Faculty Profile: Yael Niv
Assistant professor, Department of Psychology and the Princeton Neuroscience Institute
Yael Niv wanted to work in a place where she could share intellectual interests with a highly collaborative group. Princeton's active cluster of neuroscientists and cognitive psychologists, she says, has given her just what she was looking for.
Niv was also drawn by the idea she would be working with colleagues who possessed a common language of thought -- those who use computational models to study behavior and the brain.
"It is great, and rather rare, to find such an environment," said Niv, who worked as a postdoctoral fellow with Jonathan Cohen, co-director of the Princeton Neuroscience Institute, before joining the faculty in 2008. She earned her Ph.D. from the Hebrew University of Jerusalem in 2007.
Niv is seeking to understand reinforcement learning, a term that describes how humans come to choose the right actions in the face of rewards and punishments. Unlike the supervised learning of formal environments where a learner is told what are the "right" and "wrong" answers, reinforcement learning covers the arena of everyday experience where the environment provides the learner with less obvious feedback. Rewards and punishments -- from smiles and frowns to money or the lack of it -- shape the countless physical behaviors and social interactions that constitute reinforcement learning.
"We learn to behave in the world from what the world does back to us," Niv said. "People are really good at this kind of learning, but we are not sure how the brain does it."
Niv works at developing computational models that show the optimal solutions to a given problem and compares those solutions to what the brain actually does. Looking at simple tasks, she has found that the brain can quickly achieve the best solution.
In tests with more elaborate structures containing all sorts of hidden variables, it is harder to derive optimal solutions. And it is difficult to understand how the brain processes these choices. Niv is working on devising models that will lead her to a better understanding.
Learning takes place, she explains, when people make what are known as "prediction errors." That is, they expect one thing to happen, but another event occurs instead. Either they get more of a reward than they expected, a "positive" prediction error, or they receive no reward when they expected something, a "negative" prediction error. Levels of the brain chemical dopamine are altered when prediction errors occur and scientists like Niv can use imaging techniques to observe learning in action.
The work could shed light on problems like addiction, which results when people engage in flawed decision-making and opt for short-term rewards over long-term negative consequences. "How can things go wrong so badly?" Niv wonders. "The goal is to understand the system in order to be able to help with the disordered system."