2019-01450 - Post-Doctoral Research Visit F/M Modeling the filter bubbles from recommendation algorithms

Contract type : Public service fixed-term contract

Renewable contract : Oui

Level of qualifications required : PhD or equivalent

Fonction : Post-Doctoral Research Visit

About the research centre or Inria department

The Inria Rennes - Bretagne Atlantique Center is one of Inria's eight Centers and has more than thirty research teams. The Inria Center is a major and recognized player in the field of digital sciences. It is at the heart of a rich R&D and innovation ecosystem: highly innovative PMEs, large industrial groups, competitiveness clusters, research and higher education players, laboratories of excellence, technological research institute, etc.

Context

Inria Rennes is seeking a postdoctoral researcher for joining the WIDE team, which gathers 12 researchers (faculty, and PhD students) in the field of distributed systems and algorithms, with a focus on privacy preserving and scalable computation.

Assignment

Recommendation algorithms are pervasive in our online interactions:
they are personalizing the web browsing and shrink the flow of
information presented to a user. Numerous works from academia have
been specifically dedicated to the improvement of those
algorithms. Yet there are important complaints about the so called
"filter bubbles" or "echo chambers" that those algorithms are
producing to their users (e.g., loss of content diversity [1]). Very
few works are focusing on the user side: what is the information that
a curious user might gather from the observation and the analysis of
the recommendations he/she gets? Those algorithms can solely be
interacted with by the user in a black-box manner; yet, researchers
[3,4] have shown in another context that information indeed leaks from
online black-box algorithms. We have shown in [2], regarding the
observations and analysis of the YouTube recommendations to a given
user, that those recommendations can be captured under a graph
structure, and that this structure allows for an analysis with complex
network algorithms.

The postdoctoral researcher will work on the design of algorithms for
the user-sided quantification of the filter bubble effect.

He/she will for instance address the following research questions:

• What is the most convenient and powerful data structure for the modeling
of recommendations made to users, and to quantify the filter bubble effect?
• Once the data structure is fixed, which algorithms to analyze it? Are
existing ones powerful enough, or do we need to re-design new ones?
• Can we address the problem with statistical techniques such as ”capture-
recapture” for the observation of specific types of recommendations?
• Can we provide impossibility proofs, by making for instance the parallel
with differential privacy, that no information leak can occur for certain
classes of recommendation algorithms?

[1] Exploring the filter bubble: the effect of using recommender systems on
content diversity, Tien T. Nguyen and Pik-Mai Hui and F. Maxwell Harper and
Loren Terveen and Joseph A. Konstan, WWW 2014.
[2] The Topological Face of Recommendation, Erwan Le Merrer and Gilles
Trédan, Complex Networks 2017.
[3] Stealing machine learning models via prediction APIs, Florian Tramr
and Fan Zhang and Ari Juels and Michael K. Reiter and Thomas Ristenpart,
USENIX Security 2016.
[4] Inferring Networks From Random Walk-Based Node Similarities, Jeremy
Hoskins and Cameron Musco and Christopher Musco and Charalampos Tsourakakis,
NIPS 2019.

Main activities

Review current related work proposals.
Model and design alternative methods to simple statistical measures.
Implement algorithms using common libraries for contribution to open-source software.
Coordinate with researchers in the team to participate to the effort on the topic.

Publish and disseminate results.
Program the algorithms for possible integration to web-browser plugins.
Coordinate with related efforts in the team / community.

Skills

Strong background in mathematics (e.g., graph theory, linear algebra, probability/statistics) and algorithms.
Excellent English writing and speaking skills are required.

Benefits package

  • Subsidized meals
  • Partial reimbursement of public transport costs
  • Leave: 7 weeks of annual leave + 10 extra days off due to RTT (statutory reduction in working hours) + possibility of exceptional leave (sick children, moving home, etc.)
  • Possibility of teleworking (after 6 months of employment) and flexible organization of working hours
  • Professional equipment available (videoconferencing, loan of computer equipment, etc.)
  • Social, cultural and sports events and activities
  • Access to vocational training
  • Social security coverage