Second round of SERC seed grants awarded to MIT scholars

The MIT Schwarzman College of Computing’s Social and Ethical Responsibilities of Computing (SERC) has awarded a second round of seed grants. These grants aim to inspire MIT researchers to make a meaningful contribution to responsible technology development and deployment.

This year’s call for proposals covered a range of topics, from AI policies in the classroom to forecasting air pollution. Eight projects, led by MIT researchers from seven departments across all five schools and the college, have been selected to receive up to $75,000 in funding.

“We were thrilled to receive over 65 proposals for this second round of seed grants, reflecting the depth and diversity of interest in the social and ethical responsibilities of computing across MIT,” says Nikos Trichakis, associate dean of SERC in the MIT Schwarzman College of Computing and J.C. Penney Professor of Management. “Submissions came from all corners of the Institute and were reviewed by two faculty panels representing all five schools and the college. In that sense, the selected projects reflect what MIT collectively considers some of the most promising and impactful research currently underway in this critical area.”

The eight projects and research leads are:

• “Label-Efficient Selection and Verification of AI Models,” led by Sara Beery, proposes an efficient framework to assess machine learning models for the best performance for a given dataset and target goal.
• “Show, Tell, and Adapt: Aligning Robot Behavior via Multimodal Communication,” led by Andreea Bobu, proposes a framework that bridges the communication gaps between robots and human expectations via multimodal explanation and active learning.
• “Toward Responsible, Virtue-Based AI Policies in Schools,” led by Cynthia Breazeal and Daniella DiPaola, proposes a virtue-based approach to AI governance and usage in K-12 schools with educational tools that enable communities to make decisions based on shared values.
• “A Framework for Machine Learned Forecasts of Air Pollution in the Contiguous United States,” led by Tamara Broderick and Arlene Fiore, proposes the building and implementation of machine learning graphical interfaces to increase the progress, accuracy, and reliability of air quality forecasting.
• “Ethical Frameworks for Responsible Deployment of Computer Vision in Public Interactive Installations,” led by Joshua Higgason, proposes the development of ethical frameworks to guide the implementation of computer vision technology in public interactive installations.
• “Addressing Regressive Property Taxation with Fairness-Constrained Differentiable Machine Learning,” led by Haihao Lu, proposes the development of a novel machine learning pipeline to mitigate property tax regressivity, which disproportionately affects lower-income property owners.
• “Personal Benefit, Collective Harm: Differential Impact of AI on Individual and Crowd Wisdom Judgment,” led by Drazen Prelec and Eric So, proposes to glean a deeper understanding of reliance on AI, and whether an AI can be adapted to serve the interests of a collective rather than a single user.
• “A New Criterion for AI Alignment: Moral Meshing,” led by Bernardo Zacka and Bailey Flanigan, proposes an examination and further understanding of how AI systems can fail to mesh with our moral values when transitioning between bureaucratic to algorithmic decision-making and how they can be modified to ensure a better fit.