Hello there, I’m Divya!

I’m a graduate student at New York University majoring in Computer Science.

At NYU, I am the lab manager and junior Laboratory Associate jointly under the Computation and Decision-Making Lab (PI: Dr. Mark Ho) and AI Thought Partner Lab (PI: Dr. Ilia Sucholutsky) where I am exploring decision-making in humans and AI systems and the circumstances under which people will trust AI systems to make decisions on their behalf. My work delves into the intersection of learner modeling and diagnostic inference, focusing on how observers both human and algorithmic, can accurately reconstruct the latent cognitive processes that drive observable behavior. I am particularly interested in the ‘reverse-engineering’ of student logic, that is identifying the specific mental models, procedural shortcuts, and systematic misconceptions that lead to a given outcome.

In my Master’s thesis under Professor Todd Gureckis at the Computation and Cognition Lab, I am studying how humans integrate social testimony with direct experiential learning, particularly when these two sources of information are presented at different points in time or come into conflict.

In my work, I explore how the timing, frequency, and reliability of testimony influence an individual’s ability to form accurate mental models of the world. Using computational modeling and behavioral experiments, I aim to characterize how agents weigh social information (testimony) against firsthand experience, and how these weightings shift dynamically as new data is encountered.

I am particularly interested in representations in humans and artificial intelligence systems which enable efficient language learning, belief updating, and decision making. I have been drawn to how using human learning principles can improve artificial intelligence systems. My goal as a researcher is to understand fundamental issues about how people actually learn, corporate and make decisions and use this to contribute to the growing interdisciplinary effort that connects cognitive modelling and machine learning.

My bachelor’s thesis was advised by Professor Joannes Sam Mertens from The University of Catania where I designed deep learning pipelines using CNN and LSTM autoencoders to capture temporal driving patterns and behavioral signatures from raw sensor data, enabling robust profiling of individual drivers across identical driving conditions.

I’m on a personal quest to find the best hot fudge out there, because priorities. So far, the top spot goes to the hot fudge from What’s the Scoop, but the search (happily) continues.

Key Interests : Machine Learning, Cognitive Modelling, AI