Recommender Systems, a Popular Media Introduction
A Collection of News Articles by Popular Media that Explain Recommender Systems in a Way Everyone Can Understand.
In 2004, Chris Anderson wrote “The Long Tail”, an article that motivates the need for recommender systems. In The Long Tail, Anderson depicts the shift from a world of content scarcity limited by the laws of physics and retailers, to a world of content abundance and digital content, where recommender systems have the potential to fulfill our niche tastes. 16 years later, the world Anderson depicts is a reality, yet recommender systems keep advancing to fulfill the third rule he proposes in the article: “Help me find it”.
There has been a few industries that have widely exploited recommender systems for the benefits of the customers. Pushed by the boom of streaming, and its every-day use, music, news and video have been at the forefront of recommender systems in popular media— for the good and the bad.
In a 2013 Wired.com interview, “The Science Behind the Netflix Algorithms That Decide What You’ll Watch Next”, Xavier Amatriain and Carlos Gomez-Uribe explain how Netflix recommended you what to watch — either via streaming or DVD rental back in 2013. Netflix has talked to popular media about their recommender systems in many other occasions. In 2017 and 2018 interviews with Todd Yellin (“This is how Netflix’s top-secret recommendation system works”), Netflix explained once again how customers get to the recommendations.
Music recommendations have been covered from many more news outlets,
from business, music specialized outlets, and technology. Moreover, music streaming platforms are also interested in explaining how recommendations work both to their users and the content creator in order to attract both of them. For example, “Pandora’s Content Recommender: Using Data to Match Brands with Content Strategies”, or Spotify’s “Five Ways to Make Your Discover Weekly Playlists Even More Personalized”.
In a pre-Cambridge Analytica era, The Echo Nest (now part of Spotify) shared some very interesting insights with Wired UK in their article “Echo Nest knows your music, your voting choice”. They explained how the Echo Nest systems understand music to make recommendations, and how that enables them to understand audience beyond music.
Of course, not all the coverage of mainstream media for recommender systems is positive. As the technology evolves, there are several areas of concern for the public; one is the quality of recommendations — sometimes, recommender systems get it really really wrong. Another main area of concern is privacy. As recommender systems tend to be data rich, understanding how they use our personal data is vital. These topics are discussed in “Personalization Has Failed Us”, a NY Times article by Thorin Klosowski.
In “Up Next: A Better Recommendation System” (Wired Magazine), Renee DiResta discusses recommendation systems from an ethics point of view, particularly when it comes to the spread of misinformation, or creating dangerous content rabbit-holes like the ones the NY Times described for youtube in their article “On YouTube’s Digital Playground, an Open Gate for Pedophiles”.
Other interesting generic articles explaining recommender systems include the 2015 “The Evolving Landscape of Recommendation Systems” by Mehrdad Fatourechi in Tech Crunch, or “Building the Next New York Times Recommendation Engine” by Alexander Spangher for the NY Times.
Finally, one can see recommendations as a much wider problem space that is not only about selecting what the user might click, listen or watch next. As a matter of fact, recommendation has been widely successful when interesting and familiar patterns of user experience have been used. For example, in Spotify’s Discovery Weekly (“Why Spotify’s Discover Weekly Playlists Are Such A Hit”) or Netflix Artwork Personalization (“Artwork Personalization at Netflix”)
