Do algorithms keep playing the same old song?

What is the impact of algorithmic recommendations on the diversity of music being streamed? Using data provided by Deezer, researchers at the Records Project formulated an initial breakdown of the question.

Music streaming sites give their users access to libraries containing tens of millions of tracks. In order to navigate these inexhaustible catalogs, algorithms generate suggestions based on each user’s listening history. However, does this risk isolating them in an all-too-familiar “bubble”, a kind of echo chamber in which they are exposed to the same tastes over and over again? The question has been extensively examined and discussed in the context of political content on social networks, but little research has been done from the point of view of culture and music.

The Records Project set out to fill this gap, bringing together researchers in the social sciences as well as computer science to analyze the vast amounts of data provided by streaming services. The Deezer website contributes to this pledge by providing the team with an anonymous listening history for its users.

“Sociologists usually work with declarative surveys, which creates recurring issues of reliability,” says Philip Colangon, a sociologist who specializes in cultural practices and tastes at OSC. “Just like in politics, people change their reactions depending on what they think should be highlighted.” In turn, researchers in the Records Project benefit from high-quality information, and are spared the daunting task of searching for spillovers from Internet users’ activity.

“This data allows us to compare what happens depending on whether or not the user follows the recommendations of the algorithms,” explains Camille Roth, a CNRS researcher at the Marc Bloch Center who specializes in the social sciences, mathematics and computer science. “I check whether ‘echo chambers’ are created, how the information is distributed depending on the form of listener interactions with the algorithms, and whether any broad user categories can be identified.”

Stuck in a filter bubble?

In contrast to social networks, using streaming sites is a solitary activity, and users seem to be more involved in choosing the content they are exposed to. Social interactions may exist, but they are much less developed. In this context, how prevalent are the notorious “filter bubbles”, the mechanisms that supposedly isolate Internet users by suggesting overly similar recommendations?

“The logset’s work reflects the usual view about the role of algorithms in the formation of filter bubbles,” stresses Roth. “Instead of checking whether the recommendation affects user behavior, we study how the user deals with it. We have learned that there are different situations, and the effect of the recommendation and the effect of filtering vary accordingly. On the issue of online filter bubbles, it is necessary to distinguish different categories of users”.

In fact, some listeners use streaming sites as a personal library that plays only music they have personally chosen. Others use it like radio, leaving more or less room for randomness and discovery. Two basic scenarios emerge: listeners either rely on recommendations that are managed solely by artificial intelligence and rely on their listening history, or on so-called “editorial” recommendations in the form of public playlists compiled by other users or by streaming staff, often With a specific topic and sometimes with the help of algorithms.

Same old song?

Broadcast service users can thus be represented schematically in a triangle whose three angles are: “organic” listeners who do not follow recommendations, listeners who rely primarily on algorithmic recommendations, and listeners who prefer editorial recommendations. The researchers studied listening diversity within these user groups, in other words, the rates of quantity and frequency of songs being played, as well as their overall popularity. On the other hand, musical genres were not taken into account in this study.

In fact, “we should not give too much credence to the designations of musical styles – they are based on arbitrary decisions, especially in the case of more famous artists,” insists Manuel Muslim, senior researcher at Deezer, where he has worked for the past eight years after earning a degree. PhD in Musical Signal Processing. “In our work, rating Maître Gims as pop or rap can have a significant impact on results.”

Accordingly, the records project decided to evaluate diversity according to two criteria: first, the relationship between the number of different tracks played and the total number of plays, and secondly, the overall popularity of the songs in question. The behavior of the three main user groups can then be compared to the playlists of about 30 highly rated French radio stations.

The results showed that users who often follow the Deezer algorithm hear a variety of music, less popular than most music FM stations, while those who prefer editorial recommendations focus more on successful artists.

What do users want?

“Our analyzes have led us to discard theories under which automated recommendations systematically restrict users’ choices, or conversely ensure the display of a diverse range of content, including lesser-known artists,” explains Thomas Lowell, CNRS researcher at Géographie. Citi laboratory. Measurements show great variability from group to group in the use of these tools and the resulting effects. But at this point, we know nothing about listeners’ expectations, nor about the contexts in which they use these tools – nor about the users themselves, for that matter. Our survey mechanism, which combines listening history analysis with a questionnaire as well as in-depth individual interviews, will enlighten us on these points.”

Because the data from Deezer is completely anonymous, it does not include any demographic or social information other than the gender and age that each user declares when creating an account. Colangon concludes, “These findings are also pertinent to another question, which is to what extent today’s major cultural divide is not related to the difference between serious art music and its popular counterpart, but rather to access to a variety of musical references.” “We can already see indications of that here, but there is more work to be done.”

Learn more about the Records Project website:


Leave a reply:

Your email address will not be published.