The DSSG-X public lectures are free to attend. You can register for attendance by emailing us at firstname.lastname@example.org
All of our public lectures for 2021 have now happened. We look forward to next years lecture series.
Macroeconomic Forecasting: a comparison of popular approaches and the impact of the Covid pandemic.
Ana Beatriz Galvao
Abstract: In this talk, I will present the main results of Carriero, Galvao and Kapetanios (2019) [A comprehensive evaluation of macroeconomic forecasting methods, International Journal of Forecasting, 35: 1226-1239], including a discussion on the relevance of mixed-frequency models for nowcasting. I will also describe how the effects of the Covid pandemic on macroeconomic data have affected popular forecasting models' performance. Finally, I will discuss the recent advances that consider non-gaussian disturbances, methods to deal with outliers, and nonlinear machine learning methods.
Ana Beatriz Galvão is Professor of Economic Modelling and Forecasting at Warwick Business School, University of Warwick. She joined Warwick in 2013 from the School of Economics and Finance of Queen Mary, University of London. Ana is a CEPR Research Fellow in the Monetary Economics and Fluctuations Programme. She serves as Associate Editor of the International Journal of Forecasting and the Journal of Applied Econometrics. She is an economic fellow of the Office of National Statistics. She was recently (2020) elected fellow of the International Association for Applied Econometrics. Her recent research on economic forecasting and macroeconomic uncertainty has been funded by the ESRC (2013-2015) and the ONS via Economic Statistic Centre for Excellence (2017-2020).
Responsible data science in the COVID-19 era
Dr David Leslie
Abstract: During the first two waves of COVID-19, data scientists were put under an unprecedented amount of pressure to produce rapid-response insights that could assist clinicians, epidemiologists, and health officials to tackle the pandemic. On the whole, this stress test yielded mixed results. Often pushed beyond the limits of their normal practices, researchers faced myriad challenges around data management, data quality, methodological interoperability, model reporting and validation, consent, algorithmic bias, and interpretability. In this talk, I explore this range of issues, and I offer some constructive steps toward building resilience and readiness in the data science community through the development of responsible data innovation practices and protocols.
Dr David Leslie is the Ethics Theme Lead within the public policy programme at the Alan Turing Institute. Before joining the Turing, David taught at Princeton’s University Center for Human Values (UCHV),and held academic appointments at Yale’s programme in Ethics, Politics and Economics and at Harvard’s Committee on Degrees in Social Studies, where he received over a dozen teaching awards. He now serves as an elected member of the 9-person Bureau of the Council of Europe’s Ad Hoc Committee on Artificial Intelligence (CAHAI)and is the author of the UK Government’s official guidance on the responsible design and implementation of AI systems in the public sector.He is on the editorial board of the Harvard Data Science Review and is a founding editor of the Springer journal, AI and Ethics
The realities of data science no one talks about
Abstract: Data science is massively growing across all industries but despite the buzz, many data science projects in the industry fail to make it into production. This talk will cover many of the challenges, and solutions, faced by the Data Science Team at LV = GI as they grew to become a crucial part of the business. Data science in practice is not just building trendy models; to have an impact we need to ensure that the models are used and continue to perform. As such we need to overcome challenges in 3 key areas; data, technology and people.
Richard Angell is a lead data scientist at LV= GI. Richard has 6+ years experience working in the insurance industry as a data scientist. Whilst at LV= GI, he has been helping to lead the team as they implemented their first model to now having 20+ live across the business, as well as growing the team to 40+.
Fairness & Robustness in Machine Learning for Dermatology
Dr Celia Cintas
Abstract: Recent years have seen an overwhelming body of work on fairness and robustness in Machine Learning (ML) models. This is not unexpected, as it is an increasingly important concern as ML models are used to support decision-making in high-stakes applications such as mortgage lending, hiring, and diagnosis in healthcare. Currently, most ML models assume ideal conditions and rely on the assumption that test/clinical data comes from the same distribution of the training samples. However, this assumption is not satisfied in most real-world applications; in a clinical setting, we can find different hardware devices, diverse patient populations, or samples from unknown medical conditions. On the other hand, we need to assess potential disparities in outcomes that can be translated and deepen in our ML solutions. In this presentation, we will discuss how to evaluate skin-tone representation in ML solutions for dermatology and how we can enhance the existing models' robustness by detecting out-out-distribution test samples (e.g., new clinical protocols or unknown disease types) over off-the-shelf ML models.
Celia Cintas is a Research Scientist at IBM Research Africa - Nairobi, Kenya. She is a member of the AI Science team at the Kenya Lab. Her current research focuses on the improvement of ML techniques to address challenges on Global Health in developing countries and exploring subset scanning for anomalous pattern detection under generative models. Previously, a grantee from the National Scientific and Technical Research Council (CONICET) working on Deep Learning techniques for population studies at LCI-UNS and IPCSH-CONICET as part of the Consortium for Analysis of the Diversity and Evolution of Latin America (CANDELA). She holds a Ph.D. in Computer Science from Universidad del Sur (Argentina).
Using smartphones to study environmental behaviour in a field-experimental setup: A Pilot Study
Ecological sustainability is the defining challenge of our time. Here, a methodological approach is suggested that could help investigate how environmental behaviour (transport behaviour, energy consumption, food consumption, goods consumption, wasting) dilemmas can be overcome on an individual level in real life by using smartphones to collect daily behavioural data field-experimental setup. Results from a pilot study are presented to discuss the feasibility and potential of this approach. The pilot shows that studying social dilemma behaviour via smartphones is feasible and can reveal insights about possible interventions for behavioural change.
Viktoria Spaiser is Associate Professor in Sustainability Research and Computational Social Science at the School of Politics and International Studies, University of Leeds, UK. She is also affiliated with the Leeds Institute for Data Analytics (LIDA) and the Priestley International Centre for Climate in Leeds. Before her time at Leeds, she worked as a postdoctoral researcher at the Stockholm Institute for Futures Studies and the Department of Mathematics, Uppsala University in Sweden. Viktoria Spaiser is interested in sustainability research and specifically how societies can make a rapid, fair and empowering transition to zero-emissions / zero-pollution.
Data Quality in Data Science
Dr. Stephanie Eckman
The insights we get from data science models are affected by the quality of the data that go into the models. Data can suffer from errors of representation (some cases are missing) and errors of measurement (the values in the data set are wrong). Both can cause problems for data scientists training models. I will discuss common data sources and the errors that we find in them. Drawing parallels with survey data collection, I will provide ideas on how we can address these errors to improve data science models.
Dr. Stephanie Eckman is a Fellow in the Survey Research Division at RTI International. She has a Ph.D. in Survey Statistics & Methodology from the Joint Program in Survey Methodology at the University of Maryland. She specializes in understanding data quality and the social construction of data.
Getting Good at Doing "Data for Good"
We know that data science has great potential to contribute to solving social challenges, but for all of the hype around it, that seems easier said than done. We've seen many projects launched using data for good over the last decade, but we seem to still be in search of a grander impact. In this talk, Jake Porway, co-founder and previous Executive Director of DataKind, will talk about what he's learned over years of helping social organizations apply data science to social impact. He'll also talk about how the experiences of the DSSG program translate to the broader world and give a peek of what might be in store ahead for us.
Jake Porway loves seeing the good values in bad data. As a frustrated corporate data scientist, Jake co-founded DataKind, a non-profit that harnesses the power of data science and AI in the service of humanity by providing pro bono data science services to mission-driven organizations. In his decade there he has experienced the trials and tribulations for non-profits, governments, and even corporations in using data and algorithms well. He is now a fellow at data.org, a platform for partnerships foris advancing the field of data science for social impact, where he is currently creating landscapes and frameworks to make sense of this new wild west. He has a bunch of degrees related to mathy things. He lives in upstate NY with his favorite person and two of the worst cats in history.
20th August 2021
Data analysis and modelling to reduce the burden of infectious disease in the ‘bottom billion’ low-income populations
Dr. Deirdre Hollingsworth
Abstract: Neglected tropical diseases are a group of infections which have been grouped together because they predominantly affect the lowest income populations – sometimes called the ‘bottom billion’. This neglect means that there is limited data to inform quantitative analysis. However, modelling can still support policy decisions around optimal use of interventions. In the last year, models have been used to inform estimates of the likely impact of the indirect impact of the COVID-19 pandemic on these infections due to delayed implementation of control strategies. Many of these infections have slow epidemic growth rates, meaning the impact of delays will accrue gradually, but they are also unlikely to be diagnosed through normal health services, particularly during these times. I will review these results and look to the future of how modelling and data science can support these important programmes.
Dr. Deirdre Hollingsworth is an infectious disease epidemiologist who uses mathematical models and statistical analyses to study the evolution and transmission dynamics of infectious diseases with the aim of informing the design of more effective control interventions. She is particularly interested in neglected tropical diseases, a group of diseases which cause suffering amongst the poorest populations of the world. She is a professor at the University of Oxford Big Data Institute where she leads the NTD Modelling Consortium, an international network of neglected tropical disease modellers.
Her research foci are lymphatic filariasis, visceral leishmaniasis and a group of intestinal worms (soil transmitted helminths or STHs) which affect a large number of children and adults in low-income settings. She has ongoing interests in the transmission and evolution of HIV in both Africa and European/North American settings as well as malaria and influenza.