✖ YouTube's recommendation algorithm, music discovery and extremism
And: the future of music and creative tech; marketing on one streaming platform only; rapid testing = no social distancing?; AEG's plans to come through the pandemic are contactless
We all know about YouTube’s recommendation algorithm. We love the music we’ve discovered through it. We’re scared of the opaque, black-box nature of the algorithm. That we’re right to be scared of the workings of YouTube rabbit holes has just been confirmed again by a report about the Christchurch shooting. Quoted in an article about this report in Wired, Becca Lewis, a PhD candidate at Stanford, states that
“[i]t’s clear that the core of [YouTube’s] business model has an impact on allowing this content to grow and thrive. They’ve tweaked their algorithm, they’ve kicked some people off the platform, but they haven’t addressed that underlying issue.”
Scary as that is, I think that the same reasons that YouTube was instrumental in radicalisition of the Christchurch shooter help people discover new music on the platform.
So how does it work?
YouTube’s recommendation algorithm mainly works by taking into account the related co-views of any given video. In other words, it’s our accumulated viewing that helps to predict what you want to watch next after the video you’re currently watching. In a kind of self-perpetuating feedback loop the recommendations than serve two purposes:
They keep you on the platform by feeding you more of what you want to see and hear
They create a kind of network of videos that cluster around the same co-viewing stats
“underlying cultural logic of similarity within each cluster that is produced by the technologically-mediated and aggregated practices of usages by listeners and uploaders.”
Again, it’s usage that drives the creation of clusters. Furthermore, the clusters found by Airoldi et al break apart into more classically determined genres such as Grime, K-Pop, Jazz, etc. and more situational or marketing-specific genres such as Chillout, Glee Music, Hair Dryer Sound. The recommendation clusters, then, not just reinforce existing cultural indicators, but also help to solidify new ones. This results in something like Relaxing Background Music becoming popular leading to more videos will with that tag appearing on the YouTube platform.
Sonic, social, and cultural clusters
As Steve Goodman [aka Kode9] has shown, in his excellent book Sonic Warfare, that music has a unique ability to affect and that understanding this affect is necessary to study anything related to sonic culture [p. xiv]. From Goodman, I take the understanding that all vibrations connect us, good and bad, as they shape what we feel before we organise those feelings into known categories. The question of who takes advantage then becomes the question of who can tap into the right vibrations, which is as much about metadata as it is about sound.
Malcolm James, in his perceptive book Sonic Intimacy, has shown how the vibrations of Grime helped steer it to success on and through YouTube. James states that “grime’s treble (and MC) sonics and the lo-fi ethics and aesthetics of YouTube videos” [p. 24-25] created a sonic intimicacy between artist and viewer, but also between viewers. The latter, because watching those videos on YouTube, mainly on your phone, helped solidify the entire sonic identity of Grime. The clustered viewing extended into real world vibrations, for example at parties, while online the visual and sonic aesthetics fit perfectly with the way the sound was perceived through a phone speaker or earphones.
Intimacy, or closeness, is one of the key characteristics of successful influencers. Music has always created a sonic intimacy and platforms provide troves of data to base decisions on when it comes to placing and marketing a video on YouTube. There are scores of websites and videos that attempt to tell you how to ‘game’ the recommendation engine and get a higher number of views. All of these tips, however, do not allow us a closer understanding of how the algorithm works. As Tania Bucher has pointed out in her book If…Then it’s not the algorithm that’s necessarily bad or good, it’s more important to consider “how and when it matters” [P. 148].
It seems then, that those who can play into the clustered nature of the recommendations to bring across their music - or their message - can take advantage. Perhaps, though, it’s not a question of who can take advantage, but of how social patterns are exacerbated through recommendations. If I listen to a Radiohead video on YouTube and through a series of recommendations end up discovering the music of Hatis Noit she has a new fan and I have a new musical love. Yet, this same sequence also means I will be given more of the ‘genre’ that YouTube positioned Radiohead and Hatis Noit in which isn’t necessarily what they wanted or where they see themselves (right, Thom Yorke?).
Conclusion: where music benefits, extremist views do too
Similar to clusters of music videos related to genre, situation, and behaviours, YouTube provides clusters of extremist videos that trap the viewer. Music enthusiasts get excited about delving ever deeper into certain genres and finding perhaps ever better music to study or work-out to. It’s that same rabbit-hole experience that allows a person mildly attracted to conspiracy theories to find themselves within a cave of white supremacist noise. There isn’t an easy way out of this, because the YouTube recommendation algorithm has to be understood within its own situation, through its own creators, and within its own ‘genre’. The black box I referred to in the introduction doesn’t exist per se, instead there’s a power dynamic at play and each time you upload a video to YouTube you play a part in that dynamic as it shifts and determines.
Want to discuss today’s newsletter with a friend or colleague?
🖖 In Chuck Fishman’s latest installment of Don’t Split the Stream he goes into seven reasons to focus your marketing attention on one streaming platform to maximise revenues.
“Artists and their marketing teams should build on top of the results of popularity algorithms by touting … public stream counts.”
🔬 Samsung published research into the music evolution that happened this year and how that might evolve over the next decade. It’s another confirmation of the coming of age of the creator talking about
The smartphone studio
🎼 One to keep an eye on is Craig Vear’s new research project on the future of music and creative technology. He’ll look into how digital scores affect creativity and musicianship. There will be many new commissions around key themes such as
➿ So many music catalogues are changing hands this year. Dan Runcie of Trapital highlights three reasons why for the BBC:
Data makes it easier to identify the most valuable songs
Royalties are consistent revenues, hardly affected by economic factors
It’s currently cheap to borrow money making it easier to put up lump sums
🇿🇦 The South African Cultural Observatory released a report that shows that almost half of all professional musicians in South Africa are close to being forced out of the industry. One of their solutions is to establish a music desk to help musicians and government work together and cut through bureaucratic red tape. Another solution is to provide government support in the switch to digital.
💉 One festival where there will be no social distancing is Unum in Albania. They will rely on rapid testing instead.
💡 AEG’s plans to come through the pandemic with renewed energy see them focusing on
building new venues
upgrading older venues (and contracts)
turning to tech, which seems to be all about going contactless
💼 Will Dzombak, manager of Wiz Khalifa talked to MBW about livestreaming beyond the pandemic:
“Moving forward, I think there will be a way at every concert to buy access digitally to it. So if someone is on tour, there will be a way for people [at home] to buy in and be like, ‘Oh, I want to be at Wiz’s show in Oklahoma.’ It will spin into that at some point to generate extra income for people.”
I’ll take this opportunity to show my hypocrisy and share a link to a fantastic performance by Hatis Noit at the London Jazz Festival via YouTube. She ‘s the first performing artist, but I can highly recommend watching Daniel Thorne and Anne Müller too.