What if the Musical Genre is No Longer Enough? Digital Methods for Investigating the Illusion of Categorisation

Working Papers from MOCReN Lisbon 2025: Music and Online Cultures in a Changing Platform Ecosystem NOVA University Lisbon, 19–21 June 2025 1 WHAT IF THE MUSICAL GENRE IS NO LONGER ENOUGH? DIGITAL METHODS FOR INVESTIGATING THE ILLUSION OF CATEGORISATION Mattia Merlini University of Milan Mattia Zanotti University of Milan Abstract: To reflect on the contemporary relevance of musical genres, it is necessary to integrate a historical tradition of theoretical thinking with research on platform power and internet cultures which can be validated by novel empirical investigation methods. For the latter, digital tools used to analyse and aggregate metadata, become key resources for research. In our paper, exploiting the potential of digital tools, we first present some perspectives on Spotify’s genre classification system; then, by studying genre on YouTube and TikTok, we show how these are perceived and reworked by platform users. In this way, we demonstrate how in some cases the non-transparency of data provided by platforms should be compared with other types of analysis, directly carried on internet cultures. Moreover, it should be kept in mind that all metadata we can access via our digital tools is in the hands of corporations that have the power to change their accessibility and their function within the system at their will, thus affecting not only the kind of agency they perform on current communities of listeners and musicians and their conception of genres, but also our epistemological grasp on the matter. We use the case study of the recent changes in Spotify’s staff to show how the use of music genre in the Spotify Wrapped feature has changed in the last three editions, thus showcasing the power that corporate decisions have on the manipulation of the imaginaries regarding genres, being them existent or made-up. Keywords: digital methods, musical genre, platform capitalism, Spotify, streaming platforms Introduction Musical genres are one of the favourite topics of popular music studies, with several relevant monographs and papers (e.g., Brackett 2016; Fabbri 1981; Fabbri 2012; Holt 2003; Holt 2007; Krogh 2025; Lena 2012; Moore 2001; Negus 1999) entirely devoted to the theme. In popular music, as opposed to art music, genres are understood as 2 discursive concepts – to use Franco Fabbri’s definition: ‘a set of socially accepted rules regulating musical events’ (Fabbri 1981: 52) – which can change over time and are usually connected with specific subcultures, practices and business models. Studying the dynamism of genres can be challenging, especially in an era in which the classical functions of genres – as described by Simon Frith: to organise music making, music listening and music selling (Frith 1996) – are perhaps more heavily than ever influenced by opaque corporate strategies, that can now embed specific conceptions of genres in the UIs of streaming apps and in the ways their algorithms work. The shape of the ongoing activity of genre definition traditionally pursued by Fabian Holt’s “center collectivities” (Holt 2007) is then to some extent redefined in the streaming age, and for this reason we felt the need to get back to these (and other) theoretical concepts to recontextualise them in the contemporary situation, using digital methods to analyse specific genre-related practices that take place within online networked cultures. Now, let us present some results of our ongoing research. Genre as Socio-technical Musical Imaginary and Digital Methods Theory One potential avenue for reframing the concept of genre is to draw upon the notion of the imaginary – understood as a collectively imagined social order that shapes technological and cultural projects (Jasanoff & Kim, 2009). In digital music culture, this notion helps us explore how genre is not only imposed top-down by industries or platforms but also co-constructed by users. Drawing from Bucher’s (2017) idea of the algorithmic imaginary, we see how user perceptions of algorithms, and the feelings they produce, influence how genres are experienced and reshaped online. Platforms like Spotify do not just categorise music; they structure listening habits, shape expectations, and foster new communal imaginaries around musical objects. This leads us to consider digital platforms as sociotechnical environments. They centralise production, circulation, and consumption, while also enabling highly granular, user-generated categorisation. Genres here emerge not just from formal labels but from the clustering of shared listening habits, social networks, and platform-mediated interactions. The term “socio-technical”, according to Caliandro (Caliandro et al., 2024: 17), is useful to emphasise how the imaginary – referring to the consumer in this case - is not only the result of the individual's consumption habits, but rather the result of the interaction and co-creation between the consumer, society, and the technological infrastructures (platforms) to which they have access. So, by combining Holt’s concept of genre as a network of shared codes (2003) with Glenn McDonald’s Spotify-based idea of genres as communit[ies] of artists, practice and/or listeners (McDonald 2024), we can describe how digital genres today are both community-driven and platform-mediated. Ultimately, the socio-technical musical imaginary becomes a powerful heuristic for understanding how genres are continuously redefined through the interaction between users, technological infrastructures, and cultural norms. It shifts the focus from fixed classifications to dynamic, collective meaning-making processes shaped by the affordances and constraints of platforms. This enables new forms of connection and clustering among musical and cultural objects, driven not only by patterns of consumption but also by the shared elements embedded in the platforms’ operational logic. 3 Trying to Explore SMI: Digital Methods A central question in our research is whether socio-technical musical imaginaries can be studied. Drawing on Jasanoff and Kim (2015), we use interpretive methods that focus on meaning-making and the interplay between structure and agency. In this framework, digital platforms are not just tools for research, but also research environments. This dual role allows us to take seriously everyday digital practices – what Beers call the “quirks of digital culture” – as entry points into the deeper dynamics of platform-based music consumption. Using a digital methods approach, we analyse the data platforms make available – genre tags, paratexts, usage patterns, recommendation systems – through a lens inspired by assemblage theory (Langlois; Sharma), integrating technical, social, and cultural dimensions. This enables a crossplatform perspective that focuses on Spotify, YouTube, and TikTok. To investigate these imaginaries, we combine native and external tools. TikTok’s Creative Center, though limited by selective data sharing, offers insights into trends and engagement. We complement it with Zeeschuimer, a browser extension that collects data through scrolling and scraping, and 4CAT, an open-source tool for data analysis. Together, they help reveal how users engage with the imaginaries embedded in TikTok’s algorithmic culture. Spotify provides access to public listening data through various third-party tools. The SpotiGeM Hub (University of Milan) allows exploration of playlist metadata, while tools like Every Noise at Once and Organize Your Music help visualise genre structures and user preferences. Although now offline, the Spotify Artist Network once offered valuable insights into listener-based artist connections. On YouTube, the YouTube Data Tools (YTDT) provide access to metrics like view counts and subscriber numbers and enable the creation of Video Co-commenting Networks – visualisations that reveal connections between videos based on shared user activity. These patterns of co-engagement help identify interpretive communities and genre-related clusters. More broadly, such tools allow us to trace how socio-technical musical imaginaries are constructed, circulated, and reshaped by both platforms and users, showing how digital infrastructures actively influence musical meaning and engagement. Platform Capitalism and Opaque Data We know it: Platforms like Spotify, YouTube, and TikTok function at the crossroads of economic, cultural, and technological forces, redefining how music is accessed, circulated, and interpreted – often beyond the control of the traditional music industry (Srnicek, 2017). This shift has not erased the role of the industry but has altered the environment in which power operates. Music consumption data is now central to understanding cultural trends, yet this data is mediated – and owned – by corporate platforms. While platforms present themselves as empowering tools, they simultaneously extract value from users’ behaviours in processes known as datafication. Crucially, all metadata and analytics remain under corporate control. Platforms can modify how data is accessed or categorised, influencing not just musical interactions, but also how researchers can study them. 4 In this context, genres do not disappear – they evolve. On platforms, they become flexible cultural codes shaped by tags, memes, emotional engagement, and algorithmic logics. No longer just industry-imposed categories, genres are now co-produced by platforms and users through a mix of bottom-up and top-down dynamics – namely, socio-technical musical imaginary. This requires moving beyond platform-provided data and combining quantitative analysis with close observation of internet cultures and user practices. Made-up Genres on Spotify in the McDonald Era So, while it is true that thanks to digital methods we can acquire a lot of valuable information to study genres and the related socio-technical musical imaginaries, the unequal power balance between us and the corporations can lead to unexpected results. A case that demonstrates this problem takes us back to a previous case study of ours (Merlini 2025), which originated from a Spotify “Wrapped” report one of us got in 2022, featuring “voidgaze” in the “favourite genres” list. The problem is that he had no idea of what voidgaze even was. This situation introduces the phenomenon of the “made-up genres”, a sort of side effect of Spotify’s “genrefication” strategy coordinated by Glenn McDonald, the founder of the “Echo Nest” system acquired by the Swedish streaming platform in 2014 and put at the core of its algorithmic system ever since. The Echo Nest aimed at building a “musical brain” capable of associating with each track metadata acquired both from the acoustic analysis of the track (afforded by the socalled “machine listening tools”) and from discursive information gathered from “web crawling” activities on countless internet pages (Albertario 2020; Eriksson 2016; McDonald 2024). These tools helped the Spotify team in their task of giving a name to those clusters of collective listening patterns that, according to McDonald’s vision, could be revealing of latent musical categories (Johnston 2018; McDonald 2023; McDonald 2024). The huge network of labels thus shaped could then be used to guide some of the most important processes of Spotify’s algorithmic recommendation system. We employed our full range of digital methods to study “voidgaze” (Merlini 2025), but in the end, the most interesting discovery arrived when we were contacted by McDonald himself, who had read our preliminary work and opened our eyes to a truth that can be very unpleasant from a researcher’s perspective. That is, that probably our whole work was flawed because they had been fine-tuning the domain of some genre-labels (including voidgaze) in the period between the publication of our “Wrapped” our study; and that, moreover, the Spotify API, providing those same digital tools with metadata, did not grant access to the totality of it, since some “provisional genre labels” and additional sonic parameters mapped by the “musical brain” remained hidden, thus making some phenomena very difficult to explain from the outside (McDonald 2023). Spotify Wrapped (2022-2024) and Changing Corporate Strategies So, what we learned from this is that the game can become very unfair when you are playing with a corporation that can change the rules according to its strategic priorities. But opaque and unstable metadata are only part of the problem. This became even clearer after a handful of months from our conversation with McDonald, when we learned that his collaboration with Spotify had come to an end. When the time of the 5 new edition of “Wrapped” arrived, we immediately noticed that McDonald’s absence was very tangible. Indeed, instead of picturesque and fascinating unknown labels, we found very broad and straightforward genres – like soul, rock and electronica – included in the list of our favourite ones. Yet this was but the beginning of a season of important changes: an announcement made on “Spotify for Developers” on November 27th 2024 explained that the platform had ‘decided to roll out a number of measures with the aim of creating a more secure platform’, resulting in removing access via Web API to a lot of useful information, like audio features and audio analysis. This change, which applies to new apps or existing apps in development mode, will prevent new digital tools from integrating these features, which are a substantial part of what is needed to study genres on Spotify. One week later, we were surprised again as we were finally able to explore the 2024 edition of our “Wrapped”, and found out that made-up genres were back… but in a very different guise. “Witchy permanent wave art pop”? “Robbery songwriting”? “Medieval pirate tavern”? “Pink pilates princess strut pop”? “Tarantella theatre laboratory”? “Farmer market wrestling italo dance”? As we went on reading through hundreds of these labels, they started to acquire a creepy taste, and we could not help asking ourselves: do they not sound like a sort of tentative AI surrogate of Glenn McDonald’s work? The launch announcement of Wrapped 2024 on Spotify’s news website seems to suggest that this is exactly what they are: ‘Through the combination of Spotify’s powerful personalization technology and generative AI, we’ve created hyperpersonalized Wrapped experiences that connect millions of listeners worldwide with the music and audio they love’. Given that most of the words included in the “genres” we have collected can be found in the original Echo-Nest-based genre network ENOA, it is plausible that the AI-based system used those labels as an initial dataset for building user-specific labels, perhaps adding some relatable terms. So, what about genres, now? We can see at least three kinds of genres at work in the current version of Spotify. There are more or less recognised, old-styled, “conventional” genres, being integrated in the Spotify App’s UI to basically serve as collectors of editorial and algorithmic playlists (Bonini & Gandini 2019; Hagen 2015); then, there are AI-made-up genres generated by chopping down and reassembling the labels in the old Echo-Nest-based classification system, thus pursuing that “viral awareness” that, according to McDonald, was always the main goal of Wrapped (McDonald 2023). Users can now get back to sharing their surprising, sometimes even grotesque, “fictional” genres on social networks – thus giving life to new socio-technical musical imaginaries. Finally, those same old labels are apparently still leading the algorithm behind the scenes, but they are now merely “functional” genres that have abandoned any claim of being recognised by actual communities or having a role outside of the algorithm. As we cannot wait to see whether the triad conventional-fictional-functional will eventually become a tetrad (or just a couple) next December, again we are reminded that unfortunately, as researchers, we can do little more than playing around with our tools, building ephemeral frameworks to perhaps understand a little part of this everchanging situation, while always running behind our objects of study. In this fluid and opaque situation, we are subject to corporate power just as common users are, and it is 6 perhaps to the latter that we should watch, to ask ourselves whether the illusion of categorisation has after all become any more delusional than it used to be. References Albertario, G. (2020). Dalla pirateria allo streaming. i metadati come specchio della trasformazione della musica digitale: Il caso di Oink [M.A. Thesis]. Università degli Studi di Milano. Bonini, T., & Gandini, A. (2019). “First week is editorial, second week is algorithmic”: Platform gatekeepers and the platformization of music curation. Social Media + Society, 5(4). Brackett, D. (2016). Categorizing sound: Genre and twentieth-century popular music. University of California Press. Bucher, T. (2017). The algorithmic imaginary: Exploring the ordinary affects of Facebook algorithms. Information, Communication & Society, 20(1), 30-44. Caliandro, A., Gandini, A., Bainotti, L., & Anselmi, G. (2024). The platformisation of consumer culture: A digital methods guide. Amsterdam University Press. Eriksson, M., Fleischer, R., Johansson, A., Snickars, P., & Vonderau, P. (2019). Spotify Teardown: Inside the black box of streaming music. MIT Press. Fabbri, F. (1981). A theory of musical genres: Two applications. In D. Horn and P. Tagg, Popular Music Perspectives (pp. 52-81). IASPM. Fabbri, F. (2012). How genres are born, change, die: Conventions, communities and diachronic processes. In S. Hawkins (Ed.), Critical musicological reflections (pp. 179191). Ashgate. Frith, S. (1996). Performing rites: On the value of popular music. Harvard University Press. Hagen, A. N. (2015). The playlist experience: Personal playlists in music streaming services. Popular Music and Society, 38(5), 625-645. 7 Holt, F. (2003). Genre formation in popular music. Musik & Forskning, 28, 77-96. Holt, F. (2007). Genre in popular music. University of Chicago Press. Jasanoff, S., & Kim, S.-H. (2009). Containing the atom: Sociotechnical imaginaries and nuclear power in the United States and South Korea. Minerva, 47(2), 119-146. Jasanoff, S., & Kim, S.-H. (Eds.). (2015). Dreamscapes of modernity: Sociotechnical imaginaries and the fabrication of power. The University of Chicago Press. Krogh, M. (2025). Musical genre: Assemblage, abstraction, and digital terms. Bloomsbury Academic. Langlois, G. (2011). Meaning, semiotechnologies and participatory media. Culture Machine, 12, 1-27. Lena, J. C. (2012). Banding together: How communities create genres in popular music. Princeton University Press. McDonald, G. (2024). You have not yet heard your favourite song: How streaming changes music. Canbury Press. Merlini, M. (2020). A critical (and interdisciplinary) survey of popular music genre theories. Interdisciplinary Studies in Musicology, (20), 79-94. Merlini, M. (2025). Voidgaze and phantom genres: Functions and dysfunctions of music genres on streaming platforms. In C. Dysers, P. Edwards, & J. Lochhead (Eds.), The music of absence: An aesthetics of loss in the new millennium (in print). Edinburgh University Press. Moore, A. F. (2001). Questioni di stile, genere e idioletto nel rock. Musica/Realtà, 22(64), 85-102. Negus, K. (1999). Music genres and corporate cultures. Routledge. 8 Sharma, S. (2013). Black twitter? Racial hashtags, networks and contagion. New Formations, 78(78), 46-64. Srnicek, N. (2017). Platform capitalism. Polity Press.