One of the beauties of our post-pandemic world (when we all, by necessity, became very online) is that digital communities broadened in scope and deepened in engagement (and, maybe unknowingly, provided more of the most valuable resource training algorithms: data).
New waves of trends, culture, and communities brewed on the obscurity of the internet as algorithms studied and amplified our interactive behaviors online.
In some ways, this new internet culture boom entered an entirely new phase of human development.
No collection of visual artefacts or style spared the forced birth of a new “aesthetic”, encoding new patterns in algorithmically fueled social platforms and ultimately shaping new identities and brands for the youth to cling onto.
Naturally, as most visual content has an audio companion these days, I turned to music to understand how these niche aesthetics and content styles played into communities and fandom identities.
It’s easy to dismiss some of these nomeclatrues as cultural sludge, but when there are artists and fandom behind this new founded cultural aesthetic, I find it can take on a new meaning.
"Sigilkore" and "Indie Sleaze" are prime examples of how internet aesthetics emerged and gave way to artists that help articulate cultural movements in a way that few other tools can for today’s youth.
Spotify’s algorithmic/curatorial approach to the personalized music listening strategy is a direct reflection of these emergent subgenres as well.
In the last year, Sptofiy has added hundres of new playlists, personalized and editorial, to reflect these rifts on the internet.
With access to genre tags in the Spotify API, and thankfully with public sources like Glenn Mcdonald’s Every Noise At Once and volt.fm, I started off by identifying 60 key playlists like ‘undercurrents’ and ‘gloomcore’ that reflect internet trends and are directly targeted at gen z, and created after 2020.
Using playlist history information for all these playlists over the last few years, I started to question several things. How responsive is Spotify editorial to micro-trends on social/short form platforms? How much importance does Spotify put in thee "middle tier" playlists? Are there any "token" underground artists that are reshaping how Spotify's editorial teams program these playlists - how will this impact their rise? and so on...
To lay this all out, my first goal was simply mapping the “curated” emerging artist scene with Gephi, and understand the proximity and magnitude of these artists in the cultural sphere of Spotify’s ecosystem.
Data Collection
Note: I had this idea in the end of 2023 and collected the data at that time.
Within the 60 playlists I pulled current artists & any playlisted artist within the last 28 days.
Each artists was also assigned their total spotify streams (at the time) and total tiktok creations made (for the playlisted song) with the foresight of being able to scale the size of the niche, as well as labels with the foresight of identifying any labels that might be holding high market share in niche spaces.
Gephi - Preparing Data for Network Graphs
The first goal of this project is to successfully build a bi-model network, meaning it contains two different elements: artists and playlists.
Essentially, each node will represent an artist and each artist is connected to other artists through shared “emerging” playlists with an edge.
The tables were formatted to suit the nodes and edges formality in a network graph.
Credits to the Duke University Library Brown University Instructions on Gephi and Miriam Possner’s guide to Gephi. for helping me get through this (!).
Nodes.csv
the nodes file tells Gephi all the possible nodes in a network.
A node is represented by a circle within the Gephi visualization whereas the edges file tells Gephi how all the nodes are related (or connected).
The nodes file should at least have the columns "Id" and "Label."
The node table can also include attributes.
Attributes offer a way for you to distinguish between your nodes by categorizing your data by, for example, color, size, or age.
In this study, each artist will represent a node and the attributes will quantify the total tiktoks and streams for the playlisted track.
Label | ID | Streams | TikToks |
.com, Flowerbabe | 1 | 45524 | 0 |
@ | 2 | 167262 | 0 |
✝✝✝ (Crosses) | 3 | 242760 | 0 |
100 gecs | 4 | 10603305 | 1300 |
12 RODS | 5 | 129667 | 0 |
1o, 95TOYO | 6 | 0 | 0 |
26fix | 7 | 914296 | 0 |
2hollis | 8 | 1024048 | 0 |
2WEI, Bri Bryant | 9 | 865097 | 140 |
347aidan | 10 | 597786 | 11 |
4cf | 11 | 160611 | 8 |
Edges .csv
tells Gephi how the nodes are connected.
It has the columns Source, Target, and Type.
Source refers to a node that you've identified and labeled in your nodes.csv file, which is identified with the
Target also refers to a node you've listed in your nodes.csv file.
Type refers to how the two nodes are connected.
If the source drives the relationship (for example, a sender of a letter versus a receiver), the relationship is "Directed."
In this example, the sender of the letter is the source and the receiver of the letter is the target. If the relationship goes both ways -- for example, the graph visualizes friendships, the graph will be undirected.
Here is an example of what our edges table looked like:
Custom | Source | Target | Type | Weight_By_Size |
15 | 314 | 314 | Undirectional | 1 |
15 | 314 | 95 | Undirectional | 1 |
15 | 314 | 718 | Undirectional | 1 |
15 | 314 | 900 | Undirectional | 1 |
15 | 314 | 259 | Undirectional | 1 |
15 | 314 | 274 | Undirectional | 1 |
15 | 314 | 929 | Undirectional | 1 |
15 | 314 | 273 | Undirectional | 1 |
15 | 314 | 816 | Undirectional | 1 |
15 | 314 | 146 | Undirectional | 1 |
15 | 314 | 378 | Undirectional | 1 |
15 | 314 | 295 | Undirectional | 1 |
Source numbers are allocated to each artist.
Custom is the code number for each playlist.
Context
Nodes: 1676
Edges: 97050
Directed Graph
Read PT II for the full analysis.....
Comments