Nobel Conference 57

All sessions will be live streamed and archived unless a presenter requests otherwise. All activities are in Lund Arena unless otherwise noted.

Tuesday, October 5, 2021 - Session 1
Time Event
9 a.m. Musical Prelude 
Gustavus Wind Orchestra
9:30 a.m.

Lisa Heldke, Director, Nobel Conference
Rebecca M. Bergman, President of the College
Nobel Conference 57 Introduction, Thomas LoFaro, 2021 Nobel Conference Chair

10 a.m.

Data-Driven Decision Making: Now and Imagined

Lecture by Talithia Williams, PhD
Professor of Mathematics, Harvey Mudd College

Post-lecture comments by Lisa Heldke and Jillian Downey, Assistant Professor of Mathematics, Computer Science and Statistics

11 a.m. Panel Discussion and Audience Q&A
Tuesday, October 5, 2021 - Session 2
Time Event
12:10 p.m. Musical Prelude
Gustavus Wind Symphony
12:30 p.m.

How Much Evidence Do You Need? Data Science to Inform Environmental Policy During the COVID-19 Pandemic

Lecture by Francesca Dominici, PhD
Clarence James Gamble Professor of Biostatistics, Population and Data Science;
Co-Director, the Data Science Initiative, Harvard University

On December 7, 2020, the New York Times reported that then-President Trump declined to tighten soot rules. This was despite strong evidence of the adverse health effects including a link to COVID-19 deaths. Francesca Dominici will provide an overview of data science methods--including methods for causal inference and machine learning--to inform environmental policy. This is based on Dominici’s work analyzing a data platform of unprecedented size and representativeness. The platform includes more than 500 million observations on the health experience of more than 95% of the US population older than 65 years old linked to air pollution exposure and several confounders. Finally, the talk will provide an overview of Dominici’s studies on air pollution exposure, environmental racism, wildfires, and how they also can exacerbate the vulnerability to COVID-19.

Post-lecture comments by Lisa Heldke and Melissa Lynn, Assistant Professor of Mathematics, Computer Science and Statistics

1:30 p.m.

From the Village Watchman to Actionable Data: A Challenging Journey

Lecture by Michael Osterholm, PhD
A Regents Professor and McKnight Presidential Endowed Chair in Public Health; Director, Center for Infectious Disease Research and Policy at the University of Minnesota

Post-lecture comments by Lisa Heldke and Karl Larson, Professor and Program Director of Public Health

2:30 p.m. Panel Discussion and Audience Q & A
Tuesday, October 5, 2021
Time Event
3:30 p.m.
Explore specialized topics in a smaller, more interactive online format. 
6 p.m. Gallery Talk, Schaefer Art Gallery 

Arlene Birt, "Background Stories" 
Meet artist Arlene Birt and learn more about her creative process and work. The exhibit, Putting Data Into Context, is in creative conjunction with Nobel Conference 57, Big Data Revolution. Sponsored by the Art and Art History Department.

Arlene Birt, is an infodesigner, visual storyteller, public artist and educator. She incorporates behavioral psychology to visually explain the stories behind products and places and to help individuals connect emotionally to seemingly distant environmental topics. She’s the founder of Background Stories, a company that creates “clear visuals of complex stories.” Her exhibition for this conference includes installation and participant based artworks that use data as a means of visual creativity. 

Visit the Schaefer Art Gallery website for exhibit images and the Zoom link to the gallery talk.

7 p.m.

Nobel Conference Concert
Bjorling Recital Hall

OboeBass! presents American Vein, New Music for Oboe and Bass
Featuring Carrie Vecchione, oboe and Rolf Erdahl, double bass

What do the arts and big data have in common? How do they intersect? Is the creative act uniquely human? This concert features works composed for OboeBass! since 2019 by composers who provide points of departure for these questions.

Live concert is free and open to the public.
Event will be livestreamed via the Nobel Conference website.

Carrie Vecchione, oboe, and Rolf Erdahl, bass, are OboeBass! Called “pioneers” by Minnesota Public Radio, they concertize widely around the U.S., and toured Norway in 2017. They have commissioned over forty new oboe/bass pieces and recorded six CDs, effectively creating the oboe/bass duo genre.

Time Event
9 a.m. Music Prelude
Gustavus Symphony Orchestra
9:15 a.m.
Overview of the Day
9:30 a.m.

Interpretable Machine Learning

Lecture by Cynthia Rudin, PhD
Professor of Computer Science, Electrical and Computer Engineering, and Statistical Science; Director, Prediction Analysis Lab, Duke University

With machine learning have come serious societal consequences from using black box models for high-stakes decisions: flawed bail and parole decisions, racially-biased models in healthcare, and inexplicable loan decisions in finance. Transparency and interpretability of machine learning models is critical in high stakes decisions. However, there are clear reasons why organizations might use black box models instead: it is easier to profit from inexplicable predictive models than transparent models, and it is easier to construct complicated models than interpretable models. Most importantly, there is a widely-held belief that more accurate models must be more complicated, and more complicated models cannot possibly be understood by humans. Both parts of this last argument lack scientific evidence and often are not true in practice. In many cases, carefully constructed interpretable models are just as accurate as their black box counterparts on the same dataset. Cynthia Rudin will discuss the interesting phenomenon that interpretable machine learning models are often as accurate as their black box counterparts, using examples encountered throughout her career: manhole fires in New York City, caring for critically ill patients and predicting criminal recidivism.

Post-lecture comments by Lisa Heldke and Jessie Petricka, Associate Professor of Physics

10:30 a.m.

Justice in Machine Learning/AI for Health Care

Lecture by Pilar Ossorio, JD, PhD
Professor of Law and Bioethics, University of Wisconsin

Health care organizations and health care providers are using advanced algorithmic approaches (machine learning or artificial intelligence “ML/AI”) for a variety of purposes. For instance, organizations use ML/AI to assess the quality of health care services, make systems more efficient, and determine which patients need extra follow up and care coordination. Health care providers use ML/AI to identify recommended treatments, predict patient outcomes, and help with diagnoses. However, growing literature on ML/AI indicates that algorithms or combinations of them can reproduce race, gender, class and other social biases, and a small literature has now shown how ML/AI used in health care also incorporates a variety of pernicious and unfair biases. In this talk, we will consider how such biases become encoded in ML/AI for health care and some means for decreasing bias. We will also discuss how we might use the technology to identify existing unfair biases within health care systems with the aim of ameliorating them. 

Post-lecture comments by Lisa Heldke and Phil Voight, Associate Professor of Communication Studies

11:25 a.m. Panel Discussion and Audience Q & A
Wednesday, October 6, 2021 - Session 4
Time Event
1:15 p.m. Music Prelude
Gustavus Jazz Ensemble
1:30 p.m.

Child Protection: Too Much and Not Enough 

Lecture by Rhema Vaithianathan, PhD
Professor of Health Economics; Director, Centre for Social Data Analytics,
Auckland University of Technology

 By the age of 18, one in three American children will have been investigated for suspected child abuse or neglect by a child welfare agency. Yet despite this surprising level of surveillance, the rates of serious injuries and deaths from child maltreatment remain high; on average, more than five American children die every day from abuse or neglect. After every tragic maltreatment death, agencies tend to increase their rates of investigations, casting the net even wider and bringing more families and children into the system. This means a child protection system that was meant to only be used to investigate rare cases of abuse or neglect is overwhelmed with large caseloads of children, many of whom are at minimal levels of risk. Research tells us that predictive risk modeling tools are good at estimating the risk of future harm for children who are referred to agencies. But these are high stakes decisions, and a human centered approach is essential. As Vaithianathan and her team learned when they developed and implemented the Allegheny Family Screening Tool–a world-first use of predictive analytics to help triage referrals about abuse and neglect–that means using data in an ethical, transparent, and trusted way to achieve the social licence needed to proceed.

Post-lecture comments by Lisa Heldke and Kate Knutson, Professor of Political Science

2:30 p.m.

Discriminating Data

Lecture by Wendy Chun, PhD
Canada 150 Research Chair; Leader, the Digital Democracies Institute, 
Simon Fraser University

An exploration of polarization as a goal—not an error—within current practices of predictive data analysis and machine learning. These methods encode segregation, eugenics, and identity politics through their default assumptions and conditions. Correlation, for instance, which grounds big data's predictive potential, stems from twentieth-century eugenic attempts to “breed” a better future. Recommender systems foster angry clusters of sameness through homophily. Users are “trained” to become authentically predictable via a politics and technology of recognition. These predictive programs thus seek to disrupt the future by making disruption impossible. 

Post-lecture comments by Lisa Heldke and Colleen Stockmann, Assistant Professor of Art and Art History

3:30 p.m. Panel Discussion and Audience Q & A
4:30 p.m. Closing remarks with Tom LoFaro, Karl Larson and Lisa Heldke