In today's political climate, that word means a lot to us. And yet what is it, exactly? Is it a political tactic? A method of voter suppression? Surely yes, but what does that mean for affected communities? When a politician seeks to gerrymander a state, they do so by making one of quintillions of possible cuts one can make to partition a state into equal-sized pieces. So how do they choose? And how should we?
Researchers at the Metric Geometry and Gerrymandering Group have begun to understand the scope of the space of all of these possible plans. It is a huge landscape, with more orientations for districts than we could ever imagine. Yet through Markov Chain Monte Carlo (MCMC) sampling, we can at least start to explore this wildnerness.
The principle, in its intuition, is fairly simple. Given a districting plan, we propose a modification. If that modification passes muster, we have a new plan. These changes are random, but carefully calculated such that we have a real chance at obtaining a representative sample of this space of plans.
So what do these plans look like? How do they affect people living in the state? Research doesn't tend to look at this - the primary aim is to calculate summaries in order to detect instances of gerrymandering. But we have taken a different approach, looking to show for the first time what these entirely statistically generated plans look like. In how many ways can we define "community"? Soon, we will find out.
Click on Iowa, Pennsylvania, or Georgia to begin.
Click here to read the paper associated with this project. Special thanks should go to Arvind Satyanarayan, professor of 6.894 for this term, for advice on how to proceed given the extremely tight schedule and the nature of a single-person team. Also thanks to Soya Park, the teaching assistant for this term, for her guidance. Finally thanks should be given to Daryl Deford, Postdoctoral Associate at MGGG and the supervisor to the author, and Justin Solomon, professor and PI for MGGG.