The missions of the CCS Lab are (1) elucidate the neurocognitive mechanisms of decision-making using computational approaches and (2) develop cost-effective markers of psychiatric disorders, especially addictive disorders. To accomplish the missions, (1) we seek to develop optimal assays probing cognitive and neural mechanisms underlying addictive disorders and related phenotypes, and (2) we use cutting-edge computational and statistical methods to maximize prediction accuracy and generalizability.

Characterization of Decision-Making Systems Using Computational Approaches

Recent advancements in reinforcement learning and decision neuroscience now provide a coherent and quantitative framework to investigate decision-making processes: How we perceive reward and risk, learn from past experiences, and select appropriate actions. Abundant evidence also suggests that the brain has multiple systems for decision-making: the hard-wired Pavlovian system, which sets a strong prior on our actions when we are faced with rewards or punishments and the instrumental system, which is further divided into habitual (efficient but inflexible) and goal-directed (effortful but flexible) systems. These systems are known to interact with one another, but the exact nature and the neural mechanisms of the interaction remain unclear. To disentangle the decision-making processes, we use computational modeling, which allows for the generation of quantitative explanations, creating new and testable hypotheses, and large-scale phenotyping of individuals using statistical/computational methods. We also use functional magnetic resonance imaging (fMRI) to probe the neural substrates of underlying cognitive processes and to examine the interactions among decision-making systems at the neural level.

Computational Psychiatry: Quantifying the Decision-Making Deficits in Psychiatric Disorders.

Comorbidity among psychiatric disorders is one of the major problems in the current diagnostic system. Mood/anxiety disorders, for example, share several behavioral symptoms and underlying brain circuitries. Thus, a single framework is needed so neurocognitive mechanisms underlying psychiatric disorders can be explained in a systematic way. We use decision-making as a common language to understand normative and abnormal behavior in a quantitative way. Using computational modeling and the decision-making framework, my colleagues and I have identified specific neurocognitive processes altered in various clinical populations including patients with schizophrenia, bipolar disorder, pediatric bipolar disorder, anxiety disorders, pure stimulant and opiate addiction, chronic cannabis use , HIV, and eating disorders . This line of research will lead to the characterization of psychiatric patients based on their underlying neurocognitive processes, which might have an enormous impact on the diagnostic classification of psychiatric disorders and the development of more effective individualized treatments.

The Role of Emotions in Decision-Making and Human Ideologies

Emotions are deeply embedded in our value-based and social decision-making processes (e.g., regret avoidance; but note that a view of segregating “cognition” and “emotion” is becoming obsolete). We build computational models of counterfactual and social decision-making to quantitatively characterize the impact of emotions on our decision-making. Emotions also play a surprisingly large role in structuring our abstract beliefs and values (e.g., political views). For example, our recent work using fMRI and statistical learning algorithms shows that neural response to simple emotional images alone can accurately predict political orientation. We believe it has broad ramifications for how we understand human nature in that such simple perceptual processes are so tightly coupled to our abstract political beliefs. With our colleagues, we seek to understand how we could make rational and optimal decisions despite our physiological traits, and how we could understand and cooperate across divides in our abstract but fervent beliefs.

Development and Applications of Computational and Statistical Methods

We are profoundly interested in the development and applications of advanced computational and statistical methods to various research topics. For example, we use machine learning approaches to fMRI and clinical data to identify multivariate patterns of variables (e.g., brain activity) that are critical to out-of-sample predictions and generalizable to new samples. For fMRI analysis, we use model-based fMRI analysis with hierarchical Bayesian parameter estimation, a method that combines computational modeling, hierarchical Bayesian analysis, and general linear modeling of fMRI time series. Our results suggest that using HBA for model-based fMRI not only improves the characterization of individual differences (i.e., parameter recovery), but also the signal-to-noise ratio of fMRI signals. We also use various model comparisons methods (e.g., Bayesian Information Criterion (BIC), Watanabe Akaike Information Criterion (WAIC), generalization criterion, simulation methods) to find the best-fitting and most plausible model of human decision-making.


The CCS Lab is located in Building #16 (Rooms M309, M513, and A107) at Seoul National University (see Contact Us). Facilities include latest Mac/Windows desktops, MacBook Pros, two high-performance computing Linux servers (Dell PowerEdge T640 and T630) for parallel neuroimaging/modeling analyses, a transcranial direct current stimulation (tDCS) device, fNIRS devices, a current stimulator (DS8R), server backups, an eye-tracking system (Gazepoint GP3 HD), software for facial expression analysis (iMotions), network printers, and other computing resources. We also use and have full access to a 3T Siemens research-dedicated fMRI facility located at Seoul National University.