Call for transparency on music streaming algorithm ‘bias’

A report on the impact of recommendation algorithms on the UK music industry reveals that most music creators and industry professionals believe they result in unfair outcomes.

However, while such views are held by the majority, the report believes there is “limited evidence” either proving or disproving whether these algorithms amplify existing bias or introduce new biases.

Any consideration of the impact of recommendation algorithms used by digital service providers, adds the report, needs to acknowledge the “wider history and condition of the UK’s music industry”.

READ MORE:

Rather than view them in isolation, algorithms should “be considered one among several different cultural intermediaries that influence how consumers engage with music”.

The report, commissioned by the Department for Culture, Media and Sport (DCMS) also states that the majority of listens (70%) on digital services “remain unguided by recommendation algorithms”.

The research, the latest in an ongoing programme of work addressing issues identified by the DCMS Select Committee’s Inquiry into the Economics of Music Streaming, found that a lack of transparency on the part of digital platforms is a major contributor to concerns about how the algorithms work.

Current approaches to transparency “are not sufficiently alleviating the concerns of consumers, creators, and other stakeholders across the music industry”.

The result is that “feelings of mistreatment and suspicion, whilst not necessarily justified, are widespread”.

Popularity and demographics

The biases examined in the report fall broadly into two categories: popularity bias – i.e. the more well-known an artist is, the more their music is recommended – and biases according to demographic characteristics, such as gender, race and age.

Perceived biases highlighted in the report show that music creators remain highly suspicious of how algorithms work.

Of those creators who responded to a survey, 89.2% agreed or strongly agreed that they are concerned that “bias in recommendation algorithms could lead to certain artists, or labels, being prioritised over others”.

85.3% of respondents expressed concern “that algorithmic bias could lead to music from certain genres being prioritised over others”, while 67.7% were concerned that “certain demographic groups (e.g., different ethnic groups, genders etc.)” were being prioritised over others.

Concerns were also expressed about the relationship between major labels and digital service providers. It was also felt that introducing the capability for creators to forgo earnings in return for better algorithmic promotion “will create an uneven playing field”.

Transparency and research

In acknowledging that its findings are inconclusive due to the need for “greater access to data and more complete datasets”, the report calls for further action on transparency by digital platforms.

This should include a standardised approach to transparency across digital providers that could draw on the Algorithmic Transparency Recording Standard developed by the Centre for Data Ethics and Innovation and Central Digital and Data Office for use in the public sector.

Other action it recommends includes clearer signposting of when a playlist is curated by algorithms, editors, or a combination of both, and better communication between digital providers and music creators/their representatives, around access to data about how their music is consumed.

It also calls for more content tailored to a non-academic audience, “about how their recommendation algorithms work”.

The report concludes that there are number of areas where further research is needed in order to better understand the use and impact of recommendation algorithms.

In particular, it believes that an absence of evidence calls for closer working between academics and digital service providers.

Perpetuating bias

Naomi Pohl, General Secretary of the Musicians' Union, said that one of the stand-out statements from the report is that '"regardless of user gender, music streaming platforms appear to predominantly recommend white male artists to users at a significant rate".

"This is deeply concerning and reflects a bias not only from music streaming platforms but in the wider industry. Where bias exists, artificial intelligence and algorithms will absorb and perpetuate it.

"Human playlisters may have unconscious bias but they can at least make a conscious effort to offer diversity. This is just one issue with algorithmic recommendation and we look forward to further research."

Silvia Montello, CEO of the Association of Independent Music, told ArtsProfessional that the need for greater transparency around “how and where algorithms are used” is vital.

She said: “If indeed 70% of music consumed [on digital platforms] is user-driven and not algorithmically served – a statistic which is questionable and needs further expansion and clarification – then communicating this transparently to the music listening audience would also help provide further reassurance.”

Transparency is particulary important for those in the sector who have less leverage in terms of financial and marketing clout, she explained.

“The independent music community in particular will need more consistent policy and real-world evidence that popularity bias doesn’t continue to disproportionately surface major-label artists.

“Popularity bias works for the mainstream but against the more niche genres which are key for the rich, diverse and world-leading musical output from the UK industry.”

Montello added that algorithms that create playlists based on serving up music of the kind already listened to by a consumer “can be very helpful in promoting the niche genres a user engages with”, which in turn can be helpful for the independent sector.

She continued: “This does of course work against the discovery of music outside of a user’s own ‘wheelhouse’, which they may wish to explore.

“This is why radio in all its forms, including BBC Introducing and specialist shows on national broadcasters, is still vitally important for true music discovery.”