Zack Boner

I am a Ph.D. Student in Computer Science at Duke University, advised by Dr. Cynthia Rudin. I research interpretable machine learning with the objective of making AI more useful for real world domain experts. I aim to understand why simple models perform as well as black box models in a wide variety of tabular data domains. Towards this goal, I am investigating the simplifying effect of noise for several hypothesis spaces, especially sparse decision trees. I also study the underlying causes of the Rashomon Effect, where there are many approximately-equally-good models for real world data. Finally, I seek to understand the role of missing data in exacerbating the Rashomon Effect.