CAREER: Banalytics: Behavioral Network Analytics with Data Transparency

Columbia University, New York NY

Investigators

Abstract

Data from customers become central to many companies' business, and tomorrow's society's progress critically depends on scientists and public decision makers accessing citizens' data. But how to reconcile progress with privacy? This dilemma, everyday, is getting worse, because we lack an abstraction where analytics -- the science of identifying and exploiting individual types and trends -- are transparent to users and they can effectively own and trade their data. Today's analytics on proprietary data, and its highly debated need for regulation, seems a necessary evil for a lack of credible alternative. This project aims at unlocking this tussle by (1) building behavioral analytics that are compatible with distributed systems of personal data, (2) showing how context and social influence can be better leveraged, and (3) enabling incremental deployment along mutual benefits to make privacy economically efficient. The principal investigator will explore a new abstraction -- Behavioral Networks -- and show its tractability using recent results in matrix factorization, algorithms exploiting social influence, and analysis of networked incentives. Banalytics will demonstrate its applicability to various analytics tasks, including data collection, behavioral targeting, content rating and recommendation. Broader Impact: The need to regulate proprietary access to personal data is heavily discussed towards a new Privacy Bill. Unfortunately, controlling personal data used behind closed doors is highly complex, making enforcement of top-down regulation either ineffective or potentially disastrous for the free and thriving life of the web. The Banalytics project explores and makes available alternative designs empowering users, to inform this important societal debate towards better self-regulation. Not only are sound solutions to reconcile privacy and progress essential, but they need to be disseminated and understood broadly. Banalytics aims to impact the education of students at large -- not only future engineers but also the future journalists informing our citizens -- on the management of personal information and its technical, economic and societal aspects. Through close collaboration with the Columbia School of Journalism, the principal investigator will promote a unique interdisciplinary education program to allow all students and citizens, especially those who might not otherwise express interest in computer science, to be engaged in this effort.