ENGS 105.1 - Principles of Causality

Helmets increase head injuries. Faster drivers arrive later. Hospital patients are less likely to have cancer when their bones are broken. But don’t throw out your helmet, break your leg, or invest in transportation inefficiency just yet... Causality is the beacon of science and the foundation of policy, but causal relationships can be lost in a labyrinth of correlation. Instead of caveating correlative studies, this course establishes the principles on which scientists and engineers can answer true causal questions. At the core of this pursuit is experimentation. The class will begin by building a formal understanding of randomized control trials. We will generalize these principles to build causal inference methods for observational (i.e., non-experimental) data. As we dig deeper into these methods, we will discover that causal structure plays a key role in their correctness. The final portion of this class will focus on the task of learning these causal structures using observational data and/or efficient experimental design.

ENGS 20 or COSC 10, and ENGS 27 or ENGS 93; or permission of the instructor.

Coding assignments will be done in Python to take advantage of state-of-the-art libraries. Past experience with Python will not be necessary to take the course. Students who have earned credit for 85.14 or ENGG 199.19 may not enroll in this course.