
Generated by Microsoft Copilot. "AI with Smart Watch and Stethoscope." 2024.
Artificial Intelligence (AI) and Machine Learning (ML) are being used more and more across nearly every aspect of our lives. This is in part because such services are more readily available, and also because their predictions are improving. Such increasing ubiquity of our seeming to give over our decision-making authority to data-driven algorithms brings to mind the specter of a computer overlord without humanity’s best interest in mind, such as Skynet as imagined in the Terminator movies. Wisely, we humans are increasingly interested in guardrails and guidelines for AI and ML. As has already been considered in many areas of science, these should be informed both by potential benefits and harms. AI and ML products should be providing accurate out-of-sample estimation/prediction for the types of problems claimed to be in-scope for that product. They should also respect privacy and lead to improved experiences that are fair and just.
These are the types of problems that Department of Statistics faculty Subhadeep Paul and Arnab Auddy will be able to explore due to their new College of Arts and Sciences (ASC) Natural and Mathematical Sciences (NMS) Exploration Grant. This nearly $50K grant will enable them to support a Graduate Research Associate (GRA) and provide materials and supplies to pursue this work.
To ground their work in real world practice, they plan to focus on two use cases. Electronic Health Record (EHR) databases collected at most modern health care facilities can be used to advance precision medicine. To maximize this potential, we must find ways to continue to protect patient privacy through limits on sharing patient data and ways to harmonize measures or records of similar patient characteristics across different health systems. Wearable mobile devices pose similar challenges for harnessing the power of large-scale data analysis. For example, Apple Intelligence claims its tools “[d]raw on your personal context without allowing anyone else to access your personal data — not even Apple”. Subhadeep and Arnab will be working with their team to advance solutions to these challenges with new statistical tools related to heterogeneous knowledge transfer, federated in-device ML, and privacy preserving AI training.
This work is supported by the NMS Exploration grant with matching funds committed by the Department of Statistics in part through the generosity of donors to the 50th Anniversary Statistics Graduate Student Support Fund and Graduate Fellow Fund.