Shall we start with a gripe? It’s time to stop talking so distrustfully about ‘the algorithms of the streaming services’. Algorithms are just the instructions used in programming, and indeed in life. Cooking a meal, tying your shoelaces and finding your way to the office are all just algorithms – albeit running on the squidgy pink computer inside your skull rather than an electronic device.
Getting back to streaming, these are digital services driven by all kinds of algorithms. There’s a value to tightening up our language and specifying ‘recommendation algorithms’ (because that’s what we’re usually worrying about with streaming). Otherwise we risk just ending up like a politician demanding tech giants ‘get rid of algorithms’ and that’s a bad scene.
Gripe over, why are we banging on about algorithms this morning?Because the UK’s Centre for Data Ethics and Innovation has published its report on what impact the streaming services’ recommendation algorithms might be having on music consumption.
And? We’ll applaud the early reminder that “These systems employ a variety of technologies and techniques depending on the specific ways in which they are deployed. The idea of ‘a’ or ‘the’ recommendation algorithm is oversimplified” for starters: it’s a useful base for talking about the ensuing issues.
“There are widely held beliefs that the use of these technologies might serve to unfairly advantage certain groups at the expense of others,” adds the report’s intro. “This report finds that evidence proving or disproving whether these technologies embed, amplify, or introduce unfair biases is mixed, and at times inconclusive.”
So, this isn’t a giant smoking gun of a report that proves Spotify, Apple Music and co are destroying music. But it does highlight some of the areas of concern about how they’re changing listening habits – and why that may not always be a positive improvement.
A few quick points from its findings. One: “The belief that recommendation algorithms result in unfair outcomes is widespread.” That’s a problem for streaming services, who get rapped by the report for a level of transparency that is “not sufficiently alleviating the concerns of consumers, creators, and other stakeholders across the music industry”.
To put that another way: even if the streaming services’ recommendation algorithms are not biased in the ways people think they are, the fact that they do think this is a problem for the DSPs, It’s a trust issue.
(In some areas the report suggests that if anything, streaming might be counteracting historic biases in music discovery: for example, the fact that “since 2018, there appears to be a slight shift in UK streaming data towards less popular artists and tracks” as a streaming-driven correction to the historic ‘popularity bias’ of the charts.)
Two: “The majority of listens on DSPs remain unguided by recommendation algorithms.” That’s a key point: that these algorithms are not yet the dominant factor driving music discovery, even though we often assume that they are. The ballpark figure is that around 30% of streams are “guided” (in the report’s terminology) while 70% are “user-led”.
Three, and this is a biggie: “Regardless of user gender, music streaming platforms appear to predominantly recommend white male artists to users at a significant rate,” claims the report. It notes that a “low share of music by female artists, both on platforms and in the industry more widely, and music critics’ recommendations” are also factors here, and points to some studies showing that “recommendation systems may increase the number of women’s songs listened to” (our italics).
More research required, is the conclusion here, as it is for the question of whether music recommendation systems are biased on race / ethnicity grounds, where “there is a lack of substantial research”. This lack seems noteworthy in itself: why has nobody found the funding for such studies? Perhaps they will now.
It’s an even-handed report, and that makes it a helpful contribution to the wider debate about how streaming services are changing listening. It also strikes a welcome note of anti-division, pointing out that some of the potential problems – demographic-driven biases for example – are as much in the interests of DSPs to understand and solve (because they want to keep users happy) as for rightsholders and musicians.