Research
Penn Media Accountability Project (PennMAP)
PennMAP is building technology to detect patterns of bias and misinformation in media from across the political spectrum and spanning television, ratio, social media, and the broader web. We will also track consumption of information via television, desktop computers, and mobile devices, as well as its effects on individual and collective beliefs and understanding. In collaboration with our data partners, we are also building a scaleable data infrastructure to ingest, process, and analyze tens of terabytes of television, radio, and web content, as well as representative panels of roughly 100,000 media consumers over several years.
COVID – Philadelphia
Our team is building a collection of interactive data dashboards that visually summarize human mobility patterns over time and space for a collection of cities, starting with Philadelphia, along with highlighting potentially relevant demographic correlates. We are estimating a series of statistical models to identify correlations between demographic and human mobility data (e.g. does age, race, gender, income level predict social distancing metrics?) and are using mobility and demographic data to train epidemiological models designed to predict the impact of policies around reopening and vaccination.
High-Throughput Experiments on Group Dynamics
To achieve replicable, generalizable, scalable, and ultimately useful social science, we believe that is necessary to rethink the fundamental “one at a time” paradigm of experimental social and behavioral science. In its place we intend to design and run “high-throughput” experiments that are radically different in scale and scope from the traditional model. This approach opens the door to new experimental insights, as well as new approaches to theory building.
Common Sense
This project tackles the definitional conundrum of common sense head-on via a massive online survey experiment. Participants are asked to rate thousands of statements, spanning a wide range of knowledge domains, in terms of both their own agreement with the statement and their belief about the agreement of others. Our team has developed novel methods to extract statements from several diverse sources, including appearances in mass media, non-fiction books, and political campaign emails, as well as statements elicited from human respondents and generated by AI systems. We have also developed new taxonomies to classify statements by domain and type.
News
Duncan Watts and CSSLab’s New Media Bias Detector
The 2024 U.S. presidential debates kicked off June 27, with President Joe Biden and former President Donald Trump sharing the stage for the first time in four years. Duncan Watts, a computational social scientist from the University of Pennsylvania, considers this an ideal moment to test a tool his lab has been developing during the last six months: the Media Bias Detector.
“The debates offer a real-time, high-stakes environment to observe and analyze how media outlets present and potentially skew the same event,” says Watts, a Penn Integrates Knowledge Professor with appointments in the Annenberg School for Communication, School of Engineering and Applied Science, and Wharton School. “We wanted to equip regular people with a powerful, useful resource to better understand how major events, like this election, are being reported on.”
What Public Discourse Gets Wrong About Misinformation Online
Researchers at the Computational Social Science Lab (CSSLab) at the University of Pennsylvania, led by Stevens University Professor Duncan Watts, study Americans’ news consumption. In a new article in Nature, Watts, along with David Rothschild of Microsoft Research (Wharton Ph.D. ‘11 and PI in the CSSLab), Ceren Budak of the University of Michigan, Brendan Nyhan of Dartmouth College, and Annenberg alumnus Emily Thorson (Ph.D. ’13) of Syracuse University, review years of behavioral science research on exposure to false and radical content online and find that exposure to harmful and false information on social media is minimal to all but the most extreme people, despite a media narrative that claims the opposite.
Mapping Media Bias: How AI Powers the Computational Social Science Lab’s Media Bias Detector
Every day, American news outlets collectively publish thousands of articles. In 2016, according to The Atlantic, The Washington Post published 500 pieces of content per day; The New York Times and The Wall Street Journal more than 200. “We’re all consumers of the media,” says Duncan Watts, Stevens University Professor in Computer and Information Science. “We’re all influenced by what we consume there, and by what we do not consume there.”