Youtube Suggestions and Clickbait
During class, we discussed diffusion throughout networks, cascades, and whether or not a node (a person) would adopt the strategy of other nodes. YouTube’s suggestion algorithm is designed to suggest videos to the user that are most likely to make that user watch is much as possible. The algorithm takes into account many factors, such as watch time, views, and likes, to determine what videos to suggest. A user is thus given the change to decide whether or not to adopt the strategy of others who have viewed this video given the information that it was suggested, meaning it has a good amount of views, likes, watch time, and would fit you well.
In class, we discussed how a Nash Equilibrium can ensure from choices like this, i.e. whether or not to adopt certain strategies given the payoffs for those strategies and the choices of others. In the case of YouTube videos, the payoff is enjoyment from watching the video.
As the goal of the suggested videos is to maximize the user watch time, watch time of the videos plays the biggest factor. But, unfortunately, as the article points out, the suggestions are often really bad, really clickbait and controversial videos are suggested. For example, in the past, at the top of many people’s suggestions were videos about how the Sandy Hook shooting was fake and how there were no planes in the sky during 9/11.
YouTube’s machine learning algorithm has been suggesting these videos as it “thinks” these are the best videos to get people to continue watching. This is troubling as the more of these videos it suggests, the more likely it is for these videos to cascade throughout a social network. According to the article, “Researchers at Harvard conducted their own test and found that the algorithm was more often drawing viewers to extreme content and unfounded right-wing conspiracy theories”. Unfortunately, there “is no quick fix for this”. In summary, the suggestion algorithm and the likelihood of the suggested videos beginning a cascade relates to network diffusion which we have discussed in class.