I am interested in designing high-performance machine learning methods that make sense to humans. Here is a short writeup about why I care.
My current focus is building interpretability method for already-trained models (e.g., high performance neural networks). In particular, I believe the language of explanations should include higher-level, human-friendly concepts.
Previously, I built interpretable latent variable models (featured at Talking Machines, and MIT news) and creating structured Bayesian models of human decisions. I have applied these ideas to data from various domains: computer programming education, autism spectrum discorder data, recipes, disease data, 15 years of crime data from the city of Cambridge, human dialogue data from the AMI meeting corpus, and text-based chat data during disaster response. I graduated with a PhD from CSAIL, MIT.
I gave a tutorial on Interpretable machine learning at ICML 2017, slides are here .