Workshop Details

For the past 15 years, my team and I have been advancing AI and multiagent systems research towards social impact, focusing on topics of public health, conservation and public safety. We have focused on addressing a key cross-cutting challenge: how to effectively deploy our limited intervention resources. In this talk, I will present results from work in using AI for addressing challenges in public health such as Maternal and Child care interventions, HIV prevention, and TB prevention. Achieving social impact in these domains often requires methodological advances. To that end, I will highlight key research advances in multiagent reasoning and learning, in particular in, restless multiarmed bandits, influence maximization in social networks, and decision-focused learning. In pushing this research agenda, our ultimate goal is to facilitate local communities and non-profits to directly benefit from advances in AI tools and techniques.

[Video]   [Slides]

Speaker Bio: Milind Tambe is Gordon McKay Professor of Computer Science and Director of Center for Research in Computation and Society at Harvard University; concurrently, he is also Principal Scientist and Director "AI for Social Good" at Google Research. He is recipient of the AAAI Feigenbaum prize, IJCAI John McCarthy Award,  AAMAS ACM Autonomous Agents Research Award, AAAI Robert S. Engelmore Memorial Lecture Award, and he is a fellow of AAAI and ACM. He is also a recipient of the INFORMS Wagner prize for excellence in Operations Research practice and Rist Prize from MORS (Military Operations Research Society). For his work on AI and public safety, he has received Columbus Fellowship Foundation Homeland security award and commendations and certificates of appreciation from the US Coast Guard, the Federal Air Marshals Service and airport police at the city of Los Angeles.

[Presentation Slides]

AI  can help face healthcare needs and challenges, being the catalyst for a profound transformation in the healthcare arena  and helping improve efficiency, effectiveness, and responsiveness, as well as equity in the delivery of public health and healthcare services, by developing and facilitating current practices, and by introducing new methods of surveillance and action both at the individual (clinical medicine) and community/population (public and global health) levels.   AI is anticipated to uncover new links between climate, climate-related disaster exposure, and the burden of disease (especially, in terms of mental health), helping policy- and decision-makers particularly in low- and middle-income countries (LMICs), such as those belonging to the Global South to achieve SDG3 ("health and wellbeing for all") and SDG5 ("gender equality") . In this talk, I will present an AI-based framework and AI-based algorithms that we designed for achieving SDG3 and SDG5  in the Global South, leveraging and capitalizing on our experience with the “Africa-Canada Artificial Intelligence and Data Innovation Consortium” (ACADIC) Project.

Ensure healthy lives and promote well-being for all at all ages is one of the United Nations Sustainable Development Goals (SDGs). Among the health problems, depression is a serious mental health disorder that can have a significant impact on individual's well-being and quality of life. Individual factors and urban environments play an important role for psychological well-being, depression and stress. However, how the daily experienced urban environment is related to mental well-being remains an open question. To address this issue, my recent research has focused on investigating how microscale urban environments (such as the design of urban and living spaces), travel modes, and daily activities impact momentary well-being in elderly individuals. This was achieved by combining data from GPS trackers, smartphone questionnaires on mental health, socio-economic surveys, and geographic information on daily exposure environments, using a Geographically explicit Ecological Momentary Assessment (GEMA) approach. Results from the HANC-Mindmap project will be presented.

Online platforms have a wealth of data, run countless experiments and use industrial-scale algorithms to optimize user experience. Despite this, many users seem to regret the time they spend on these platforms. One possible explanation is misaligned incentives: platforms are not optimizing for user happiness. We suggest the problem runs deeper, transcending the specific incentives of any particular platform, and instead stems from a mistaken revealed-preference assumption: To understand what users want, platforms look at what users do. Yet research has demonstrated, and personal experience affirms, that we often make choices in the moment that are inconsistent with what we actually want. In this work, we develop a model of media consumption where users have inconsistent preferences. We consider an altruistic platform which simply wants to maximize user utility, but only observes user engagement. We show how our model of users' preference inconsistencies produces phenomena that are familiar from everyday experience, but difficult to capture in traditional user interaction models. A key ingredient in our model is a formulation for how platforms determine what to show users: they optimize over a large set of potential content (the content manifold) parametrized by underlying features of the content. Whether improving engagement improves user welfare depends on the direction of movement in the content manifold: for certain directions of change, increasing engagement makes users less happy, while in other directions, increasing engagement makes users happier. We characterize the structure of content manifolds for which increasing engagement fails to increase user utility. By linking these effects to abstractions of platform design choices, our model thus creates a theoretical framework and vocabulary in which to explore interactions between design, behavioral science, and social media. By improving our understanding of how algorithms and platforms interact with human psychology, we can work towards societal goals of health and well-being.

[Presentation Slides]

Computing is critical to achieving the UN SDGs (hence this workshop :). Yet, computing also has a large impact on the environment that negatively impacts at least several of the SDGs. Examples of computing’s impacts on the environments include: energy consumption leading to greenhouse gas emission, water consumption, and e-waste. In this talk, I will briefly review datacenters’ energy consumption and corresponding greenhouse gas emission and work that has been done to increase energy efficiency and reduce emission. I will also touch very briefly on datacenters’ water consumption.

[Presentation Slides]

The UN Goal to make cities and communities sustainable emphasizes the importance of affordable housing. Yet, millions of Americans and hundreds of millions worldwide experience homelessness each year. The scarcity of housing assistance to relieve the burden of homelessness presents technical and ethical challenges for policymakers. Findings from a community-based research initiative demonstrate how algorithmic decision-making can incorporate local priorities to reduce the incidence of homelessness equitably across subpopulations. The transparent and replicable decision-making process provides a sustainable approach to help achieve the UN aim of leaving no one behind.

Ensuring the fairness and equitability of algorithms for allocating public resources is challenging since the definition of fairness is highly intersectional, multi-modal, and domain-specific. In the context of limited resource allocation, we present new results showing that these definitions cannot hold simultaneously in the presence of heterogeneous responses to different “treatments” among different groups. Using real-world administrative records, we demonstrate the fairness trade-offs stemming from such heterogeneity across groups in responses to homeless services. Our findings raise concerns about seemingly natural fairness metrics in the design of policies to address homelessness and in social policy more generally, especially when stakeholders can have competing objectives.

Sera Linardi is the Founding Director of the Center for Analytical Approaches for Social Innovation (CAASI) at the University of Pittsburgh. The center connects students across disciplines to local social justice movements by incubating student-driven data science projects from community requests. This talk will cover the two different ways the UN Sustainable Development Goals are used by CAASI's nonprofit partners. The first approach utilizes the interconnection between all 17 goals to cross reference multiple community indicators. The second focuses tightly on a single SDG to elevate a hyperlocal issue to the international stage.

Artificial Intelligence (AI) is a rapidly advancing field. Novel machine learning methods combined with reasoning and search techniques have led us to reach new milestones: from computer vision, machine translation, and Go world-champion level play, to self-driving cars. These ever-expanding AI capabilities open new exciting avenues for advances in new domains. I will discuss our AI research for advancing scientific discovery for a sustainable future. In particular, I will talk about our research in a new interdisciplinary field, Computational Sustainability, which has the overarching goal of developing computational models and methods to help manage the balance between environmental, economic, and societal needs for a sustainable future. I will provide examples of computational sustainability problems, ranging from biodiversity and wildlife conservation to multi-criteria strategic planning of hydropower dams in the Amazon basin and materials discovery for renewable energy materials. This work was featured in the Communications of ACM, in a cover article entitled Computational sustainability: computing for a better world and a sustainable future . I will also talk about our work on AI for accelerating the discovery for new solar fuels materials, which has been featured in Nature Machine Intelligence, in a cover article entitled, Automating crystal-structure phase mapping by combining deep learning with constraint reasoning. In this work, we propose an approach called Deep Reasoning Networks (DRNets), which requires only modest amounts of (unlabeled) data, in sharp contrast to standard deep learning approaches. DRNets reach super-human performance for crystal-structure phase mapping, a core, long-standing challenge in materials science, enabling the discovery of solar-fuels materials. DRNets provide a general framework for integrating deep learning and reasoning for tackling challenging problems. For an intuitive demonstration of our approach, using a simpler domain, we also solve variants of the Sudoku problem. The article DRNets can solve Sudoku, speed scientific discovery, provides a perspective for a general audience about DRNets. Finally, I will highlight cross-cutting computational themes and challenges for AI at the intersection of constraint reasoning and deep learning. 

[Video]   [Slides]

Speaker Bio: Carla Gomes is the Ronald C. and Antonia V. Nielsen Professor of Computing and Information Science, the director of the Institute for Computational Sustainability at Cornell University, and co-director of the Cornell University AI for Science Institute. Gomes received a Ph.D. in computer science in artificial intelligence from the University of Edinburgh. Her research area is Artificial Intelligence with a focus on large-scale constraint reasoning, optimization, and machine learning. Recently, Gomes has become deeply immersed in research on scientific discovery for a sustainable future and, more generally, in research in the new field of Computational Sustainability. Computational Sustainability aims to develop computational methods to help solve some of the key environmental, economic, and societal challenges to help put us on a path toward a sustainable future. Gomes was the lead PI of two NSF Expeditions in Computing awards. Gomes has (co-)authored over 200 publications, which have appeared in venues spanning Nature, Science, and a variety of conferences and journals in AI and Computer Science, including five best paper awards. Gomes was named the “most influential Cornell professor” by a Merrill Presidential Scholar (2020). Gomes was also the recipient of the Association for the Advancement of Artificial Intelligence (AAAI) Feigenbaum Prize (2021) for “high-impact contributions to the field of artificial intelligence, through innovations in constraint reasoning, optimization, the integration of reasoning and learning, and through founding the field of Computational Sustainability, with impactful applications in ecology, species conservation, environmental sustainability, and materials discovery for energy” and of the 2022 ACM/AAAI Allen Newell Award, for contributions bridging computer science and other disciplines. Gomes is a Fellow of the Association for the Advancement of Artificial Intelligence (AAAI), a Fellow of the Association for Computing Machinery (ACM), and a Fellow of the American Association for the Advancement of Science (AAAS).

The question whether a given economic mechanism or computer science algorithm can help achieve the UN sustainable development goals might, at first sight, seem to be very complex, but clear. If progress is made towards achieving one or several of these goals, it is bound to be difficult to impute a causal role to a given economic mechanism or computer science algorithm in such achievements. But at least the question seems to be clear enough. This seeming transparency is deceptive, for two main reasons. A first reason stems from the fact that, while some UNSD goals are articulated as envisioned end-states, such as the end of poverty, others are articulated as endeavors, such as “promote peaceful and inclusive societies”. A second reason is that, because answering the above imputation question is so complex, spending too much time and energy to answer it might be counterproductive. Therefore, philosophical reflections are needed to clarify the kind of contributions to these goals that can be expected from products of scientific research. I will argue that, if one wants to claim that a given economic mechanism or algorithm contributes to achieving this or that development goal, one should be prepared and willing to justify this claim publicly, that is, to argue in favor of that claim publicly, using a very unusual kind of rhetoric, based on actively searching for counterarguments aimed at undermining the claim at issue.

Popular discourse about extreme weather events often invokes climate change as a cause.  But extreme weather events have occurred throughout the time for which high-quality records of weather and climate are available, including well before the global warming trend accelerated in the late 20th century.  Quantifying the effects of climate change on the likelihood of extreme events has been challenging, but methods have now been developed for determining the contribution of climate change to individual extreme events.  These "event attribution" methods will be briefly discussed along with some examples.

[Presentation Slides]

I'll speak about a few ongoing and potential future directions where algorithmic mechanism design might be helpful toward the UN Sustainable Development Goals.  These include an ongoing project with the Mainstreet Districts in the City of Boston, the "money burning" objective which is highly relevant in healthcare, and more.

Voting is a widely-used methodology to make a collective decision for a group of agents. In this talk, I will introduce a new notion of group fairness, present a machine learning framework and a constrained optimization framework to design new voting rules that are fair and efficient, and illustrate the three-way trade-off between economic efficiency, fairness, and privacy.

The talk is based on Farhad Mohsin, Ao Liu, Pin-Yu Chen, Francesca Rossi and Lirong Xia. Learning to Design Fair and Private Voting Rules. Journal of Artificial Intelligence Research. 2022.

[Presentation Slides]