Programming languages are abstractions that democratize access to complex ideas. Programmers wrestling with uncertain data—from sensors, machine learning, and more—would benefit from language support for probabilistic reasoning, but using such a language currently requires statistics expertise. We have developed Uncertain, a programming abstraction for probabilistic reasoning embedded in existing languages. Uncertain<T> uses hypothesis tests to make statistical decisions automatically. This automation, possible because of a careful language design that avoids problematic patterns, allows programmers to reason about probabilities at a comfortable level. In this talk I’ll describe Uncertain<T>’s design features and show its effectiveness in case studies. Finally, I’ll share some recent work to reiterate the importance of programming languages as tools for democratizing ideas.

I’m a second-year computer science PhD student in the PLSE and Sampa groups at the University of Washington, advised by Emina Torlak, Dan Grossman, and Luis Ceze. We develop new formal methods and programming languages techniques, from program synthesis frameworks to consistency models, to solve systems problems.