Quantitative Neuroeconomic Modeling of Intrinsic Motivation during Memory Performance in Adolescents at Risk for Psychosis

Students

2021
Engineering and Applied Sciences

Project Summary

Under Dr. Daniel Wolf’s mentorship this summer, I had the opportunity to apply my background in math and computer science to explore the realms of neuroscience, psychology, psychiatry and behavioral neuroscience through a quantitative lens. The lab’s primary focus is in neuroeconomics - an interdisciplinary field of understanding the human decision making process in which the brain rationalizes external circumstances and weight possible actions using knowledge of past experiences to make optimal choices which maximize economic rewards (both extrinsic and intrinsic). Modeling these processes offers insight into an individual’s reward response. This helps to address an unmet therapeutic need in the psychiatric community by elucidating motivational deficits, which may help serve as indicators for adolescents at risk of psychotic disorders.

The project I was specifically involved in evaluated patients’ confidence levels through  successive trials of a memory performance task. The confidence ratings of the patients undergoing the experimental task offered a data point from which to extrapolate further information such as the patient’s expected intrinsic and extrinsic reward from the task as well as the patient’s rate of learning. In order to perform this quantitative analysis, I implemented code for a reinforcement learning model in MATLAB. Reinforcement learning is a type of machine learning premised on the notion that learning occurs solely when the outcome of an individual’s action differs from the individual’s expectation. This prediction error causes the individual to update her expectation for future reference accordingly as modeled by specific mathematical equations. This approach is commonly used in behavioral psychology. Once having implemented this learning model, I wrote code to regress the patient behavioral data against the reinforcement learning model. This allowed me to extrapolate information about the patients’ learning rates, initialization points, and other parameters -- more generally this enables the lab to quantitatively assess how the performance of healthy patients differs from that of unhealthy patients.

My overall experience allowed me to gain invaluable skills and insight into many interesting fields of study. As a computer science major, my work over the summer helped me gain more experience with algorithm development and modular coding. In the process, I read many papers and textbooks to learn more about the field of neuroeconomics and the biological basis of decision-making. Reinforcement learning -- and machine learning, more generally -- has many far-reaching applications beyond the medical field. Moreover, I was able to recognize the paramount importance of statistics and mathematical analysis in drawing significant data-based conclusions: even relatively simple mathematical approaches can often allow scientists to extract meaningful signals from noisy empirical data. Overall, I learned a ton about many prolific fields of study as well as the scientific research process itself. I look forward to using my experiences and knowledge gained to continue to explore the various applications of neuroeconomics.