Research Overview

I am on the job market this year for faculty or other research-oriented roles in computer science, information science, and related fields! Please get in touch if you know of any relevant opportunities!

Online communities are an integral part of the daily life of billions of people, with the average member spending more than two hours online daily. While online communities can bring people together, they also have been associated with harms such as misinformation, polarization, and poor mental health. My work mitigates these harms and empowers online communities to be their best through better governance.

My work makes online communities better through three research activities: I characterize communities’ values and how they vary between different communities and community members. I assess current governance practices across thousands of communities to identify the most promising strategies. I deploy good governance through novel systems and partnerships with communities.

I combine and advance methods in computational social science to produce actionable insights that are grounded in an understanding of communities’ varied needs. I use causal inference methods and conduct experiments in real communities to ensure the robustness of my results.

My experience as a longtime moderator grounds my research and drives impact. As a result of my research, I was invited to join the Reddit Moderator Council, where I help guide Reddit’s product decisions. I also am a Computational Science Fellow at Reddit, which has deployed my methods used my results to develop new moderator tools.

CHARACTERIZING COMMUNITY VALUES

Numerous researchers study how to make online communities ‘better,’ however, community members’ values for their own communities are not well understood. In order to inform the rest of my work, I conducted the largest and most comprehensive surveys of online community members’ values to date. I contributed a taxonomy of community values, and used this taxonomy to measure how values very both between and within different communities. My taxonomy has been used by other researchers as well as Reddit itself.

ASSESSING GOVERNANCE AND OUTCOMES

Community moderators must make countless decisions about how to govern their communities, yet mostly lack a data driven understanding of which governance practices are most appropriate. My work develops scalable methods to measure governance and outcomes across hundreds of thousands of online communities to assess which practices are most promising, including moderator team composition, rules, and their enforcement.

Effectively measuring community outcomes is challenging. I developed a new method to detect and classify community members’ publicly stated perceptions of their moderators, and developed novel methods to assess the political bias and factualness of content shared in communities. This work won an Outstanding Analysis award at ICWSM 2021. Putting governance and outcomes together, I identify which governance practices are most associated with positive attitudes towards moderators and community members’ own preferences measured using my survey data. My work shows that existing community governance affordances such as up- and down-voting are can meaningfully reduce the visibility of problematic content, and that moderators are best received when those moderators are active, engaged, and enforce rules appropriately.

DEPLOYING GOOD GOVERNANCE

An understanding of good governance is meaningless if it cannot be deployed in real online communities. I partner with communities and platforms to deploy my research through two avenues: through systems building and deploying interventions as part of randomized controlled trials (RCTs), as well as through collaborations with the Reddit Moderator Council and other groups to disseminate best practices and translate my insights into impact. In ongoing work, I am conducting an RCT in a community with 1.7 million members to evaluate an LLM tool which challenges contributors to write higher quality comments. When active experimentation is not possible, I have employed causal inference methods, such as SHERBERT, which I developed to exceed the performance of every other commonly used method on many tasks.

I leverage my decade of experience as a moderator to bridge the ‘implementation gap’ between researchers and moderators. By sharing best practices from my results with the Reddit Moderator Council, moderator recruiting strategies informed by my work have been adopted by over 87 communities with more than 115.4 million combined members, and Reddit itself has begun using my methods to quantify community values in their own metrics. I also champion better data access for researchers, including a CSCW SIG and an ICWSM invited talk.