Artificial intelligence is poised to play an increasingly large role in societies across the world. Accordingly, there is a growing interest in ensuring that AI is used in a responsible and beneficial manner. A range of perspectives and contributions are needed, spanning the full spectrum from fundamental research to sustained deployments.
This seminar series will explore how artificial intelligence can contribute to solving social problems. For example, what role can AI play in promoting health, access to opportunity, and sustainable development? How can AI initiatives be deployed in an ethical, inclusive, and accountable manner?
Organizer: Andrew Perrault (firstname.lastname@example.org)
Human-Centered Machine Learning for Dangerous Mental Health Behaviors Online
Abstract: Research and industry use machine learning to identify and intervene in physically dangerous behaviors discussed on social media, such as advocating for self-injury or violence. There is an urgent need to innovate in data-driven systems to handle the volume and risk of this content in social networks and its propagation to others in the community. However, traditional approaches to prediction have mixed success, in part because technical solutions oversimplify complex behavior and the unique interactions of dangerous communities with both individuals and platforms. The difficulties in computationally handling these circumstances threatens the applications of these techniques to pressing social problems.
In this talk, I will describe my work in human-centered machine learning, an approach that refocuses technological innovation on the needs of humans, communities, and stakeholders. I study this through dangerous mental illness behaviors in online communities, like opioid abuse, suicidal ideation, and promoting eating disorders. First, I will talk about my work in building novel and human-centered prediction systems that make robust and accurate assessments of mental illness signals across several conditions. Then, I will discuss recent research on a crucial part of machine learning pipelines - generating labels for training data. I have found alarming gaps in construct validity and rigor that jeopardize the state-of-the-art – and I’ll discuss our current work on how we’re attempting to fix this. Together, these inform an agenda for human-centered machine learning that is scientifically rigorous and more considerate of social contexts in data, providing a pathway for more impactful and ethical problem solving in computer science.
Using AI to Augment HIV Prevention Interventions for Homeless Youth: Results from a Large Clinical Trial
Abstract: Each year, there are nearly 4 million youth experiencing homelessness (YEH) in the United States with HIV prevalence ranging from 3 to 11.5%. Peer change agent (PCA) models for HIV prevention have been used successfully in many populations, but there have been notable failures. In recent years, network interventionists have suggested that these failures could be attributed to PCA selection procedures. The change agents themselves who are selected to do the PCA work can often be as important as the messages they convey. To address this concern, we tested a new PCA intervention for YEH, with three arms: (1) an arm using an artificial intelligence (AI) planning algorithm to select PCA, (2) a popularity arm—the standard PCA approach—operationalized as highest degree centrality (DC), and (3) an observation only comparison group (OBS). We tested this approach with 704 youth between 2017 and 2019. We found that PCA models that promote HIV testing, HIV knowledge, and condom use are efficacious for YEH. Both the AI and DC arms showed improvements over time. AI-based PCA selection led to better outcomes and increased the speed of intervention effects. Specifically, the changes in behavior observed in the AI arm occurred by 1 month, but not until 3 months in the DC arm. Given the transient nature of YEH and the high risk for HIV infection, more rapid intervention effects are desirable.
Scientific Natural Language Processing and the Fight Against COVID-19
Abstract: This talk will describe the dramatic creation of the COVID-19 Open Research Dataset (CORD-19) and the broad range of efforts, both inside and outside of the Semantic Scholar project, to garner insights into COVID-19 and its treatment based on this growing corpus of research papers.
Measuring Economic Development from Space with Machine Learning
Abstract: Recent technological developments are creating new spatio-temporal data streams that contain a wealth of information relevant to climate adaptation strategies. Modern AI techniques have the potential to yield accurate, inexpensive, and highly scalable models to inform research and policy. A key challenge, however, is the lack of large quantities of labeled data that often characterize successful machine learning applications. In this talk, I will present new approaches for learning useful spatio-temporal models in contexts where labeled training data is scarce or not available at all. I will show applications to predict and map poverty in developing countries, monitor agricultural productivity and food security outcomes, and map infrastructure access in Africa. Our methods can reliably predict economic well-being using only high-resolution satellite imagery. Because images are passively collected in every corner of the world, our methods can provide timely and accurate measurements in a very scalable end economic way, and could significantly improve the effectiveness of climate adaptation efforts.
Addressing Problems in Global Health with Non-traditional Data and Machine Learning
Abstract: In this talk, I will present examples of my work on using non-traditional data from satellite images, social media, and other Internet sources to address global health problems. I will also discuss the challenges associated with using data from these sources. To conduct effective research using these data, it is important to consider and incorporate into analytical processes the distinct social, cultural, and economic context in different countries and communities.
Real-time pandemic planning and response: experiences from the ongoing COVID-19 pandemic
Abstract: COVID-19 pandemic represents an unprecedented global crisis. Its global economic, social and health is already staggering and will continue to grow. Computation and, more broadly, computational thinking plays a multi-faceted role in supporting global real-time epidemic science especially because controlled experiments are impossible in epidemiology. High performance computing, data science and new sources of massive amounts of data from device-mediated interactions have created unprecedented opportunities to prevent, detect and respond to pandemics.
In this talk, using COVID-19 as an exemplar, I will describe how scalable computing, AI and data science can play an important role in advancing real-time epidemic science.
Antidepressant Recommendation Systems: An Ongoing Journey for Technical and Clinical Impact
Abstract: How to choose antidepressants is an important question for patients who need medications to help manage their depression. In this talk, I will discuss how our ongoing effort toward clinical relevance -- that is, our search for ways to assist with antidepressant treatment recommendations -- has led to insights in core machine learning (we corrected a decade-old flaw in supervised generative modeling) as well as HCI (how psychiatrists react to incorrect recommendations). I will share how we select directions with interesting technical questions and societal impact, as well as what kinds of directions are good candidates for ML at all. (I hope to make some of this a more informal discussion.)
Work in collaboration with Roy Perlis, Tom McCoy, Michael Hughes, Eric Sudderth, Gabe Hope, Leah Weiner, Melanie Pradier, Maia Jacobs, Krzyzstof Gajos.
Machine Learning Applications During COVID-19
Abstract: Dr. Majumder will discuss three of her research group’s recent publications at the intersection of machine learning and epidemiology to answer a wide range of research questions about the ongoing COVID-19 pandemic. Broadly, the talk will cover agent-based models for epidemic dynamics, natural language processing for bibliometrics analysis, and novel digital data sources for misinformation surveillance.
Lessons from Imperfect Attempts to Serve (from Academia)
Abstract: The Center for Analytical Approaches to Social Innovation (CAASI) was founded in the Fall of 2019 to connect the tools of quantitative social science and computing to the practical needs of local organizations that serve vulnerable populations. While some of CAASI's projects have fit neatly into existing research categories, this is not the case for our most recent initiatives that arose in response to George Floyd's death this May. This talk will contrast these different approaches to community engagement through two specific examples: a field experiment with a provider of reintegration services to the formerly incarcerated and a volunteer-driven effort to build a tool to understand the process of holding the police accountable.
Stevie Chancellor (Northwestern University)
Dr. Stevie Chancellor is the CS + X Postdoctoral Fellow in Computer Science at Northwestern University. Her research combines approaches from human-computer interaction and machine learning to build and critically evaluate human-centered systems, focusing on high-risk health behaviors in online communities. Her work has been featured in The Atlantic, Wired and Gizmodo. Stevie recently received her doctorate in Human-Centered Computing from Georgia Tech, and will start as an Assistant Professor in Computer Science and Engineering at the University of Minnesota in 2021.
Eric Rice (Northwestern University)
Eric Rice is an associate professor and the founding co-director of the USC Center for Artificial Intelligence in Society, a joint venture of the USC Suzanne Dworak-Peck School of Social Work and the USC Viterbi School of Engineering. Rice received a BA from the University of Chicago, and an MA and PhD in Sociology from Stanford University. He was a postdoctoral fellow at the University of California, Los Angeles. He joined the USC faculty in 2009. In 2012 he received the John B. Reid Early Career Award through the Society for Prevention Research. He specializes in social network science and theory, as well as community-based research. His primary focus is on youth experiencing homelessness and how issues of social network influence may affect risk-taking behaviors and resilience. For several years he has been working with colleague Milind Tambe to merge social work science and AI, seeking novel solutions to major social problems such as homelessness and HIV. Rice is the author of more than 100 peer-reviewed articles in such publications as the American Journal of Public Health, AIDS and Behavior, the Journal of Adolescent Health, Pediatrics, Child Development, and the Journal of the Society for Social Work Research. He is the recipient of grants from the National Institute of Mental Health, the California HIV/AIDS Research Program, the Army Research Office and other agencies.
Oren Etzioni (Allen Institute for AI)
Dr. Oren Etzioni is Chief Executive Officer at AI2. He has been a Professor at the University of Washington’s Computer Science department since 1991. His awards include Seattle’s Geek of the Year (2013), and he has founded or co-founded several companies, including Farecast (acquired by Microsoft). He has written over 100 technical papers, as well as commentary on AI for The New York Times, Wired, and Nature. He helped to pioneer meta-search, online comparison shopping, machine reading, and Open Information Extraction.
Stefano Ermon (Stanford University)
I am an Assistant Professor in the Department of Computer Science at Stanford University, where I am affiliated with the Artificial Intelligence Laboratory and a fellow of the Woods Institute for the Environment.
My research is centered on techniques for scalable and accurate inference in graphical models, statistical modeling of data, large-scale combinatorial optimization, and robust decision making under uncertainty, and is motivated by a range of applications, in particular ones in the emerging field of computational sustainability.
Elaine Nsoesie (Boston University)
Dr. Nsoesie applies data science methodologies to global health problems, using digital data and technology to improve health, particularly in the realm of surveillance of chronic and infectious diseases. She has also been appointed as a BU Data Science Faculty Fellow, as part of the BU Data Science Initiative at the Hariri Institute for Computing. The Data Science Faculty Fellows program assembles a cluster of uniquely talented faculty whose expertise transcends traditional disciplinary boundaries to enable fundamental advances in data science. Dr. Nsoesie completed her PhD in Computational Epidemiology from the Genetics, Bioinformatics and Computational Biology program at Virginia Tech, and her PhD dissertation, Sensitivity Analysis and Forecasting in Network Epidemiology Models, at the Network Dynamics and Simulations Science Lab at Virginia Tech BioComplexity Institute. After postdoctoral associate positions at Harvard Medical School and Boston Children’s Hospital, Dr. Nsoesie joined the faculty of the Institute for Health Metrics and Evaluation (IHME) at the University of Washington.
Madhav Marathe (University of Virginia)
Madhav Marathe is an endowed Distinguished Professor in Biocomplexity, Director of the Network Systems Science and Advanced Computing (NSSAC) Division, Biocomplexity Institute and Initiative, and a tenured Professor of Computer Science at the University of Virginia. Dr. Marathe is a passionate advocate and practitioner of transdisciplinary team science. During his 25-year professional career, he has established and led a number of large transdisciplinary projects and groups. His areas of expertise are network science, artificial intelligence, high performance computing, computational epidemiology, biological and socially coupled systems, and data analytics.
Finale Doshi-Velez (Harvard School of Engineering and Applied Sciences)
Finale Doshi-Velez is a John L. Loeb associate professor in Computer Science at the Harvard Paulson School of Engineering and Applied Sciences. She completed her MSc from the University of Cambridge as a Marshall Scholar, her PhD from MIT, and her postdoc at Harvard Medical School. Her interests lie at the intersection of machine learning, healthcare, and interpretablity.
Maimuna Majumder (Boston Children’s Hospital and Harvard Medical School)
Dr. Maimuna (Maia) Majumder is a computational epidemiologist specializing in emerging epidemics and a recent graduate of the Engineering Systems program at MIT’s Institute for Data, Systems, and Society (IDSS). In between her graduate studies and her current position at CHIP, Maia spent a year at the Health Policy Data Science lab at Harvard Medical School’s Health Care Policy department as a postdoctoral fellow. During her masters and doctoral studies at MIT, she was funded through a graduate fellowship at HealthMap. Prior to Maia’s arrival at MIT, she earned a Bachelors of Science in Engineering Science (with a concentration in Civil and Environmental Engineering) and a Masters of Public Health in Epidemiology and Biostatistics at Tufts University. While at Tufts, Maia was a field researcher with the International Centre for Diarrheal Disease Research, Bangladesh (ICDDR,B), where she worked with clinic patients (and their data) to learn how to better tell their stories. Her current research interests involve probabilistic modeling, artificial intelligence, and “systems epidemiology” in the context of public health, with a focus on causal inference for infectious disease surveillance using digital disease data (e.g. search trends; news and social media). She also enjoys exploring novel techniques for data procurement, writing about data for the general public, and creating meaningful data visualizations. As of January 2019, Maia has been engaged in pandemic response efforts and is a leading expert in COVID-19 epidemiology.
Sera Linardi (University of Pittsburgh)
Sera Linardi is an Associate Professor at the Graduate School of Public and International Affairs (GSPIA) at the University of Pittsburgh, where she founded and directs the Center for Analytical Approaches to Social Innovation (CAASI). She received her Ph.D. in Social Science at the California Institute of Technology after working as a computer scientist at Adobe Systems. She bridges academic research and practical challenges in public/social services provision, specifically around prosocial behavior, information aggregation, and behavior economics of the poor. Her research has been published in economics, management, and political science journals (Journal of Public Economics, Management Science, Games and Economic Behavior, British Journal of Political Science) and won the 2016 Midwest Political Science Association Best Paper in Comparative Politics Award. Her work is currently supported by the NSF and the Heinz Endowment.