How to Safeguard Researchers Who Study Social Media

To analyze social media data, researchers at NYU’s Center for Social Media and Politics (CSMaP) will label a sample of tweets, Facebook posts, or YouTube videos in a given dataset. The process, called coding, is standard across the social sciences.

Researchers working on projects based at the lab will code thousands of pieces of content — into categories such as “healthcare,” “the economy,” “positive” or “negative” toward Joe Biden, or “not relevant” — before feeding the results into machine learning classifiers to train the underlying algorithms.

This coding work can be challenging: Though we have guidelines in place to flag extreme posts, our researchers could still come across hateful or offensive content while doing their work. Potentially harmful material can even be the subject of our research — from our study investigating who is more likely to share fake news online, to our work tracking hate speech on Twitter during and after Donald Trump’s 2016 U.S. presidential campaign.

Even a random sample of relatively benign social media posts can skew negative, especially during times of crisis: “It’s a constant barrage,” said CSMaP Research Scientist Megan Brown, who regularly analyzes YouTube content. “I’ll see that tons of people died from COVID and even the tiger at the zoo has COVID. It’s like, ‘Great, now it’s time for dinner.’”

Exposure to this kind of content is a byproduct of our mission, which is to study how social media use affects politics, and use social media data to study broader political attitudes and behavior. The Center houses the Social Media and Political Participation (SMaPP) Lab, where we develop new and innovative ways to use social media data to carry out our mission. The data we analyze contain content from across the social media ecosystem; inevitably, they at times capture the kind of harmful material that is present on social media platforms and is very difficult to moderate.

Only a few of our research projects involve explicitly sensitive topics, but we knew we still needed a plan to mitigate the worst effects of this exposure. Wellness programming for a computational social science lab is a relatively new concept, so we turned to advice from journalists and other experts who view the same user-generated content as our researchers: A major insight we adopted was the importance of making this a workplace culture and management issue.

It’s why we’re committing to a set of best practices and values to create a healthy and productive research environment in our lab. The following plan, which we developed by consulting other researchers and drawing on relevant literature, is a work-in-progress that we’re constantly seeking feedback on. By sharing our learnings with the research community, we hope to foreground the topic of wellness and promote building researcher resilience as an important organizational value.

Recognize a researcher’s individual agency

We’re committed to fostering the kind of culture that empowers our researchers to achieve mental wellness on their own terms. We believe there isn’t a one-size-fits-all model: The way these challenges might affect a researcher depends on their background, experiences, level of wellness before starting a project, and resilience while doing the work.

Our researchers work in a variety of contexts at the lab: Projects focusing on hate speech or radicalization could lead to what clinical psychologists call “vicarious trauma,” which describes the harm caused by repeated exposure to disturbing images or stories. On the other hand, more routine tasks (like coding political tweets) could still cause emotional or psychological distress due to the unfiltered nature of user-generated social media content.

It’s why we make it clear that researchers can leave a project at any time without consequences. Co-director Joshua Tucker, for example, told Brown that she should prioritize her mental wellness while working on the YouTube project, and that she could take breaks from it whenever she needed to. “It was just really nice that I knew I was supported by my colleagues — that I could step away from it at times, and that was OK,” she said.

Provide training and establish clear guidelines around wellness

We crafted a formal Wellness Statement to share with researchers and research assistants before starting a project.

We also host training meetings to introduce the statement to researchers and point them to resources offered by the lab and the broader NYU community.

We warn researchers of the pitfalls of social media data analysis, including possible exposure to violent images and other explicit or disturbing material. And we are creating wellness routines for researchers to refer to while working.

Check in regularly and be mindful of varying needs

Researchers’ roles at the lab will vary along with the level of support they require: A group of research assistants coding 500 tweets per week this semester may require more attention than a pair of researchers writing a brief literature review. We regularly check in with project team members at all levels, and we are working to improve that process in Fall 2020 using lessons learned from the past two semesters.

We also remind researchers to monitor their routines and behavior. This means asking researchers to reflect on their diet, exercise regimen, sleep cycle, stress level, and recreational activities. We encourage those who notice changes from their baseline levels to reach out to us so we can adjust their workload and direct them to university resources such as the NYU Wellness Exchange.

Build a strong organizational support system

Research shows a trusted support system can boost mental wellness and improve performance.

We encourage our researchers to build strong social ties both within the lab and outside of it. We do so by planning social events and creating opportunities for relationships to develop organically, such as encouraging lab members to eat lunch together and organizing group Zoom meetings during times of social distancing.

Moving forward, we plan to combine the power of each researcher’s individual agency with the collective support structures we’re building within the lab. Our goal is to create an environment in which researchers — and ultimately the research itself — can thrive. We will report our progress as we continue to learn, grow, and improve our approach to wellness.

Natasha Gordon is the Program Administrator at the Center for Social Media and Politics, supporting Center management and administration. She is a master’s student in Political Science at NYU researching democratization and sociotechnical development.

Jen Rosiere Reynolds is the Center’s Lab Manager. Before joining CSMaP, Jen worked in national security and received her master’s degree in Government from Johns Hopkins University. She focused her research on free speech and violent extremist radicalization online.

Many thanks to Andrea Lampros and the U.C. Berkeley Human Rights Center team for sharing resources and providing a starting point for our research environment and wellness planning.