**Quick summary:**
The overarching goal of my research and this NSF CAREER project is to reconcile the risk and value of personal data networking, by designing algorithms leveraging social mobile behaviors and incentives.

**Brief thematic overview**:
Readers, bloggers, shoppers, are constantly seeking new information and sharing them using social networking tools. The mundane expectation is that by accessing and sharing more data, you certainly spend time and expose yourself but you are better informed to make important decisions. In other words, it?s worth the effort and the risk. But is it? How much are we actually aware about concrete risks? And are we really seeing that this process rewards its participants efficiently? What about when the data are collected and used without your awareness and consent?

Behavioral analytics leverage personal data about user behaviors collected and used for personalization, decision and optimization. It is arguably a historic driver of societal progress, but also one of the most controversial. It already challenges core values such as personal privacy and societal fairness in equal access to content and services.

Behavioral analytics so far are hindered by two main defects:
- it provides *no transparency*: users are releasing information and permissions to use data without being aware of the induced risks and how the information they provide affects their service. At best they have crude information and guideline about which data are known and rarely how they are used.
- it provides *no incentive*: users are generally facing a binary choice, convenience of a service free of charge in exchange of full data disclosure. At best, they are offered a coarse choice of opting-out of some personalization features.

This project, in contrast, works on providing transparency and incentive in today?s behavioral analytics:
(Th.I) Quantify the threats posed by personal data disclosure across multiple domains, in the face of emerging plausible risks: re-identification across domains and discrimination.
(Th.II) Build transparency tools informing users on what their data reveal, how their data are being used, and techniques guaranteeing a fair use of their data.
(Th.III) Design incentives for systems leveraging users participation and/or data, to align individual?s interest with global efficiency.

Outcomes:
- This 5+ year effort lead to 14 papers in tier-1 conferences like ACM SIGMETRICS, Web Conf, COSN, CCS, USENIX Security, multiple ones received nomination and awards. Those results were picked up by multiple media outlets (including, each for a different result, in The Washington Post, The New York Times bits blog, The Economist, MIT Technological Review and the Times of India). One article was downloaded more than 48,000 times, most of those articles are already cited, three of them more than 100 times.
- Six Ph.D students including 3 female and 1 Hispanic received direct support for their research through this project. Nine undergraduate students, including 5 female, 1 African American, 2 Hispanic conducted research and authored papers. One of them was a finalist (top-4) for the CRA Undergraduate Researcher award. Three of those are currently enrolled in Ph.D (Stanford and Berkeley).
- This research lead to new educational initiative, including three new tutorials given at ACM SIGMETRICS and The Web Conference, a revamped social network class to address the urgent needs to incorporate ethics of algorithmic impact, as well as educational project in collaboration with the school of Journalism and Social Work at Columbia.

Last Modified: 07/21/2019
Modified by: Augustin Chaintreau