Exploring Computational Modeling

By Frances Grace Hart
September 3, 2024

On July 27, 2024, more than 60 scientists ranging incareer stages and from a variety of disciplines convened at the University of Pennsylvania Stephen A. Levin Building for SRNDNA’s one-day intensive Computational Modeling Workshop.

Dr. Debbie Yee (Brown University) opened the day with an introduction to cognitive computational modeling, including why these models are useful and various strategies for implementing them.

Following this introduction, Dr. Angela Radulescu (Mt. Sinai Center for Computational Psychiatry) and Alana Jaski (Brown University) acquainted attendees with reinforcement learning, covering such topics as the Markov decision process; model-based vs. model-free reinforcement learning; artificial neural networks; and the potential for virtual reality technology to aid computational research.

Next, Dr. Bob Wilson (Georgia Institute of Technology) discussed drift diffusion models (DDMs) of decision formation, including topics such as moving dot tasks; the interrogation paradigm; signal-to-noise ratios; stochastic differential equations; and value based DDMs.

The final lecture was presented by Dr. Gregory Samanez-Larkin (Duke University), a founder of SRNDNA. Dr. Samanez-Larkin’s presentation, titled “The Promise of Computational Models to Decision Neuroscience of Aging,” recounted how SRNDNA was inspired by two significant societal shifts in the early 2000s: the rise of shared decision-making in healthcare and the transition from pensions to 401(k) retirement plans. Years later, SRNDNA continues to connect researchers committed to unraveling the complexities of decision-making. Dr. Samanez-Larkin concluded by highlighting the potential for computational modeling to not only satiate scientific curiosity, but also offer solutions to society’s most pressing issues.

The day closed with two parallel hands-on tutorials on reinforcement learning and drift diffusion models, followed by a lively reception. 

In summary, the workshop provided a friendly and accessible introduction to rigorous computational methods for decision neuroscience, aging, and related topics. A post-event survey revealed overall positive appraisals, with attendees highlighting the collaborative atmosphere and cultivation of new professional relationships. Attendees appreciated the opportunity to explore promising avenues of scientific inquiry in a supportive setting. Many remarked that computational modeling, while compelling, often seemed daunting due to its perceived steep learning curve. However, workshops like this lower the barrier to entry, providing a welcoming environment for those without prior or extensive experience to gain the confidence, foundational knowledge, and skills to implement these methods in their own work.

SRNDNA looks forward to facilitating more opportunities like this in the future.