This work is the result of a collaboration between researchers at the Chaire PcEn and the students Jordan Louis, Laura Swietlicki, Marvin Mayard and Thomas Roux of the Master MMI (Interactive Multimedia) directed by David Bihanic at the University Paris 1 Panthéon-Sorbonne. Based on data collected with a Chrome extension developed by Nicolas Herbaut, associate professor in computer science, the students had the task of reflecting on their different levels of exploitation and visual processing. In addition to the graphic projection of the data through visualisation techniques, the students were asked to suggest creative means of "highlighting" the data. The data sets made available to them included, for several Netflix users, the exact position of each content on their home page, which allowed students to reconstitute the visual hierarchy, specific to each profile, in the contents that appeared on-screen.
The critiria on which the series and movies are selected are quite obscure. Nonetheless, through the content they choose to highlight or neglect, it is not less than 19 millions of users per month that are influenced by the US streaming giant. The desires and thoughts of the user are incentivized by the thumbnails they see everytime they browse the website. However, the movies and series featured on the user's screen have been carefully put together beforehand. Consequently, the user is then subjected to a pictorial dictatorship: what they will watch is already chosen for them. The impact of these choices can be quite significant, which is why we can question them and their meaning.
With the data that we collected from the PcEn (laboratory of research from l'École des médias et du numérique de la Sorbonne), we wish to depict the evolution of recommendation of movies and series over time (Prime Time) and space (Prime Space) in the interface of the home page of Netflix. Most importantly, we want to draw attention to the different variables (date of release, IMDB rating, whether or not it is a Netflix production as well as the type of content) took into consideration for the algorithm of Netflix. We can then better analyze a specific interface output from the algorithm.