Zenna Tavares bio photo

Zenna Tavares

I'm an researcher at Columbia University, formerly at MIT. I think about how to build computer programs that think as humans think. You can contact me on zt2297[at]columbia[dot]edu

Email Twitter Github

Who Am I

I’m an associate research scientist and Innovation Fellow at the Zuckerman Institute and Data Science Institute of Columbia University. I came to Columbia from a postdoc position working for Armando Solar Lezama and with the Computational Cognitive Science group run by Joshua Tenenbaum. I completed my Ph.D in the same group.

I am from London. I have two big brothers: Kibwe who cofounded architecture/animation studio Factory Fifteen and musician/designer/artist Gaika.

What do I do

I work to understand and mechanize the various forms of human reasoning. This primarily consists of developing algorithms and languages for (Bayesian) probabilistic modeling and causal inference. I use a variety of tools in programming languages and analysis, statistics, and (deep!) machine learning. I like to design little programming languages that synthesize these ideas, such as Omega.jl.

What do I do (with less the mumbo jumbo)

I have two goals. First, is to understand how humans reason, which is to say, how they come to derive knowledge from observing and interacting with the world. Second, I aim to construct tools that advance science, engineering and (hopefully) the humanities. Perhaps surprisingly, there is a sizable chunk of fertile ground that allows me to do both more or less simultaneously.

News

  • Our workshop Databases and AI has been accepted at Neurips 2021!
  • Our workshop Out-of-distribution generalization and adaptation in natural and artificial intelligence has been accepted at Neurips 2021!
  • I’ve joined the program committee for Languages for Inference (LAFI) @POPL 2022
  • I’ve jointed Columbia as an Innovation Fellow. Time to innovate!
  • Our paper “Causal Probabilistic Programming Without Tears” has been accepted at PROBPROG 2021.
  • Our paper “A Language for Counterfactual Generative Models” is appearing at ICML 2021.