How Virtual Recommendations Shape Your Music Preferences
Thursday, November 12, 2015
Thanks to a growing number of streaming services like Apple Music, it’s now easier than ever for listeners to discover their new favorite song or artist among millions of choices.
Online platforms that suggest new music, movies, and products based on consumers’ established preferences are powered by recommender systems—dynamic algorithms that leverage users’ virtual behavior to suggest products or content that they have not yet purchased, experienced, or considered.
Recommender systems protect users from becoming overwhelmed by the abundance of choices on the web, and can save them time in searching for what they need. But despite their convenience, research shows that recommender systems can manipulate consumers’ preferences.
According to research by Information and Decision Sciences Professor Gedas Adomavicius and his colleagues, consumers tend to report over-inflated preferences for songs, TV shows, and jokes that are displayed with a high star-rating compared to content they consume without recommender systems. Conversely, consumers report reduced preferences for entertainment presented with low star-ratings.
Consequently, Apple Music listeners may be entering into a self-perpetuating cycle in which recommender systems sneakily shape their preferences, which inform future recommendations, which further bias their music choices.
Ultimately, users are building playlists with music they may not have liked had they heard it offline.
A related study led by Adomavicius suggests consumers are also willing to pay more for products with inflated ratings that recommender systems serve them. In a set of experiments, participants were asked to name their own price for digital songs that were presented with a series of manipulated system recommendations. Results showed that when purchasing songs, each one-star increase in recommendation for a given song (on the scale of one to five stars) resulted in a 10-15 percent increase in the participant’s willingness to pay for that song.
These increasingly sophisticated recommender systems can give companies a strategic, if not unscrupulous, advantage in profiting from content that consumers might not have otherwise preferred or purchased.
“Just as savvy consumers understand the impacts of advertising, discounting, and pricing strategies, they will also need to consider the potential impact of recommendations on their purchasing and consumption decisions,” says Adomavicius.
This story appeared in the latest issue of the Carlson School Magazine. For more on students, faculty, and alumni, check out the magazine.
The study was co-authored by Carlson School Information and Decision Sciences Professor Shawn Curley, University of Arizona Associate Professor Jesse Bockstedt, and Indiana University Assistant Professor Jingjing Zhang.