Katrina Wolters L&S Social Sciences
Behavioral signatures of HMM-revealed perceptual modes in dermatology
My project uses advanced computational modeling techniques, such as Hidden Markov Models (HMMs) and Google’s Derm Foundation ML model, to better understand fluctuations and biases in humans’ perceptual attention while making skin lesion malignancy judgments. We aim to use state-of-the-art machine learning methods and models to explore fluctuations in accuracy, perceptual biases, and response behavior among participants when making dermatological judgments. My project will be among the first to explore these differences using HMMs to infer contrasting perceptual states as they naturally occur within a single, continuous dermatological judgment task. I will also be able to evaluate both within-subject and between-subject differences in these behaviors, as HMMs allow us to track fluctuations in behavior and perceptual bias across time, rather than treating each state as an isolated event in a given task. This approach allows us to capture a more dynamic and nuanced picture of perceptual attention and bias, revealing how subtle fluctuations relate to the way we see and interact with the world, with the ultimate goal of improving clinician accuracy.
Message To Sponsor
Thank you so much to Leadership for sponsoring my SURF project! I’m incredibly grateful to have this opportunity to explore new questions in the field of vision science, and your generous contribution made it possible for me to pursue this project in depth. This experience allowed me to dive into fascinating research, and also helped me to gain invaluable skills and knowledge for my future work. I’m truly grateful for your support in making this opportunity possible!