Four faculty and four students across the Cornell Ann S. Bowers College of Computing and Information Science were selected in the first round of grants from the new five-year, multimillion-dollar partnership established between the college and LinkedIn.
The awards will fund innovative research ranging from the development of foundational advances in machine learning, to improving equity, fairness, and privacy in computing methodology and practice. Through these grants, the first-of-its-kind partnership opens new channels for communication and cooperation between scientists at Cornell Bowers CIS and LinkedIn that will yield high-impact research in these areas.
“We are excited about this long-term partnership with LinkedIn,” said Kavita Bala, dean of Cornell Bowers CIS. “This targeted funding creates deeper relationships through research and networking activities, so we will come together to talk about important research problems that span industry and academia.”
Dozens of faculty and students applied for the funding and the successful recipients represent research expertise across all three Bowers CIS departments. The selection was chaired by Thorsten Joachims, associate dean for research at Cornell Bowers CIS. “It is exciting to see the breadth of faculty and student interest,” Joachims said, “and how this partnership can inform the real-world impact of our research.”
“When industry and academia collaborate to build our collective knowledge, everyone benefits," said Ya Xu, head of Data and AI at LinkedIn. "We're excited to embark on this journey with Cornell Bowers CIS, driving forward research in Data Science and AI, with a focus on improving equity."
Cornell Bowers CIS faculty received grants to support the following projects:
• Emma Pierson, assistant professor of computer science at Cornell Tech: Training the engineering workforce to develop fair algorithms
In a large-scale, randomized controlled trial, Pierson will investigate how best to train engineers to design fairer algorithms. The work will result in a publicly available algorithmic fairness class for engineers, available through Coursera.
• Karthik Sridharan, associate professor of computer science: Reinforcement Learning for Optimizing Long-Term and Short-Term Costs
Currently, machine learning-based systems are good at accomplishing short-term goals, like achieving high click-through rates, but less effective at long-term objectives, like enhancing user engagement. This project aims to develop new machine learning methods that optimize both long- and short-term goals, while preventing polarization and bias, which can occur over time with such systems.
• Allison Koenecke, assistant professor of information science: Early Stoppage of Randomized Controlled Trials on Heterogeneous Populations
Researchers often end clinical trials or A/B tests early when statistical analysis shows a clear benefit to a medical treatment. However, methods for deciding when to stop a trial early may disadvantage minority groups, which typically make up a small percentage of trial participants. Koenecke proposes to develop fair rules for stopping a trial and determine how to implement those rules in a straightforward way in real-world settings.
• David Matteson, associate department chair and associate professor of statistics and data science: Deep Generative Models for Large-scale Ranking and Temporal Datasets
For this project, Matteson will use a type of machine learning approach called Generative Adversarial Networks to correct the issues of missing and biased data in large datasets. Additionally, he aims to ensure fairness and privacy through these methods.
For the student awards, selected Ph.D. candidates will receive grants to cover one year of their research. These awards can help the students gain valuable experience in applying for and managing grant funding.
The following students were selected:
· Ruihan Wu, a computer science doctoral student advised by Kilian Weinberger, works on identifying and solving security and privacy problems in machine learning. Her research will deal with protecting data privacy when analyzing social networks and limiting negative effects from “cheaters” who create fake accounts on peer recommendation sites.
· Marios Papachristou, a doctoral student in computer science advised by Jon Kleinberg, studies the dynamics of real-world social networks, looking at how income shocks move through financial networks as a type of contagion. This work was previously used to determine an equitable system for bailouts and could be applied to other networks involving mismatched supply and demand, such as ridesharing apps, high-performance computing, ad placement, and supply chains.
· Anthonia Carter, a doctoral student in information science advised by Christobal Cheyre, is designing social and technological solutions for allocating money more equitably to under-resourced communities. She will be looking at how investing in emerging venture capitalists and first-time startup founders, especially those from historically marginalized communities, may be an opportunity to diversify investment portfolios and achieve increased returns.
· Kimberly Hochstedler, a doctoral student in statistics and data science advised by Martin Wells, investigates the imperfect algorithms commonly used to classify individuals in the criminal justice system, healthcare, and in business hiring practices. She is working on a bias correction strategy to correct misclassifications and to understand bias in social systems.
A kickoff event for the strategic partnership will occur in September, where researchers at Cornell Bowers CIS and LinkedIn will gather in Ithaca. A retreat is planned for the following May, where award winners will be invited to present research findings.
Written by Patricia Waldron, a science writer for the Cornell Ann S. Bowers College of Computing and Information Science.