COVID-19 and AI: A Virtual Conference
Session I: Landscape and Framing
Challenges Responding to COVID-19: Perspectives from a Physician and Policy Maker
In his remarks, Congressman Bera recapped the evolution of the COVID-19 virus, then talked about the ways big data and AI can help going forward. This includes helping public and healthcare professionals get through the immediate crisis and paying attention to other countries that are a month or two ahead of the U.S. to see what we can learn. He also spoke about the importance of serologic testing to understand how many people in the community have been affected by the coronavirus.
An Academic Medical Center’s Data Science Response to a Pandemic
Dr. Shah’s presentation focused on Stanford as an academic medical center and the collaborative efforts to respond to this pandemic. This work includes operational planning, clinical care decisions and broader research questions that drive science. He presented information on the models being created and the sources feeding into those models.
Issues in Responsible Reporting of COVID-19
Disease does not spread in isolation - it spreads alongside rumors, hoaxes and pseudoscience. Seema Yasmin’s talk looks at concurrent spread of disease and misinformation and disinformation about disease. How do anti-science movements and disinformation campaigns threaten public health and how might we mitigate the spread of false health news? How can reporters counter false narratives about contagion? Even well-meaning information can contribute to a sense of information overload for a public struggling to make sense of multiple information streams. Accurate, accessible information is critical during public health crises to help people follow evidence-based guidance on how to stay safe.
Global Best Practices in Controlling the COVID-19 Pandemic
A global comparison of best practices by countries/territories that have had some successes in curtailing the COVID-19 pandemic provides valuable lessons for others. The research Michele Barry will speak about is the result of an in-depth literature review and will focus on responses from four key countries/territories: Singapore, South Korea, China and Hong Kong. This is an important problem to address as the tested and tried approaches of these countries/territories may inform and shape policies of other countries, especially those more recently experiencing an increase in COVID-19 cases.
Session II: Social Impacts & Bio-Security
COVID-19 Infodemic and Crisis Informatics
In this presentation, Kate Starbird examines misinformation and disinformation in the context of crisis events, and how COVID-19 is different from any other crisis. It draws from the sociology of disaster and the social-psychology of rumoring, as well as recent historical work on disinformation. And it applies these in the context of online information-sharing during crisis events. Viewers will learn about the challenges related to the spread of mis- and disinformation as well as recommendations for people participating in these spaces and as crisis communicators.
COVID-19: Misinformation & Disinformation
The Stanford Internet Observatory is a cross-disciplinary program of research, teaching and policy engagement for the study of abuse in current information technologies, with a focus on social media. Renee DiResta’s presentation examines China's English-language state media properties and their propaganda activities surrounding the novel coronavirus. Understanding what China is communicating outwardly, to the English-speaking world, and how those narratives are being adopted, shared, and spread in both the media and social media environment enables fact-checkers and public leaders to counter the narratives where necessary, and to understand how China sees this virus in the context of its geopolitical standing.
COVID-19 & Biosecurity
Advances in biotech can both mitigate and cause catastrophic biological threats. Megan Palmer will talk about considerations for this dual use nature in addressing COVID-19 as well as preparing for and preventing future pandemics. Today biotech innovations such as diagnostics, therapies and vaccines are helping with solutions to the current catastrophe. Meanwhile, biotechnology advances and activities could lead to even more devastating accidents or intentional misuse scenarios. Megan will share some of the biotech community’s responses to COVID-19 and what this is teaching about ensuring research efforts translate to societal benefits, especially in times of crisis. She will also share some examples of policies designed to ensure oversight of pathogen research to protect against safety and security incidents.
“Foreign Bodies”: COVID-19 and Xenophobia
In her presentation, Eran Alam talked about how her research explores the virus as a vector of disease that’s foreign to the body, similar to bodies that are foreign to the imagination of the nation. She presented the way leaders and society have responded over time, including war time, regarding the foreign body as an enemy and the implications this has for society.
Session III: Tracking the Epidemic
Taiwan’s Use of Data Analytics to Control COVID-19
Transparency is critical in a democracy if digital technology is deployed to ensure protection of public health and civil liberties. Jason Wang will look at how technology has the potential to curb the spread of COVID-19, including examples where digital technology has successfully been used for disease surveillance. Taiwan provided a great example on the use of technology in case detection, contact tracing, isolation of cases, and quarantine of exposed individuals.
Tools for Estimating Unreported Infections of COVID-19
Viruses continue to mutate as they infect people, even in asymptomatic individuals. By leveraging the number of mutations between viruses captured in reported individuals, Lucy Li’s research uses mathematical models to predict the most likely number of missing infections. In this presentation, Li will talk about her research to identify the number of unreported infections using the viral genomes and WHO time-series data. Quantifying the total number of infections, not just the reported ones, is important for accurately characterizing the fatality rate given an infection, identifying changes in capacity for testing and providing an indicator for how long it will take to contain the epidemics. Data from China suggests that widespread testing for the virus even amongst asymptomatic individuals is necessary to effectively quantify infections and contain the epidemic. Furthermore, this large number of unreported infections indicates that even if the number of cases were going down, the total number of infections could remain high for a long period of time, so public health interventions would need to be maintained to prevent the number of infections from bouncing back.
Methods for Real Time Mapping of COVID-19 Cases Worldwide
In his presentation, John Brownstein presented how Harvard Medical School has been tracking COVID-19 since late December, and how the research includes a health mining tool called HealthMap, which is freely available. He also addressed the various ways to gather valuable data from the public both to help track the virus but also to understand social distancing impact.
Epidemiological Forecasting Tools for COVID-19
Reporting from one of the CDC's Centers of Excellence for Flu Forecasting, Ryan Tibshirani will speak about nowcasting and forecasting the Covid-19 pandemic at the level of individual US counties. These nowcasts/forecasts will be aggregated up to the state level and submitted as part of Covid-ILI forecasting efforts at the county and state level, assisting state and local public health officials, who can figure such forecasts into their decisions. This research adapts two existing systems for flu forecasting based on statistical machine learning and wisdom-of-crowds, respectively.
A Mobile App Intervention to Slow COVID-19 Using Crowdsourced Data
With a sufficient diagnosis rate and contact tracing accuracy, COVID-19 can be contained.
Non-pharmaceutical pandemic interventions fundamentally make a trade-off between two important social goods: loss of life from the pandemic and economic impact, which influences health and well-being outcomes indirectly. In general, non-pharmaceutical approaches to infectious disease control have the following components: filtering (picking a subset of the population) and intervention (modifying the behaviour of these people). Without good filtering, broad quarantines and social distancing are needed, incurring a huge cost in the form of negative impact on people’s lives. Tina White will speak about her research on performing automated contact tracing at scale using anonymized Bluetooth proximity sensing.
AI for COVID-19: An Online Virtual Care Approach
With half of the world’s population lacking access to healthcare services, and 30% of the adult population in the US having inadequate health insurance coverage to get even basic access to services, it should have been clear that a pandemic like COVID-19 would strain the global healthcare system way over its maximum capacity. In this context, many are trying to embrace and encourage the use of telehealth as a way to provide safe and convenient access to care. However, telehealth in itself can not scale to cover all our needs unless we improve scalability and efficiency through AI and automation. In this talk, Xavier Amatriain will describe how his work on combining latest AI advances with medical experts and online access has the potential to change the landscape in healthcare access and provide 24/7 quality healthcare. He will also describe how those approaches have been used to address the urgent and immediate needs of the current pandemic.
Knowledge Technology to Accelerate Open Science in Addressing the COVID-19 Pandemic
The response to COVID-19 involves a global community of investigators who want rapid access to emerging results and the opportunity to examine those results in the most efficient and reliable way possible. Science advances most effectively and rapidly when investigators have the opportunity to verify one another’s data and to explore those data in search of new discoveries. The urgency of the crisis makes “open science” an important goal so that researchers can share, access, and re-analyze one another’s data as soon as the data become available. In this presentation, Mark Musen will talk about the development of a Web-based application, known as the CEDAR Workbench, which makes it easy for scientists to create the metadata needed to describe their experiments—including information about the subject of the experiment, the experimental conditions, and the interventions that were made. The system also uses AI to learn patterns in the metadata, helping investigators to streamline their entry of new metadata. There is unprecedented pressure to speed up research on COVID-19, and tools that support open science can enhance the efforts of the entire community that is working so furiously to understand this virus.
Full description of CEDAR Workbench - a Web-based application developed by Musen’s team: https://metadatacenter.org
What We Can Learn From Twitter Analysis About COVID-19
Stanford HAI junior fellow Johannes Eichstaedt is a psychologist who uses social media to understand the psychological states of large populations. He examined Twitter posts to learn how the virus and social distancing are affecting our anxiety and life satisfaction and how factors such our age, education, and hometown size can impact our emotions.
Session IV: Treatments & Vaccines
Rapid Analysis of SARS-CoV-2 Genomic Content Using the Functional Genomics Platform
In her presentation, Eran Alam talked about how her research explores the virus as a vector of disease that’s forIn her presentation, Kristen Beck discussed the mutation of the COVID-19 virus and how IBM is working on the functional genomics platform. She also talked about the ways in which IBM is helping as the outbreak has intensified.
COVID-19 Open Research Dataset Challenge (CORD-19)
Anthony Goldblum will talk about how Kaggle is helping the global community and health organizations better understand the pandemic by contributing to the ecosystem of knowledge. The goal of Kaggle is to create an up-to-date picture of the state of knowledge on COVID-19 using two machine learning challenges. The first challenge involves taking the CORD-19 dataset, which contains 44,000 articles on COVID-19, and using it to answer key questions from science and health policy literature. The second challenge is to forecast cases and fatalities by city from the JHU data with the goal of identifying factors that impact transmission of COVID-19.
An up-to-date dashboard of what the scientific literature knows about COVID-19 and a nice collection of relevant datasets: https://www.kaggle.com/covid-19-contributions
Using machine learning to enable workflows to care for clinically deteriorating patients
Patients hospitalized for COVID are at high risk for getting sick rapidly, requiring ICU level care and mechanical ventilation performed in a highly coordinated manner. Clinically deteriorating patients must be identified early in order to appropriately prioritize clinical resources and to plan for escalation. As the number of hospitalized patients with COVID19 increases, it will be important to develop more consistent, reliable, and efficient workflows that do not rely solely on manual review of clinical data. In this presentation Ron Li will address how using a combination of data science, design thinking, and process improvement to build workflows enabled by machine learning may more efficiently help hospitals care for critically ill COVID19 patients.
AI-Assisted Elderly Care for Acute Infection and Chronic Disease
Seniors are the most vulnerable population in the COVID-19 pandemic. To take care of seniors while keeping them safe, a technology that can detect early symptoms of COVID-19, monitor home isolation, and manage chronic conditions at home is urgently needed. Fei-Fei Li will talk about an AI-powered smart sensor system to protect in-home elderly amid the COVID-19 pandemic. More specifically, her research will demonstrate that a privacy-protected system based on multi-modal sensor technology and activity understanding model can be used for acute infection and chronic disease management.
Identifying COVID-19 Vaccine Candidates with ML
Vaccines are one of the most powerful tools to curb a pandemic and prevent its recurrence. However, vaccine design is often a guessing game for new pathogens. With a goal of identifying fragments of SARS-CoV-2 that can be used for COVID-19 vaccines, Binbin Chen will address how AI tools built upon immunology knowledge and data can provide a better "educated guess" and increase the chances of finding an effective vaccine. Binbin and his colleagues made the full list of vaccine candidates available on BioRxiv, and are organizing a project to collect samples for the future pandemic with https://www.nextpandemic.org/.
Repurposing Existing Drugs to Fight COVID-19
The fastest way to address the urgent need to develop new treatments for COVID-19 is to find existing drugs that can be repurposed. Stefano Rensi will talk about how his research uses protein structure prediction (homology modeling) and molecular docking (physical simulations) to predict whether compounds can bind and inhibit a protein that facilitates viral invasion of cells (TMPRSS2).
From the HAI blog
- Artificial Intelligence and COVID-19: How Technology Can Understand, Track, and Improve Health Outcomes
- Tracking a Killer: Hunting the Coronavirus with Technology, AI, and Analytics
- What Twitter Reveals About COVID-19’s Impact on Our Mental Health
- Stanford Hospital Official COVID-19 site
- FSI Analysis on the COVID-19 Pandemic
- Centers for Disease Control and Prevention official site
Epidemic Forecasting - A dataset of all major efforts to reduce transmission of COVID-19
Kaggle COVID-19 - Kaggle calls for global community involvement In data gathering for COVID-19
- Global COVID-19 Map - Healthmap.org.
- COVID-19 Collaboration Opportunities