the effects of social cues on goal-directed
learning, with their neurocomputational correlates using Bayesian RL models and model-based fMRI analyses
(Aim 2). Finally, to characterize separable circuits involved in maladaptive Pavlovian computations in BPD, we
will conduct a high-resolution resting-state fMRI study of the integration of the cingulo-opercular network with
specific limbic and striatal regions (Aim 3). Altogether, our computational psychiatry approach builds on the
unique strengths of our investigative team in BPD and neurodevelopment (Hallquist), Bayesian methodology
(Oravecz), and decision neuroscience (Hallquist, Dombrovski). This work aligns well with the NIMH Strategic
Plan for Research objectives to describe the neural circuits underlying mental illness (Strategy 1.1) and to
identify biomarkers and behavioral indicators that predict change in illness (Strategy 2.2).