Johannes Eichstaedt remembers the moment he knew he wanted to switch his focus from particle physics to psychology. A decade ago, gazing at the particle accelerator at Argonne National Laboratory, he realized he felt out of place both professionally and personally.
“I realized I cared more about people than I did about particles,” he says. “I wanted to work on how humans live their lives, and I wanted it to be work that had the potential to be relevant to a lot of people, so I decided to switch into the social sciences. I chose psychology because it’s a connector between fields like cognitive science, sociology, public health, and economics. It does a lot of linkage, which I appreciate.”
Eichstaedt found himself drawn not only to the interdisciplinary potential of psychology, but also to the relatively new subfield of positive psychology, which strives to understand what makes human life most worth living. At the University of Pennsylvania, he joined a cadre of young psychologists and computer scientists anxious to take advantage of increasing government interest in measuring the happiness and mental health of their citizens.
“We thought that if these governments were doing this at a small scale, maybe we could find ways to do it globally and cheaply using social media-based indicators,” he says. “We started the World Well-Being Project in an attempt to use social media and large-scale aggregated data and run it through natural language processing to create indicators of well-being for counties and cities.” Today, Eichstaedt is a computational social scientist and the Ram and Vijay Shriram Faculty Fellow at the Stanford Institute for Human-Centered Artificial Intelligence (HAI). He continues to collaborate with many of those colleagues as they study a growing volume of social media data to try to understand the mental and physical health of users.
“Together we now have tremendous firepower, with 30 to 40 people working on this from all different aspects,” says Eichstaedt, who also directs the Stanford Computational Psychology and Well-Being Lab. “Ten years ago, we began the consortium by studying well-being, but now we want to know if we can use these methods to help understand disease and reduce mortality.”
His team uses social media data and machine learning to gain insight into a wide range of health-related issues; they’ve recently identified heightened levels of depression in the Black community following the murder of George Floyd, and tracked the public’s COVID-19 social-distancing trends and adherence to public health guidelines. One current focus is building a dataset of millions of Americans whose tweets will be analyzed over time to screen for individual communities’ mental health, which is believed to have worsened significantly during the pandemic. Using prediction models that search for words associated with negative emotions and cognitions, Eichstaedt is able to pinpoint counties throughout the U.S. where the population is at particular risk of depression and its associated physical complications.
“Communities have properties,” he says. “It turns out that these social media pipelines are really good at predicting things that are psychological in nature—things like suicides, accidents, drinking, and even atherosclerotic heart disease.”
Eichstaedt is also working with HAI Associate Director Russell Altman to develop a project tracking the opioid pandemic using drug-related language culled from social media. The project could give healthcare professionals and policymakers much-needed community health information more quickly and cheaply than traditional surveying.
Watch a conversation between Johannes Eichstaedt and Russell Altman: How Social Media Can Help Gauge Societal Health
Privacy is at the forefront of Eichstaedt’s work. When studying individuals, his team obtains the necessary permission to analyze participants’ social media feeds, follows strict privacy guidelines, and is externally reviewed for compliance and ethical research conduct. Few social media users, however, realize how much information can be revealed by allowing access to their statuses or “likes,” raising important questions on user privacy, informed consent, data protection, and data ownership. Yet the technology also holds the potential to greatly benefit public health.
“I think we’re not even foreseeing some of the most important applications for this technology,” he says, adding that they could range from predicting communities most at risk of having low birthweight babies to assessing human stress levels following climate change-related disasters such as wildfires. The technology could be especially useful for gaining health insights into populations in under-resourced areas of the world where data collection isn’t as prolific as in the U.S.
Read policy brief: AI-Enabled Depression Prediction Using Social Media
“The beauty of what we’re doing at these interdisciplinary AI intersections is that we’re really building things that haven’t existed before,” Eichstaedt says. “It’s often hard for interdisciplinary people like myself to find truly interdisciplinary environments in which to work. That’s why I’m really grateful for HAI, which has given me the chance to do my work at full force. And that’s been really nice.”
This article is part of the People of HAI series which spotlights our community of scholars, faculty, students, and staff coming from different backgrounds and disciplines.
Stanford HAI's mission is to advance AI research, education, policy, and practice to improve the human condition. Learn more.