Funds from this grant support an effort by a new non-profit, Actuate, led by Arati Prabhakar and Wade Shen, to begin to develop, build, and test “DataSafes,” an end-to-end, multi-party, and privacy-preserving system for sharing, linking, modeling, and analyzing sensitive data. Using advanced mathematical techniques like differential privacy and fully homomorphic encryption, DataSafes will provide a platform where data that would be dangerous to share openly can be analyzed in provably safe, trustworthy, and productive ways. Potential use cases include bankers who want to detect fraud patterns across their industry without identifying their clients, doctors who want to quantify disease spread without identifying their patients, or economists who want to identify poverty causes without identifying tax filers. Grant funds will allow Prabhakar and Shen to begin developing the open-source protocols, prototypes, standards, and code libraries that constitute the foundational technological architecture of the platform. Grant expenditures will be overseen by the Hopewell Fund, a fiscal sponsor for Actuate.

Behavioral economics has been highly successful at cataloguing patterns of bias, inconsistency, and even irrationality in everyday human decision-making. One of the great remaining challenges in the field, however, is to present a coherent explanation for why such behaviors exist, to derive the empirical findings of behavioral economics from theoretical models of how our brains work. This grant funds a joint project by economist Michael Woodford of Columbia University and neuroscientist Rava Azeredo da Silveira of the University of Basel, to develop more realistic models of the role that memory limitations play in generating the patterns of economic behavior documented by behavioral psychologists. Woodford and da Silveira will build a set of theoretical models that incorporate the assumption that memory carries a cost. As agents move through time, they create internalized perceptions of the world. Recalling these internal perceptions with flawless precision, they hypothesize, is very costly, and thus as we move from one period to the next, the rational reluctance to pay this “memory cost” produces an increasingly noisy representation of past internal perceptions, resulting in the systematic deviations from optimality observed by behavioral economists. Grant funds will support the theoretical modeling and comparative analysis, as well as a postdoctoral fellow and graduate student who will work on the project.