Collaboration with Humanities and Social Sciences

I have established collaborations with researchers in Humanities and Social Sciences, combining a computer science perspective with legal or economic issues. Some of these collaborations have led to publications and communications at conferences or colloquia. Others are still in progress.

Thanks to my participation in Bordeaux’s Digital Convergence Colloquia, I was able to meet legal researchers, including, with whom I worked on the legal aspects of smart contracts.

Competition disturbance by Artificial Intelligence

I’m interested in a phenomenon little exploited in Intelligence Artificial: the possibility for an algorithm to collude with another to artificially increase prices.

Consider n continuously updated AI models obtained by reinforcement learning, whose objective function is to maximize the profit of the firms that use them. Under certain experimental conditions, we have shown that it is possible that the increase in the number of competitors using the algorithm could increase prices, this behavior being akin to collusion, therefore an illegal practice, because it distorts competition.

We seek to formulate what would be the right restrictions to place on the algorithms used for pricing, and to improve the legislator’s ability to detect tacit agreements, through data analysis and statistical tools.

Checking GDPR Compliance of Blockchain Processes

This work in progress is guided by the need for today’s information systems to comply with standards for data protection. This need must be put in balance with competitiveness increases thanks to the adoption of new technologies, such as blockchain, providing guarantees in terms of data integrity.

In order to better understand the data flows, we propose the use of process mining technique (part of the “Business Process” management techniques located at the crossroads of data mining (or data mining) and analysis of business processes based on the analysis) on execution traces. It makes it possible to discover, analyze and improve existing processes and has already been used for the extraction of structured information in blockchains. This technique is also used in a more traditional information system context to verify GDPR compliance, and in particular for the right to be forgotten. The contribution of such a technique is to check whether the information systems using a blockchain actually comply with the standards. Initially, we will be able to verify compliance from execution traces, then propose solutions for correcting non-conformities in business processes and finally proceed to validate the process thus described.

collection and use of multimedia streaming consumption data

As part of the PCEN chair (Cultural Pluralism & Digital Ethics), I won a Paris 1 call for projects with researchers in economics aimed at extracting consumption data from a streaming platform with the aim of validating the hypotheses concerning the highlighting of content of European origin to comply with the SMA directive imposing a quota of community content on online platforms. In this context, we have developed a multi-platform distributed webscrapping solution, to collect data and proceed to their analysis.

I extended this project at the CRI in 2020-2021 by setting up a team made up of L3 and M1 students for the technical part, as well as 2 research master’s students for the data analysis part. We have addressed two important questions regarding platform content:

  • Can process mining be used to model the compliance of an online delivery platform’s recommendations with the SMA directive?
  • Have we won the war against conspirational content?

Based on two datasets created using the webscrapping solution, we employed AI algorithms (fastText) to verify our hypotheses regarding the creation of filter bubbles on YouTube.

We then generated process traces of the recommendation algorithm, analyzed by contextual process mining techniques, in order to establish a new compliance metric with the SMA directive, taking into account the highlighting of content on the platform.

The writing of articles from the research papers of M2 students is in progress.