Fourth-Year Update

More than two-and-a-half years ago, I wrote a post basically to prove to myself and the search engine crawlers that I had survived the first year of graduate school. Let me update you: I’m seven semesters in, and still going. I feel like I’m doing okay, but that kind of thing is really hard to judge, since I’m basically just doing research full-time. This is good: grad school is (mostly) about learning to do good research and I don’t know that I’m good at doing research but I’m certainly getting better at doing research. On the other hand, research doesn’t have the same meta-structure as coursework: timeframes are less rigid, there’s no “set amount” of work to do, old stuff doesn’t just evaporate in December/May, and the feedback system (peer review) is infrequent and high-stakes. I’d like to do a post on my experience with the peer review system as a grad student, and maybe I’ll have some time to do that over the holiday.

When last we spoke, dear Google indexer, I was in the middle of my first research project, which turned into our paper on strategic classification.
Since then, I’ve written seven more papers: four are published, two are currently in the review process (and available on arXiv), and one was just accepted and will be published in February. I’m thinking I can write three more by August (when I update my CV to say I’m a fifth-year student). While setting goals in terms of numbers of papers rather than in terms of learning something new, discovering some result, or completing a project feels like an unhealthy thing to do, I also don’t think that this is an unreasonable goal.

In that intervening time, my interests have shifted from the things I described a brief 30 months ago. Looking back, I think it’s fun to be able to read that and then write what I’m about to write, and I’m looking forward to doing it again in a few years (this does seem to be the kind of excercise which becomes less valuable if done too frequently). I’ll certainly have to do it when I put together my research statement when I start looking for jobs.

Because the last time I did this, I had minimal “real” research experience, and the topics were fittingly vague and broad. I think my interests are still pretty broad relative to other people in my position. Figuring out how to make that breadth a strength rather than a liability has been a big challenge, and it will be interesting to see how that shakes out, going forward. So, without further ado, here is my new-and-improved answer to what exactly it is I do.

1. Algorithmic Game Theory

Algorithmic game theory is poorly defined in a really nice way. I think about it as the study of systems in which strategic agents and algorithmic processes interact. Sometimes, those strategic agents are algorithmic processes and so there’s a good amount of overlap between “economics” game theory and “computer science” game theory. I’m interested in settings where a strategic agent might misrepresent their features in order to get some favorable outcome from an algorithm, as in the strategic classification paper; problems arising from individuals who might strategically obscure their information to protect their own privacy, as in our private beauty contest paper; and strategic behavior in election mechanisms, including models of candidates trying to capture voteshare and in voters trying to manipulate outcomes.

2. Algorithms and Society

Designing algorithms which incorporate social norms like fairness is a research area which has sprung up in the last decade. In short, research in this line asks how we can design algorithms which satisfy a rigorous definition of fairness (such as not making more mistakes on members of Group A than we do on members of Group B), and we did this in our fair allocations paper. Along with this, we should be looking at how individuals might be harmed by algorithmic systems and investigate who pays this price, taking a more holistic view than can be captured by whatever operationalization the dataset or model uses. My paper at AIES on fair districting looks at this a little bit, and one thing that comes out of that paper is the observation that trying to draw congressional districts in Pennsylvania to be as compact as possible makes access to representation more difficult for Democrats than Republicans, despite drawing districts based on a neutral criterion being ostsensibly “fair”.

3. Computational Redistricting

Beyond the AIES paper, I’ve done a ton of work on computational and mathematical methods in the setting of redistricting. This includes coauthoring three other academic papers, contributing to research and reports, working on software, making general audience materials introducing some of these concepts, and mentoring students at the 2019 Voting Rights Data Institute. In the wake of major court cases and reforms, a renewed popular and academic interest, and in preparation for the 2021 nationwide redistricting cycle, this is a timely and impactful subject and I’m excited to continue thinking about how computer science can make meaningful contributions toward improving representation and voting rights, especially for those who have been denied such things as a result of formal and informal modes of discrimination.

Share this post on social media: Share on Facebook Tweet Share on Google+ Submit to Reddit