4 ms·
It makes me wonder how one would research Youtube algorithms from the outside. Perhaps freshly imaged computers in random geographical locations periodically r
by tmpz22 7y ago
It makes me wonder how one would research Youtube algorithms from the outside.
Perhaps freshly imaged computers in random geographical locations periodically running selenium or other browser automation software, recording which recommendations are made for various viewing preferences?
The problem is A) the study could be invalidated by basic bot prevention (from identifying Selenium, identifying fresh accounts, etc) and B) Youtube actively preventing such research with "new" accounts and C) lack of sufficient scale could skew the results.
Ideally Youtube would partner with a transparent auditor and have internal teams work with the auditors, but there's no way Google agrees to that because it could potentially unearth some very bad practices by the Youtube team (if they do exist...).
- minimaxir 7y agoI helped contribute to YouTube algorithm research for BuzzFeed News in January: https://www.buzzfeednews.com/article/carolineodonovan/down-youtubes-recommendation-rabbithole https://www.buzzfeednews.com/article/carolineodonovan/down-y...
- tmpz22 7y ago> To better understand how Up Next discovery works, BuzzFeed News ran a series of searches on YouTube for news and politics terms popular during the first week of January 2019 (per Google Trends). We played the first result and then clicked the top video recommended by the platform’s Up Next algorithm. We made each query in a fresh search session with no personal account or watch history data informing the algorithm, except for geographical location and time of day, effectively demonstrating how YouTube’s recommendation operates in the absence of personalization. From a technical perspective this does not seem very rigorous, specifically: > We made each query in a fresh search session with no personal account or watch history I don't see how you can confidently say > effectively demonstrating how YouTube’s recommendation operates in the absence of personalization --- 1) What is a "fresh search session" 2) Were all these requests made from the same IP address? 3) What does no personal account or watch history mean? Did you use incognito mode, did you logout from chrome, had you recently been logged into an account in that browser (such that cookies may still be present)? --- The point of my comment is that this research is ~hard~ and its difficult to definitely prove that Google's algorithms are immoral unless you can research with a high degree of rigor (the study is repeatable, its been peer reviewed by experts, etc). FWIW I think this research is immensely valuable and needed, but I feel like until it is done well the "news" about it is little more then Op eds and blogspam.
- iudqnolq 7y agoWhy did you edit out the line that says you used techniques like the ones the gp mentions? That was what made your comment relevant.