Phylogenetic reconstruction of the cultural evolution of electronic music via dynamic community detection (1975{1999) Mason Youngblooda,b,1, Karim Baraghithc, and Patrick E. Savaged a Department of Psychology, The Graduate Center, City University of New York, New York, NY, USA bDepartment of Biology, Queens College, City University of New York, Flushing, NY, USA cDepartment of Philosophy, DCLPS, Heinrich-Heine University, D¨usseldorf,NRW, Germany dFaculty of Environment and Information Studies, Keio University SFC, Fujisawa, Japan [email protected] Abstract Cultural phylogenies, or \trees" of culture, are typically built using methods from biology that use similarities and differences in artifacts to infer the historical relationships between the populations that produced them. While these methods have yielded important insights, particularly in linguistics, researchers continue to debate the extent to which cultural phylogenies are tree-like or reticulated due to high levels of horizontal transmission. In this study, we propose a novel method for phylogenetic reconstruction using dynamic community detection that explicitly accounts for transmission between lineages. We used data from 1,498,483 collaborative relationships between electronic music artists to construct a cultural phylogeny based on observed population structure. The results suggest that, although the phylogeny is fun- damentally tree-like, horizontal transmission is common and populations never become fully isolated from one another. In addition, we found evidence that electronic music diversity has increased between 1975 and 1999. The method used in this study is available as a new R package called DynCommPhylo. Future studies should apply this method to other cultural systems such as academic publishing and film, as well as biological systems where high resolution reproductive data is available, to assess how levels of reticulation in evolution vary across domains. Keywords { cultural evolution, electronic music, phylogenetics, community detection, horizontal transmission Introduction for the spread of antibiotic resistance in bacteria [11], but it also occurs in vertebrates, invertebrates, and plants [12, Historically, researchers have relied on phylogenetic com- 13]. Nevertheless, Gould had a fair point. Although cul- parative methods from biology to create cultural phyloge- tural traits evolve through a process of variation, selection nies, or \trees" of culture. These methods use differences and reproduction that results in observable fissions of cul- and similarities in the cultural products of different popula- tural lineages, branches of the \tree of culture" can reunify tions (analogous to differences and similarities in DNA) to later on and frequently do so, for instance in the case of reconstruct the historical relationships between them. Tra- shared practices or customs. ditional phylogenetic methods, which assume the tree-like That being said, there is significant variation in reticula- structure typical of genetic evolution, have yielded criti- tion across cultural domains [6, 8], and depending on the cal insights, particularly in linguistics [1{3]. However, re- conditions (e.g. co-inheritance of traits) horizontal trans- searchers have debated whether cultural phylogenies are mission may or may not interfere with traditional phylo- fundamentally tree-like, or whether high levels of horizontal genetic reconstruction [14{16]. In language evolution, for arXiv:2011.02460v2 [q-bio.PE] 5 Nov 2020 transmission lead to a more reticulated structure [4{8]. For example, horizontal transmission is lower and phylogenetic example, biologist Stephen Jay Gould [9] wrote that: relationships can be reliably reconstructed [17]. In other domains, such as material culture, rates of horizontal trans- \Biological evolution is a bad analogue for cul- mission can be higher and more variable [18{20], leading to tural change [...] Biological evolution is a system phylogenies that clearly contradict the historical record [21]. of constant divergence without subsequent joining For contemporary culture in the digital age [22], where rapid of branches. Lineages, once distinct, are sepa- within-generational changes are the norm [23, 24], the nega- rate forever. In human history, transmission across tive effects of horizontal transmission on phylogenetic signal lineages is, perhaps, the major source of cultural are likely to be even more extreme. More recent advance- change." ments in network-based phylogenetics allow researchers to Certainly, this is an oversimplification. Horizontal trans- estimate reticulation [6, 7, 25{28], but these methods are mission also frequently occurs in biology and sometimes pro- typically unrooted [29{31] (i.e. cannot be used to infer vides problems for the practice of classification [10]. Hori- chronology [32]) and thus remain complementary to tradi- zontal transmission is for instance the primary mechanism tional phylogenetic reconstruction [33]. 1 Additionally, outside of linguistics it can be extremely Instead, what counts is the relative number of social inter- challenging to characterize complex cultural traits in a man- actions within (and outside) the community. ner suitable for phylogenetic analysis [26, 34]. In practice, More specifically, we are using genres of electronic mu- this means that cultural phylogenies are often limited to sic as a test case. Although musicologists recognize that very specific domains with variation that can be more eas- genres are generated by evolving communities of artists [47, ily characterized. For example, applications of phylogenetic 48], previous attempts to quantitatively map genres have methods in music have been restricted to traditional rhyth- depended on listener habits [49], instrument similarity [50], mic patterns [35, 36], individual instruments [21], the works or sub-genre tags on streaming platforms [51]. By using of a single composer [37, 38], or folk music within a single artist co-release data (who collaborates with who) we can region [39]. explicitly track how populations of artists, and the genres Given these limitations, it would be incredibly valuable that they correspond to, evolve over time. This approach to be able to construct large-scale phylogenies for complex is similar to qualitative attempts at reconstructing music cultural traits while explicitly accounting for and measur- trees that rely on historical accounts of how artist commu- ing horizontal transmission [14]. If phylogenies represent nities grow, diverge, and influence each other over time [52, changes in population structure over time [40{42], then one 53]. We chose to study electronic music because it is known way forward might be to assess population structure from for its rapid differentiation into competing genres and sub- the bottom-up. This is, of course, a complicated proposi- genres [54], particularly during the 1990s [55]. In addition, tion. Population structure in biology is typically determined collaboration links between electronic music producers are by genetic variation driven by the combined effects of evolu- already known to be important for cultural transmission [23, tionary processes such as recombination, mutation, genetic 24] and community structure [56]. drift, demographic history, and natural selection. Unfortu- In brief, our method uses the TILES algorithm [57] to nately, population concepts are hardly discussed in the cul- identify communities in a dynamic network of artists, which tural evolutionary literature (for an exception see Scapoli et are then clustered into populations using the fast greedy al. [43]). modularity optimization algorithm (Figure 1). TILES is According to philosopher of biology Roberta Millstein's an online algorithm in that it works with an \interaction definition of a (biological) population, the \boundaries of stream" of nodes, in this case artists, and links, in this a population are those groupings where the rates of inter- case collaborations. In other words, individuals enter and actions are much higher within than without" [44]. She exit communities as they form and break relationships with calls this definition the \causal interactionist population other individuals. A node is considered to be a \core" com- concept", or CIPC for short. If agents are represented as munity member if it forms a triangle with other community nodes in a network graph, where links represent interac- members, and a \peripheral" member if it is one link away tions relevant for reproduction (e.g. mating, transferring from a core node. Community composition is recomputed information), then populations are groups of agents that in- throughout this process, each time that a new link enters or teract significantly more with one another than with other exits the network. Snapshots of the community composition agents. In general, the rates of interactions between agents are then collected at a regular time interval. We chose to are lower between populations than within them, and this use the TILES algorithm because it closely resembles the feature is precisely what gives them their (sometimes fuzzy) formalization of the CIPC proposed by Baraghith [45]. boundaries. It has recently been suggested to use this \inner interactive connectivity", i.e. cohesion in cultural popula- tion structure, as the population defining criterion in cul- Methods tural evolution [45]. According to this interpretation, populations represent 0.1 Data Collection a specific kind of \nearly decomposable system" [46] in All data