To learn more about the COVID-19 related research projects Khoury faculty are working on, select a project below.
- Applying reinforcement learning to learn policies with better outcomes (Tina Eliassi-Rad)
- Behavior-training technology to end the spread of COVID-19 (Stephen Intille)
- Developing new, effective treatments to combat COVID-19 (Albert-László Barabási)
- How to effectively communicate the spread of disease (Cody Dunne)
- How to have an Election Day during a pandemic (Brennan Klein, Stefan McCabe, Tim LaRock, Leo Torres, Alessandro Vespignani, Sam Scarpino, David Lazer, Tina Eliassi-Rad, Dr. Matteo Chinazzi)
- Making remote data visualization work more effective, remotely (Michelle Borkin, Cody Dunne)
- Monitoring COVID-19 trials (Byron Wallace)
- Optimizing algorithms for behavioral interventions and diagnostic inference (Ravi Sundarum, Raj Rajaraman)
- People’s response to COVID-19 (Jacqueline Griffin, Andrea Parker, Miso Kim, Stacey Marsella)
- Predicting the spread: advising the White House, CDC and WHO (Alessandro Vespignani)
- Privacy and data protection implications of COVID-19 contact tracing, public surveillance, and data collection efforts (Woodrow Hartzog)
- Revealing privacy implications of dark patterns in COVID-19 mobile apps (Alan Mislove, Christo Wilson, Woodrow Hartzog, David Choffnes)
- Social media and public response (Lu Wang)
- The state of news media bias and misinformation about the outbreak (Lu Wang)
- Understanding net neutrality violations in response to COVID-19 (David Choffnes)
- Using disease outbreak patterns to forecast what’s coming next (Cody Dunne)
- Using Foldit to design a synthetic protein that could be used as a COVID-19 antiviral (Seth Cooper)
Applying reinforcement learning to learn policies with better outcomes
Tina Eliassi-Rad’s team is applying reinforcement learning on partially observed coronavirus contact networks to learn which policies may lead to better outcomes. This is joint work with Jan-Willem Van De Meent and Dr. Rajmonda Caceres (at MITLL).
Behavior-training technology to end the spread of COVID-19
Stephen Intille and his team are using real-time behavior recognition technology and operant conditioning methods to train people in important health-related behavior changes to stop the spread of disease. The training system would be worn by the user, designed to alert them when they use incorrect hand-washing techniques or touch their faces. To increase the impact, they are interested in deploying a prototype system into high-risk settings such as hospitals and nursing homes. With additional funding, the team would be able to support a dedicated full-time student to further develop the project and its goals.
Developing new, effective treatments to combat COVID-19
Albert-László Barabási’s Network Medicine research team identified 40 new and repurposed drugs that could potentially treat COVID-19. The findings come from a modeling tool for infection dynamics based on network science—complex math, artificial intelligence (AI), physics, and computing. Barabási’s toolset mapped how 332 proteins within human cells behave after the cell is hijacked by the novel coronavirus. These treatments target areas within the cell where the virus operates, instead of directly targeting the virus. Funding would accelerate further results by providing access to a supercomputer to quickly understand the binding process for 8,000 additional proteins. Analyzing one binding protein takes 3-6 months on a typical computer.
How to effectively communicate the spread of disease
Cody Dunne’s team, in addition to developing a visualization exploring different COVID-19 outbreaks across the world, is also exploring the best way to disseminate and communicate information about those outbreaks. As researchers, government officials, and journalists frequently use visualizations to explain the pandemic to their audiences, research is necessary to understand what types of visualizations and data are most effective. Dunne’s team has numerous questions: How do different audiences interpret these visualizations? How does viewing them affect behavioral responses? Do the visualizations effectively communicate important messages about COVID-19 and assist people in decision-making? Funding is needed to collect relevant visualizations from social media and news sources; to classify and analyze their efficacy; and create a set of information visualization guidelines to better inform this work going forward.
How to have an Election Day during a pandemic
Brennan Klein, Stefan McCabe, Tim LaRock, Leo Torres (all Network Science PhD students) are working with a team of PhD students to think ahead to November’s elections, hoping to evaluate whether holding large elections would result in new transmissions of COVID-19. The team plans to provide a retrospective assessment of the impact Wisconsin’s April 7 in-person elections may have had on infection rates. With urgent funding, the team would be able to answer a series of critical research questions: Did Wisconsin’s in-person voting result in new transmissions of COVID-19? If so, can we estimate the number and location of new infections coming from polling places? If not, can we estimate the conditions under which in-person voting would have likely resulted in new infections? The answers they find could then be used to produce estimates about what might happen if Americans go out to vote in November, providing policy briefs and information to assist lawmakers and health experts as they plan for Election Day.
The team includes Alessandro Vespignani and Sam Scarpino (both with expertise in epidemiology, working closely with the CDC on COVID-19 projections), David Lazer (with expertise in computational social science, especially in causal methods), Tina Eliassi-Rad (with expertise in machine learning and data mining), and Matteo Chinazzi (with expertise in big data and disease modeling).
Making remote data visualization work more effective, remotely
Monitoring COVID-19 trials
Byron Wallace and collaborator Iain Marshall (King’s College London) maintain “trialstreamer,” which uses machine learning and natural language processing to maintain a continuously updated database of all reports of clinical trials in humans. All identified trial reports are run through a suite of models to extract key information and make it searchable and accessible. One use of this is to monitor for all COVID-19 trials.
Optimizing Algorithms for Behavioral Interventions and Diagnostic Inference
1. To explore the foundations of policy design for controlling epidemics using behavioral interventions
2. To enable faster and more accurate diagnostic inference by adapting and enhancing the tools of machine learning developed in the context of e-commerce.
People’s response to COVID-19
Jacqueline Griffin, Andrea Parker, Miso Kim and Stacey Marsella are working on a telephone survey of people’s responses to COVID-19. Griffin’s interest is in the information seeking practices of at-risk people. Marsella’s interest is to model and predict decision-making under stress. Marsella and PhD student Nutchanon Yongsatianchot have been working on a model of decision-making under stress that they have been evaluating with hurricane evacuation behavior but now want to expand to how people are coping with COVID-19. In addition to that, a group headed by Griffin has an existing project to model the impact of disruptions in the pharmaceutical drug supply network. This project will be exploring COVID-19 related issues.
Predicting the spread: advising the White House, CDC and WHO
Alessandro Vespignani, an expert in infectious disease modeling, is using data-driven computational modeling to simulate different intervention strategies (travel bans, school closures, social distance policies) and evaluate their effectiveness. His recommendations directly contribute to decisions made by local, state, national, and international leaders, including the CDC and WHO. Vespignani’s team is one of four university teams creating models to advise the White House; their recommendations formed the basis of the Trump Administration’s decision to extend social distancing guidelines through the end of April. This work has received significant media coverage, most notably in the NYTimes on March 13 and on April 23. As the demand on Vespignani’s lab grows, he urgently needs financial resources to meet computational expenses, bolster personnel, and access additional datasets.
Privacy and data protection implications of COVID-19 contact tracing, public surveillance, and data collection efforts
Woodrow Hartzog is studying the privacy and data protection implications of COVID-19 contact tracing, public surveillance, and data collection efforts. Specifically, Hartzog is examining the privacy implications of contact tracing apps, the use of facial recognition to monitor the spread of COVID as well as enforce social distancing, and the current and proposed law and policy frameworks that govern the collection and use of health data. He plans on incorporating this research into future law review articles on trust-based privacy frameworks and electronic surveillance frameworks built around the concept of obscurity.
Revealing privacy implications of dark patterns in COVID-19 mobile apps
Alan Mislove, Christo Wilson, Woodrow Hartzog, and David Choffnes have been jointly advising a Ph.D. student investigating privacy dark patterns, which are user interfaces that manipulate users into giving up their personal data and typically then selling that data to third parties. With the rise of COVID-19, there have been a surge of mobile apps that seek to prey on consumers’ fears about the virus, and often are misleading about what data they are gathering and what they are doing with it. Funding is needed to support one Ph.D. student to continue research into these apps and their dark patterns, identifying the harms consumers may face as a result of this unknown data collection.
Social media and public response
Lu Wang designs natural language processing (NLP) methods to identify the public’s reactions to the COVID-19 pandemic from social media data. Importantly, Wang’s team focuses on what the public does (e.g. their reactions to government policies and expert suggestions), and what reasons or evidence they rely on to support their behavior.
The state of news media bias and misinformation about the outbreak
Lu Wang’s team is studying the news media coverage of COVID-19, attempting to understand how different media outlets from different countries frame the outbreak. The team is interested in understanding how journalists present the underlying issues, including government policies and behavior, suggestions to the public, blame attribution, and interpretation over conspiracy theories. With funding, the team is hoping to evaluate whether media source affects the presentation of information, and to reveal informational bias. The team plans to conduct research over multiple multilingual media corpora, comparing media in different countries and of different political leanings.
Understanding net neutrality violations in response to COVID-19
David Choffnes’ team has collected data on net neutrality violations for the past three years, finding that video streaming services are often throttled by U.S. cellular providers. As communities continue to practice social distancing and remote work, streaming sites are likely seeing an increase in traffic — but are they still being throttled, and are new services that are also seeing increased use, like teleconferencing programs, facing similar problems? The team seeks funding for two students to dedicate time to exploring those research questions for multiple upcoming months.
Using disease outbreak patterns to forecast what’s coming next
Cody Dunne’s lab is developing a prototype visualization to understand the spread in the United States by comparing it to earlier COVID-19 outbreaks in other countries, such as Italy. Because outbreak timing, government regulations, and the infection and death rates all differ widely, the team had to identify a way to standardize the data in order to compare it. They found that by aligning each outbreak by sentinel events, such as the first 400 people infected, they were able to see clear patterns and evaluate the similarities and differences. The prototype visualization allows users to view the data with numerous parameters: they can align it by the number of people infected, look at growth vs. absolute values, normalize by population or by log scale, and more. In its first two days, it was viewed more than 6,000 times — but it needs more. With additional funding, the team will be able to expand the visualization’s capabilities, as well as maintain and update its current uses. On the list of desired additions are functions like aligning by type of intervention, regional comparisons within countries, and information on each country’s criteria for infection or death from COVID-19.
Using Foldit to design a synthetic protein that could be used as a COVID-19 antiviral
Seth Cooper’s team, in collaboration with a team at the University of Washington and others, are using the internet to its fullest potential. They’re part of the Foldit project, an online game designed to crowdsource protein folding puzzles and help solve science problems, that is currently working on designing a synthetic protein that could be used as a COVID-19 antiviral. The game is very flexible and can support numerous protein-related problems, and Northeastern researchers are working to increase the effectiveness of the game and the game’s tutorial system. With additional funding, both teams would be able to develop tutorial or game content specific to COVID-19, increasing the impact and galvanizing users to continue working on a solution.