fedebarba
commented
7 years ago To start we need to visualize all the feeds impressed on our user's timelines. We are now dividing for type of feed to see how the algorithm behaves with different ways of information. The number at the end of the bars represent the average impression order. Every post is impressed on a timeline with a specific order that change every time a timeline is refreshed. To analyse it means understand the life of a feed, for how long has this feed been at the top? Also more important, why are some feeds at the top and other on the bottom? Is it just a chronological order?
Here it is possible to see for how may time a feed is impressed on a specific timeline, in this case Camila. The impression day represent when feeds had been impressed in Camila's timelines
How to know if the feeds you see on your screen are what you were really looking for? How those feeds are selected, filtered and then showed? What we want to understand is why some posts appear and others don't. To do that we need to get deeper in the aspects determining the success of a post on others, understand what the algorithm considers like a weakness and what like a strength. A quantitative analysis is necessary to test if it is true that more likes/shares/AnyOtherInteractions to a feed make it more highlighted by Facebook for users. In case this relation does not find match with data, which other quantitative element can be part in this huge battle to appear. The qualitative part of this study tries to determine how the algorithm reacts to the external inputs. Is the algorithm filtering some issue just because the country is in an electoral period? Or is the algorithm selecting a specific issue to determine what the political speeches have to be about? It can be a two-way relation between Facebook and the real word. Is Facebook so strong to determine the political agenda of a country or it is just filtering a determined political agenda?