The Niv Lab focuses on several distinct and inter-related lines of research related to understanding the computational basis of learning and decision making in the brain. We use behavior – and sometimes neural data – to constrain computational models of how experience changes future decisions and choices (i.e., learning), and we apply this to understanding mental health conditions and tailoring their treatment. 

One major theme of our research is understanding how animals and humans learn useful task representations, effectively grouping experience into latent causes or clusters that allow fast and generalizable learning. A second focus is on computational psychiatry/psychotherapy. Here we use computational models of learning and decision making to study mental health conditions, e.g., related to mood and anxiety, and transdiagnostic symptoms such as rumination. Guided by computational models, we develop behavioral assays that cut across diagnoses in order to understand the underlying axes of mental health, and to predict treatment response. 

As an integral part of our research, we run experiments on human participants. These mostly consist of behavioral experiments, many of which are conducted on large online samples. Occasionally, we conduct fMRI experiments when we have a hypothesis regarding brain mechanisms that can be answered better by looking into the brain. We also run studies on patient populations (through the Rutgers-Princeton Center for Computational Cognitive Neuro-Psychiatry). 

Our primary modeling frameworks are reinforcement learning and (Bayesian) latent-cause inference, and we make heavy use of (hierarchical) model-fitting and statistical model comparison. We also use machine-learning techniques of multivariate pattern analysis and representation similarity analysis of fMRI data, as well as classification and decoding techniques in neural data analysis. Computational modeling is not only a tool we use to analyze data; it is really how we think about function in the brain, so everything we do involves modeling in a recurrent feedback loop between experimental data and models. 

We also have other interests that are more or less tied to our main focus areas. These include investigating the flexibility of the goals we set ourselves, and, in collaboration with other labs, taking a developmental perspective on understanding mental health concerns. In general, our focus shifts slightly over time, through the interests of lab members, as we encourage lab members to focus their research on the questions they are most passionate about (of course, we accept to the lab those whose interests align with our current interests). This means that while we have some overarching themes to our work, projects may not always feel strongly interconnected. In any case, we encourage crosstalk between projects, collaboration within the lab and outside it, and we support individual exploration and growth. In an effort to uncover the “hidden curriculum” of academic work, we have a detailed lab manual that expresses our co-created values, expectations, and lab rules.

Grants supporting our research: 

For the auditorily-inclined:

An informal podcast about what the lab does (starts around minute 9)

In the press:

• Achtung Baby: Why and How We Make Decisions, SfN 2017 Annual Meeting Blogs
• Decoding the Brain, Princeton Alumni Weekly
• Learning What to Learn, Innovation Magazine
• Rewriting History: Yael Niv studies how the brain learns from experience, and how it can unlearn fears, Princeton Alumni Weekly
• This Week in Machine Learning & AI Podcast from NIPS 2017. Sketchnotes of the podcast, by Shirin of Shirin’s Playground: