Researchers at the Courant Institute of Mathematical Sciences found that Tuesday’s vice presidential debate negatively affected online discourse about Vice President Kamala Harris more so than former president Donald Trump, according to online activity immediately following the debate.
The study measured what they called “sentiment scores” between Harris and Trump three hours before, during and two hours after the debate between Sen. J.D. Vance and Gov. Tim Walz. Attitudes toward Harris plummeted by 18% from before the debate to during, while Trump saw a 23% dip during that same period. Two hours after the debate, Harris’ sentiment score had declined even further while Trump’s rebounded by 10%.
Members of NYU’s Predictive Analytics and AI Research Group, which conducted the study, analyzed online users’ feelings about both candidates in real-time by interpreting the rhetoric of user posts on Reddit, Google search trends, YouTube comments and trending tags on the social media platform X using algorithms that assess how positively or negatively language is processed.
Anasse Bari, a clinical professor at Courant and head of the research group, said he was interested in an alternative approach to election research and understanding how media affects individuals’ political perceptions.
“Our goal was to derive insights that could reveal new indicators of the election, particularly in gauging public opinion regarding both Republican and Democratic candidates,” Bari said in a statement to WSN. “It was exciting to watch the debate with my students, apply algorithms in real-time and discuss the results together.”
The same group of researchers conducted a similar experiment during the Sept. 10 presidential debate that aimed to analyze how candidates’ discussions could change viewers’ interest. Abortion — the most-searched topic after both debates — saw a 22% increase in related searches after the vice presidential debate compared to the presidential debate. Other highly searched words included “economy,” “Israel,” “tax” and “carbon emissions.”
Bari, who used predictive analytics to analyze the 2016 and 2020 elections, said that there are “many similarities” between all three elections in terms of key topics that concern voters — such as abortion and immigration, which have both been consistently salient campaign issues. Bari said that these themes “continue to dominate” this year’s election season, noting the importance of data to understand driving factors for voters and the topics they prioritize.
The researchers used algorithms to categorize the language surrounding each candidate into “positive,” “negative” and “neutral” sentiments. Positive sentiment was indicated by positive language, such as references to “improving” inflation and negative sentiment reflected criticism or resistance, while neutral sentiment consisted of neither.
Bari said that while the study exclusively includes audience members who are active on social media, its findings still provide practical insights.
“In the information age, big data and AI are here to stay,” Bari told WSN. “Tweets, Google searches and Facebook posts should not be ignored and may prove crucial in predicting a presidential election, regardless of the noise.”
Contact Mariapaula Gonzalez at [email protected].
