History-based recommendation is not a new thing – surely everyone has encountered the phenomenon that after watching an innocent YouTube video, the feed will be full of relevant recommendations on that topic. Although the recommendation algorithm has been developed for many years, it is far from perfect; at least from a user perspective.
The thing probably works, just not as we would expect: YouTube doesn’t recommend the best videos in terms of relevancy, but the ones that promise it the most revenue. There are a few stories circulating on the Internet that the algorithm, on the other hand, is downright harmful and can influence people’s way of thinking, including political opinion.
A new study e for people. Although the study has flaws, it seems that, fortunately, this is the latter case, and the link between video sharing and radicalization (or, if you like, brainwashing, as it may be not just politics but even flatland or lizards) has not been proven.
The investigation analyzed data from Nielsen based on the content consumption habits of 300 000 people. Between 2016 and 2019, they viewed 21 million clips on YouTube. The anonymous study, as appropriate, took into account data from users who did not know they were part of the study. If they knew, they could have been more cautious and possibly tried to hide certain activities and interests.
The creators first narrowed the scope of the videos; a keyword search revealed that only 3.3 percent of the videos viewed had anything to do with politics. The videos were “collected” by 1,000 channels.
Based on the analyzes, the clips were typically unable to transform people’s political views: anyone who visited left-wing content in 2016 watched such videos in 2019 as well. , and who was clicked on right-wing videos in 2016, held similar views in 2019. Based on these, radicalization and fanaticization are not typical at all – although it cannot be ruled out that such a case has occurred – it is much more characteristic that people see or seek to reinforce their own vision in the videos they watch.
however, a shortcoming of the study is that it only examined the period 2016-2019 and in 2020 the data analysis was already underway; that is, a year was dropped when the video sharing was flooded with political content due to the U.S. election.
Views and number of YouTube users however, it has increased, suggesting that more people have used the Internet to obtain political information.
Whether YouTube can radicalize this, of course, is not yet clear; this required a trend analysis. The authors of the study are the so-called. in addition to the “mainstream” (left and right groups), a distinction was made between the anti-election and the far right groups (the far left was left out because it was present with a statistically immeasurable proportion); and viewer habits were as follows. The viewership of the right-wing, far-right and anti-voting groups started at a similar level in all respects. Overall, however, the content on the right was viewed by more people – but the average viewing time did not change significantly. Anti-voting videos were watched not only by more people, but also on average for a longer period of time. , in addition, we should experience accelerating growth. The study also points out that for far-right users, YouTube is only part of a larger system; because each video is more likely to come from an external site than from a recommendation under a previously viewed YouTube video.
Overall the study found no evidence that YouTube would steer users to the far right in any way. However, the data show that the platform’s referral system does a good job of reprimanding viewers on the site as much as possible.
So much it is also worth noting that anti-voting videos do not count as explicit political content, their viewers are otherwise typically right-wing (i.e. not radical).
In addition to the above-mentioned error of the study, there are other shortcomings: for example, the creators do not know to whom YouTube gave when and what recommendations, the database only contains the reactions. In addition, the study only analyzed videos viewed with desktop browsers, not taking into account those viewed by users on mobile. And, of course, under the law of large numbers, averages can obscure the extremes of some smaller groups. or even the opposite.
Are you more interested in IT? You can find our news and analyzes for IT and infocommunication decision makers here
