Augustin Chaintreau

Mathematics, and scientifically reproducible datasets, are critical not only for algorithms to improve for all, but also to guide society on the ethical and economic consequences of differentiated treatments led by automatic decisions. Chaintreau joined Columbia after five years of research in industry; since then he works only with public and non-profit funds. Scientists and engineers in industry labs are however frequent collaborators in the joint release of data, tools, or publication of results, as well as investigative journalists.

Today’s big data is flawed, and the threats it poses are not theoretical: Chaintreau has proved with reproducible experiments that personalization algorithms in services used by millions pose moral hazards. His measurements have revealed that metrics of social endorsement misrepresent online attention. As our lives become more digital, Chaintreau has showed that the network dynamics facilitated by online interactions and sharing economies sometimes organically exacerbates various inequalities. Personal information collection and usage, however, ultimately bring benefits that we cannot forego, in various domains such as health, energy efficiency, and public policies. While at first glance, those problems appear embedded in the fabric of Big Data, Chaintreau's work has shown, on the contrary, that these trends can be reversed by addressing some of their common root causes.

What are those causes? First, today’s Big Data is opaque. That is, the algorithms run for personalization may be kept behind closed doors and are unaccountable. Second, it typically presents a fragmented view, where the data needed to answer a question may be large, difficult to collect, and spread across two or more disparate domains. Third, its impact may be felt differently among participants in complex network dynamics, where individual incentives can reinforce unfair outcomes. For each of those hard root causes analyzed by his research, Chaintreau found that mathematical analysis and novel measurement offer a foundation to build a solution that scales: reveal what caused an algorithm output among a growing set of signals, quickly reconnect identical users among the same datasets using a few location points, and mitigate the bad consequence of incentive interaction dynamics in growing networks.

Chaintreau received a BSc in pure and applied mathematics and computer science from École Normale Supérieure (ENS) in Paris in 2001 and a MSc in probability and applications in 2002 from Universit´e Pierre et Marie Curie Paris. He earned a PhD from ENS in 2006 in mathematics and computer science for his work conducted jointly with INRIA. A recipient of a NSF CAREER in 2013, and the ACM SIGMETRICS Rising Star Award that same year. In 2015, he was elected by ACM members of the Computer System Performance Evaluation research community to its board of directors following his service in 28 ACM conference program committees, including two as chair. He also played identical roles occasionally for AAAI and IEEE venues, and served as editor for four ACM or IEEE journals, including one currently as editor in chief. He joined the faculty of Columbia Engineering in 2010. 

Research Areas

• Networks and Distributed Systems
• Data Privacy
• Economics & Computation
• Game Theory
• Networked and Distributed Systems
• Sustainable Computing (Green Computing)
• Algorithm Fairness & Ethics