The Many Dimensions of Data

International Conference on Data Economy

June 22 and 23, 2017
IMT - Télécom ParisTech - 46, rue Barrault, Paris 13

The digital economy is now "data-driven". Big data, algorithms and artificial intelligence are profoundly changing the nature of competition between firms as well as customer relationship management. On the one hand, new data strategies based on smart connected objects create opportunities to challenge existing value chains and force companies to rethink the organization of their industry. Data accumulation by large companies generate increasing returns to scale that increase demand through targeting and recommendations. On the other hand, big data can improve product and service experience but raise at the same time concerns for consumers. For instance, business models based on advertising are both creating new ways to monetize audience but raise privacy, autonomy and trust concerns.

This international conference will focus on two dimensions of the data economy from a pluridisciplinary perspective. First, the event is part the official program of the French Data Protection Authority (Cnil) on ethics and algorithms and will highlight several keynotes on the topic. The second dimension that the conference will explore is the free flow of data in the digital economy, focusing on economic, legal and political aspects.

Thursday Program (13:30-18:00)

Online crime and security economics
Download the presentation

Nicolas Christin
Carnegie Mellon University

Over The Top (OTT) Video: Policy and Privacy Issues

Marvin Sirbu

Carnegie Mellon University

Accountability for Fairness in Data-Driven Systems

Anupam Datta

Carnegie Mellon University, Sillicon Valley

Our vision is to enable data-driven systems to be accountable for fairness properties. We use the term “accountable” to refer to computational mechanisms that can be used to “account for” behaviors of systems to support detection of fairness violations, as well as explain how they came about. We then leverage this understanding to repair systems and carefully account for the absence of violations.

As an illustration of this paradigm, I will introduce a notion of proxy non-discrimination and methods for making machine learning models accountable for this property. A key building block for these notions is Quantitative Input Influence -- a form of explanation for decisions of machine learning models. This work is motivated by recent discoveries of unfairness or bias in widely used data-driven systems, including our own work that discovered gender bias in the targeting of job-related ads in the Google advertising ecosystem.

Friday Program (09:00 - 12:30)

Industry 4.0 and the IoT: A View from Europe

Yann Ménière

European Patent Office, Chief economist

The current Debate on Data Ownership and Access to Data: First lessons from in-depth Industry Interviews

Klaus Wiedemann

Max Planck Institute for Innovation and Competition

Franziska Greiner

Max Planck Institute for Innovation and Competition


Update on recent developments in EC data policy

Kristina Kjerstad

European Commission
DG Communications Networks, Content and Technology
Data Policy and Innovation Unit

Commentaires Clos.