About her PhD project

Learning is a fundamental mechanism that allows animals to adapt to changing environments by using past experience to guide future choices. In humans, this capacity often takes the form of goal-directed learning, where actions are selected to achieve desired outcomes and reduce uncertainty. My work focuses on understanding how information-seeking behavior arises and how it differs from traditional reward-driven learning. While standard reinforcement learning (RL) describes how agents optimize for external rewards, many behaviors are driven by an intrinsic motivation to gain information about the environment.

To explore these dynamics, I combine recurrent neural network (RNN) modeling with human behavioral and imaging data. This approach allows me to contrast reward-driven and information-driven learning in terms of both patterns of actions and the underlying neural activity. By bridging normative frameworks with artificial neural networks, my project aims to identify the computational mechanisms by which biological and artificial agents integrate action–outcome information to form internal models. Ultimately, I seek to determine whether intrinsic and extrinsic forms of motivation rely on shared or distinct neural computations, clarifying how abstract quantities like information gain are represented in network dynamics.