Trustworthy Autonomous Recommender Systems on Music Streaming Platforms

Eight new TAS research projects announced

Currently almost 60% of the global recording industry sales are made via streaming platforms. Given the enormity of choice on these platforms, and that music listening is a low-key, routine consumption choice, consumers are more and more relying on the recommendations of autonomous recommendation systems. Streaming platforms are two-sided markets, where recommendations are deployed to enhance the user experience on the consumer side, but they also decide the fate of the investments that composers, lyricists, producers, and performers made into the music. We are going to contribute to a research on how such systems may lead to potentially tilted competition field between the content providers, and more specifically, between major labels and independents.

Reprex maintains the Digital Music Observatory and the Listen Local system for granular microdata about music use in small territories (i.e., on small country or sub-national level.) We will provide data/expertise in music streaming and recommendation systems and links to many relevant stakeholders with our considerable experience running experiments on music platforms.

A research team of the University of East Anglia (UEA) the University of Liverpool (UoL), The University of London (City), and King’s College (KCL), supported by the Competition Market Authority of the United Kingdom and Reprex won a prestigious research grant to understand how recommender systems on music streaming platforms can employ trustworthy AI.

The researchers will explore the relationship between the autonomous recommendation systems and entry barriers via simulation. Working closely with Reprex, they will simulate sets of users, and iteratively generate recommendation lists, which the simulated users will react to by deciding how long to engage for and which recommendations to listen to. Through their engagement their user profiles will be updated based on what they listen to which will feed into future recommendations.

See our Feasibility Study for [Listen Local](
See our Feasibility Study for Listen Local.

The empirical experiments of the project want to explore how autonomous recommendation systems are driving consumer choice in a real-life setting, and to establish causality between the recommendation systems and the barrier to entry. As part of the second work package, the researchers will conduct randomised trials by inviting participants to stream music through our own user interface. Reprex has extensive experience conducting similar experiments in the music domain (for various online, field experiments, and high-quality surveys.)

Link: Eight new TAS research projects announced

Daniel Antal
Daniel Antal