Video and textual tutorial, on the basic usage, the evidence page, and everything a researcher need to start analyzing YT!
A collective observation of the Youtube personalization algorithm; addressing COVID-19 and getting evidence on how recommendations works
Coordinated collection of yt search results; to spot tendencies, bubbles, and biases. real time psedonymized open data feed.
Install youtube.tracking.exposed browser extension
2021 development and expansion: YouChoose.AI.
Youtube.tracking.exposed as backend, have been sponsored by the Next Generation Internet program Ledger. The team has expanded, and we begin to move from showing why an algorithm matter to give back control.
It is in alpha testing, a few aspects still need to be addressed, check the website to get update.
Collective algorithm observations and workshop results
FiterTube: Echo chambers, Filter Bubble and Polarization
During the US elections, we realized a collective project at the Digital Methods Winter School 2021. We simulated echo chambers, we studied the construction of filter bubbles and consequent political polarization of suggestions.
Trexit: Polarization and Reinforcement during Brexit
During Brexit, we made a three days analysis of the algorithm with ten researchers from all over the world. The research aims to split the group in two and see how YT considers different activities to personalize the following recommendation.
Personalization: every second matters
Our first research with a dozen of students: we began by mapping Youtube personalization differences and distances. Watching a video for a few seconds more is enough to get different recommendations. We tested the sperimental 'clean browser'.
Updates and releases
2021
- Webslides: wetest#1 quick summary.
- Workshop: experimental introduction for groups and public events.
- The research paper on on weTest1 gots published.
- Do you know that search results is personalized? You can test it with our experimental tool.
2020
- The year begun with heated debate and articles on academic research using API or non-logged profiles. We were happy! because, someone acknowledges that you can't investigate algorithms without a passive scraper, which is precisely what we do, and we hope to be used by researchers.
- Keep improving visualization in personal page.
- Compare video from the same Author.
- The first collaborative test is happening! wetest#1, inspired by the first collective observation of PornHub algorithm
2019
- Released working browser extension, good enough for researchers and running workshops. A good high level description is in our project analysis.
- Released a page with a summary of our first experiments.
- Give control to researchers and individuals: a supported can tag their contributions, and remove any evidence they don't want to preserve. Download data in CSV format or via API.
- Public stats.