The Language of Fandom

By Patrick Bohn, October 21, 2025

Professor Venkata S. Govindarajan studies how language reflects the ups and downs of professional sports fandom.

Venkata S. Govindarajan, an assistant professor of computer science, studied the language choices of NFL fans in Reddit comment threads.

If you’re a fan of sports, you’ve almost certainly engaged in passionate discussions and debates with your favorite team’s fellow supporters during a game. Whether it’s the thrill of beating Cortland, or—if you’re a Bills fan—the agony of losing to the Kansas City Chiefs again, being a fan means riding the roller coaster of emotions that comes with following a team.

And, if you’re a digital native, there’s a strong chance this discourse takes place at the message boards and comment threads that populate the Internet.

An assistant professor of computer science at Ithaca College, Venkata S. Govindarajan describes himself as a “computational linguist at heart.” (Broadly, a computational linguist is someone who blends linguistics and computer science to understand human language.) Govindarajan’s research examines how we talk about in-groups (groups you consider yourself a part of) and out-groups (ones you don’t).

“In language, people us phrases like ‘us’ and ‘we’ when making themselves part of the in-group and ‘they’ and ‘them’ when making themselves part of the out-group,” he explains. “Usually, it’s done to reinforce good qualities about an in-group and vice versa.”

Sports, with its clearly defined lines of fandom, provides the perfect avenue for that type of analysis.

Specifically, what Govindarajan wanted to do was study how the language people use to describe in-groups changes based on the current state of the world and the current fortunes for the in-group.

“I needed a way to attach a number to indicate how good the state of the world is for the in-group independent of language," he said. “Sports is perfect for that, because the changing score and team’s odds for winning provided a clear state for the in-groups.”

“I didn’t want to analyze this in a lab setting ... which [would give] the experiments more of an artificial feel. I wanted to see how this worked in the real world.”

Venkata S. Govindarajan, assistant professor of computer science

Govindarajan chose to study the NFL because football provides frequent natural breaks, where fans can talk about what just happened, and a team’s odds for winning can change.

“I knew I didn’t want to analyze this in a lab setting, which is where most of the studies on in-group language take place, which gave the experiments more of an artificial feel,” he said. “I wanted to see how this worked in the real world.”

So, he turned to Reddit, the social media platform composed of countless subreddits—or topics—on nearly every subject imaginable. Each NFL team has its own dedicated subreddit where conversations among fans occur, and live game conversation threads can generate hundreds, or even thousands, of comments.

“What made Reddit so useful was that all the comments in each thread were timestamped, which made it very easy to connect a comment to a specific event in the game,” he said. “It was like an online sports bar with thousands of people talking at once.”

With the live game threads for 32 NFL teams during the 2021-22 and 2022-23 seasons serving as his data set, Govindarajan tracked nearly six million comments after pre-processing to filter out low-quality comments using heuristics like length and meta-data. Using open-source AI language models, he annotated more than 100,000 comments for words and phrases referring to the in- and out-groups and assigned each comment to the win probability for the in-group.

“Fans aren’t abandoning their team; instead, they’re circling the wagons and supporting them no matter what. It’s like the line from Henry V by Shakespeare: ‘We few, we happy few, we band of brothers.’”

Venkata S. Govindarajan

And because Govindarajan focuses his research on natural language use using modern machine learning tools and techniques, he utilized open-source AI language models to sort the comments into those referring to the in- and out-groups, respectively, and assigned each comment to a game state.

The sheer scale of comments made coding by AI necessary. A single human coding one comment every six seconds for 10 hours a day, five days a week, would take 2 1/2 years to go through 6 million comments. AI usage allowed for Govindarajan to study a larger scope of comments.

“The ability to use AI and technology to help me code these comments was invaluable,” said Govindarajan, who presented his findings at the 2024 Conference on Empirical Methods in Natural Language Processing, one of the foremost academic conferences on the usage of AI for text analysis.

As for the results themselves, what he found was “kind of heartwarming.”

“I thought I’d see increased usage of phrases like ‘we’ and ‘us’ among fans when things were going well for their team (the in-group)” Govindarajan said. “But I saw more of that from fans when the team was doing poorly.

“There was a sense of protectionism in the language that fans used then, almost like a rallying of the troops,” he continues. “Fans aren’t abandoning their team; instead, they’re circling the wagons and supporting them no matter what. It’s like the line from Henry V by Shakespeare: ‘We few, we happy few, we band of brothers.’”