Abstract
The increasing relevance of platformization sheds light on the role of algorithms in filtering political content, profiling audiences and defining the rules for the competition between traditional outlets and new content creators online. More importantly, algorithms learn and adapt results based on users’ activities online. But, if algorithms learn over time, how to deal with this time-varying dynamic when analysing them? The present paper brings a method for analysing YouTube search ranking and related video algorithm results over time, applied to a corpus of 1346 videos related to the war in Ukraine connected through 7934 related video links, starting on 21st November and stopping on 5th December. Results show that YouTube search and related video algorithms differ considerably in their behaviours, considering the channels and video clusters they benefited over time. It could be a dangerous bias to focus solely on one of the algorithms or presume its functioning based on collections made after – and not during – the political events they influenced. More than a matter of choosing methods, to understand how algorithms are changing the network structure of the current public sphere, it is important to develop new ones.
Get full access to this article
View all access options for this article.