A Recommender System analysis framework

Here we displasy our first findings on Youtube.
This run on free software, donate your data to the project.

SPOILERS:

  • 01. Why Youtube?

    Billions of hours automatically recommended by the algorithm

  • 02. Journals & Research

    Empirical evidences and controversial academic papers

  • 03. Methodology

    API are not a miningfull way to investigate personalization

  • 04. Personalization & polarization tests

    Can we visualize the personalization? Can we study the polarization of UK newpapers around Brexit



Youtube is more watched than TV

More than 1 billion hours of videos a day (2017), 70 percent of the watching time is spended riding a chain of recommendations driven by artificial intelligence.


Recommender System is a Black Box

YouTube published two different papers (1,2) about it's algorithm, but we still don't know how does it works exactelly, it's a Trade Secret!