Unit on Computational Decision Neuroscience (CDN)
Research focus
The Computational Decision Neuroscience (CDN) lab studies how the computations that subserve decision making are implemented in the brain and how this may be different in individuals with mental illness. We are especially interested in understanding how acute states of decompensation or exacerbated symptomatology, for example brought about by increased stress, incidental negative emotion, craving, and pain, have a biasing effect on decision-making.
Our group combines economic theory-inspired behavioral tasks, biosensor data, passive and active mobile device data, clinical information, model-based fMRI, and computational modeling to understand how these states influence value representations and computations that underlie intertemporal decisions, decisions under uncertainty, and higher-order reasoning about those decisions.
Our ultimate goal is to apply what we learn about the mechanisms by which decision-making can be acutely affected to develop interventions that could prevent acute negative outcomes in psychiatry.
Core ideas and methodological approaches
The two core ideas of our lab are:
- Decision-making is sensitive to internal states (emotional, physiological, clinical)
- Metacognition is important for guiding decision-making
We use four main methodological approaches: decision-making and other cognitive-processes from theory-based tasks, multimodal state characterization, computational modeling and model-based neuroimaging.
Value-based decision-making and psychiatry
The last few decades have brought about many theoretical advances on how humans process information about rewards or punishments to yield choice behavior. The resulting field of value-based decision-making lies at the intersection between economics, psychology, computer science, systems, and cognitive neuroscience. Computational models based on reinforcement learning, Bayesian inference, and economic choice provide testable hypotheses for how decision-making is implemented at the individual level.
More recently, the emerging field of computational psychiatry has begun to interrogate the utility of these modeling approaches to solve clinically relevant problems such as diagnosis, prognosis, and treatment selection for diverse mental health conditions.
Altered decision-making is characteristic of mental health disorders. For example, higher risk-taking is displayed in individuals with substance use disorder, and intolerance of uncertain decisions is common in anxiety. Despite a plethora of research surrounding these patterns and their neurological underpinnings, there is still much we need to understand to deliver helpful treatments to individuals with mental illness.
We conceptualize these behaviors as reflecting state-dependent and context-dependent value computations in the brain. In psychopathology, many maladaptive behaviors seem to be triggered by acute perturbations of an individual’s internal states (physical and emotional homeostasis). Our studies involve exploring these decision mechanisms under different affective states and stressors and relating them back to psychopathology.
Metacognition and psychiatry
Reasoning about one’s thoughts and behaviors —a process known as metacognition— is a very important ability. In psychiatry, metacognition is often evaluated through the individual’s level of introspection, emotional and interoceptive awareness, and self-efficacy. Individual differences in this ability map to different mental health disorders. For example, individuals with depression and obsessive-compulsive disorder often display a distorted perception of their performance, judging it to be worse than it really is. In addition, metacognition is important to psychiatry because many psychotherapeutic approaches leverage and enhance metacognitive ability.
Our lab seeks to understand how individual differences in metacognitive ability may provide resilience or increase vulnerability to acute stressor-induced maladaptive behavior. We employ computational modeling to explore the mechanism that gives rise to metacognitive judgements and relate model-derived parameters as metrics of metacognitive ability.