Machinic enculturation, copyright bots, and the aesthetics of composing mashups for machines

ABSTRACT

Sample-based media producers compose their art for machines. They implement aural effects and editing techniques to their source media, which very directly affect the sound of their music, but the aesthetics of which are not aimed at human ears; they are aimed at convincing copyright bots their media is not worth “flagging.” We offer machinic enculturation as a term descriptive of the phenomenon of adapting one’s media practices to machinic audiences. Then, using the basic operations of rhetoric (addition, omission, transposition, and transmutation) and a stylistic framework (antithesis, apostrophe, and reasoning by question and answer), we demonstrate a born-digital aesthetic, moored in the rhetorical prowess of some producers as they compose their music for copyright bots.

KEYWORDS:

• Content identification
• SoundCloud
• mashup
• remix
• machinic enculturation
• aesthetics

Arguably all automations on the Web are rules “made durable” (Latour 1990), imposing value systems on users, constraining certain modes of expression, and supporting others. Copyright bots—automations designed specifically to locate, identify, and alert to the use of copyrighted material—rigidly and relentlessly enforce public copyright law at the behest of private parties. Where automations might be studied as constraints on art, constricting the available means of expression (e.g., making some audio samples “off-limits”), in this article, we describe machinic enculturation, a phenomenon wherein machines are not simply treated as constraints on expression, but rather, as audiences with their own expectations for media content, which producers attempt to learn and meet. Entangled with mashup artists’ adaptations of craft for copyright bots is a unique aesthetic, emergent from the styles and meanings (re)produced from a complex of negotiations between humans and machines.

The article first gives background on copyright bots and some problems of artistic expression that attend the automation of copyright law. Following this, is a brief discussion of machinic enculturation as a unit of aesthetic analysis in sample-based music. A rhetorical framework is offered as a means to analytically assess the aesthetics of speaking to machines, for people. Drawing on discussion boards and public commentary on copyright bots in mashup production, we illustrate that, among the cultural practices of mashup artistry, are instances of machinic enculturation. From an exploratory analysis we demonstrate that musical compositions, which successfully “speak to” machines, carry an aesthetic, marked by moves to speak to, and past, non-transparent systems that privately apply public law.

Problems of automated enforcement of copyright law: art finds a way

In response to perceived problems of copyright infringement amid the increasingly insurmountable quantities of content that they host, platforms, like YouTube and SoundCloud, have begun to use automated “content identification” technologies (YouTube 2019; SoundCloud 2011) in concert with media-tracking solutions (Zefr 2019; see also, Kastrenakes 2015). “Copyright bot” is a shorthand term for the automated software agents that discover, flag, and/or threaten litigation for infringement of identified content. The algorithms that undergird these technologies are proprietary, so it is difficult to tell exactly how they work, but, based on public technical documentation and discussion, content identification systems operate by way of referencing known media scanned to a database and comparing it to subsequent uploads on a given platform (Audible Magic 2019; see also Kaiser 2018). When a copyright holder submits a piece of audio media for content identification, a model is generated, which tracks the characteristics of that media piece to create a “fingerprint” of the track (Audible Magic 2019). Because “fingerprinting” works by “measur[ing] differentiating perceptual characteristics of the content itself and does not rely on metadata, watermarks, or file hashes” (Audible Magic 2019), it is likely that some sort of machine learning system is used to track patterns, such as beats per minute, time signatures, timbre, chordal melody, et cetera.

Copyright bots have been critiqued as erring too heavily on the side of owners of copyrighted material, creating a chilling effect on fair use (e.g., Electronic Frontier Foundation, 2019). That is, out of fear of such repercussions as having one’s content “demonetized” or having one’s page removed altogether for copyright “strikes,” users are avoiding the use of copyrighted media, even in cases of fair use. Because the bots tend to operate from imperative rule systems, they can overreach, flagging content, despite persons’ use of media arguably falling within the bounds of fair use. A well-publicized example can be found in a legal exchange between Lawrence Lessig a legal scholar; see Lessig, 2004, and an Australian record company, after a video of Lessig’s was flagged for copyright infringement. As part of the lecture, Lessig was critiquing a few examples of mashups. When the audio from those examples made it into the video of his lecture, the video was flagged by the YouTube content identification system, and subsequently, the video was blocked from public viewing, despite the academic nature of the video, arguably placing it within the scope of fair use (see Sydell 2013).¹ As we will show in this article, some mashup artists, instead of quibbling with the bots and their owners, are adapting their compositional techniques to (potentially hostile) machinic audiences—a class of adaptation that we will eventually develop as machinic enculturation.

To be sure, others have illustrated that musical forms evolve alongside recording and composition technologies, including sample-based music and its aesthetics. Katz (2010) locates moments in which the features of recording technologies encourage particular types of musical expression. He calls these phenomena, phonographic effects. For example, sampling technology made it possible to compose Public Enemy’s “Fight the Power.” In particular, the samples used to compose the song can be conceptualized as “performative quotations,” taken from various lesser-known, but nonetheless, iconic, African American artists (162). The consequence of which is a particular form of music that celebrates the contributions of musical champions, made possible by sampling technology. We are interested to underscore the claim that technology can encourage particular compositional choices in music. Only, we are not as interested in the possibilities of compositional choice afforded by evolutions in recording technology as much as the perceived array of choices available to artists with respect to their (machinic) audiences. If sample-based music involves the technologically facilitated transformation of specific recorded music performances into melody and rhythm, as Katz discusses it (174), we are interested in the affectations added to those performances, meant to skirt the attention of algorithms, creating an aesthetic meant for machines, but appreciated by people, nonetheless feeding into the dialectic ebb and flow between aesthetics, copyright, and technology.

With regard to the dialectic (re)shapings of musical forms in sample-based music, Sinnreich (2019) defines a “five moments” framework for appreciating the dynamic, punctuated interactions of (1) laws and regulations, (2) market dynamics, (3) codes and practices, (4) music technologies, and (5) concepts of authorship that attend aesthetic innovation. Of importance to the current discussion of mashup production and copyright bots, is Sinnreich’s explanation that the lobbying efforts of the recording industry were successful in expanding copyright law to incorporate “phonographic copyright,” which includes the qualities of a given recording, not just the musical notation used to compose it (430). Consequently, market dynamics shifted to incorporate litigation as a key aspect of the business-as-usual for music production companies. Sinnreich further illustrates the influence of music technologies in shaping legal conceptions of intellectual property, exposing a related change in conceptions of authorship, which incorporates the producers of recordings, not just the writers of scores. For example, recent rulings in favor of producers using unauthorized samples, represent evolutions in law and regulation as well as conceptions of artistic craft and expression (434–435). In the case of copyright bots, we hope to draw attention to a similar evolution of musicality having to do with emerging technological practices of sample-based production, moored in techniques of making music for machines, not as constraints on art, but rather as listening audiences.

Perceptual technics are proposed by Sterne (2012) to designate the entanglements of scientific, social, and market systems to shape the outcomes of media forms (53). For example, Sterne shows that compression—the undergirding technology of. mp3 music files—has historical antecedents in telephonic technology, where sonic bandwidth was compressed to let more people talk on the same line (42); originally optimized for profit, but eventually evolving into a core feature enabling the promiscuity of the .mp3 file, or its amenability to casual sharing beyond market interests. In the case of copyright bots and art, we will demonstrate that attempts to constrict the promiscuity of some media forms, driven by market interests, have evolved practices of art that deal in promiscuous media, shaping their aesthetic value in new ways. Surely, mashup artists negotiate with the available means of composition to obtain, manipulate, and share digital music samples to make music. The means of musical possibilities are important to consider aesthetically. But, on the same token, the audiences for whom mashup artists compose their music in the first place also evolve. Increasingly, machines occupy the listening publics of mashups.

Underscoring problems of automated copyright enforcement concerning public expression are Bar-Ziv and Elkin-Koren (2018). Chief among the problems they outline is that automation makes copyright claims trivial to wager, drastically reducing risk to the complainant, while also introducing barriers to due process for defendants who might be engaging fair use. As Perel and Elkin-Koren (2017) point out, such problems are exacerbated by the legal bind between the need for decision-transparency and the trade secrets of private companies who use proprietary algorithms to enforce legal actions (193). The problem inherent to this is the lack of accountability regarding the ways private companies are assessing things like originality, substantial similarity, or permissibility of use (197). Such problems have been pointed out by others as quelling “one of the basic bargains of culture: people celebrating and participating in culture without having their energy and ‘labors of love’ exploited for profit” (Soha and McDowell 2016, 10). And, beyond this, the opaque application of copyright law in ways that prioritize the copyright owner manifests a “natural-rights based” assumption, wherein the author is imagined to have full control of their work in totality when, in actuality, the spirit of US copyright law is one that recognizes the utilitarian necessity of intertextual expression (Bridy 2019, 315)—being able to quote one’s champions or to parody one’s detractors. Others have articulated copyright bots as the “emergence of a new panoptic tower,” imposing surveillance on users, and consequently, representing corporate control of free expression (Birk 2015, 255). Because the bots can flag any use of copyrighted work, regardless of the nature of the use, there is reason to believe that the bots are not serving the law, or the freedom to engage public expression, well. Put in terms of the current argument, copyright can be applied to close off opportunities of expression, founded in problematic assumptions about what it means to “own” an idea. And so, it follows that when copyright enforcement is automated, it takes something that already might be potentially deleterious for art, and makes it further rigid, relentless, closed.

Legal scholars have already begun exploring necessary revisions to copyright law to better protect producers who might be engaging fair use, while keeping sight of the necessity of automated copyright scanning on the information-saturated internet, despite the imperfections inherent to automated assessment of fair use (e.g., Solomon 2015; Menell 2016; Depoorter and Walker, 2013). Others offer expeditiously pragmatic approaches to the problem. Perel and Elkin-Koren (2017), for instance, offer “black box tinkering” as a means for users to exercise their rights to test copyright identification algorithms by systematically uploading (non)copyrighted material in order to discover the thresholds and procedures of automated copyright enforcement (202). Perel and Elkin-Koren position the insights garnered from black box tinkering as those supporting social activism (212). If one can learn how “fair” the bots are, the more equipped that person will be to defend themselves in court or to motivate revisions to copyright law itself.

What we will describe as machinic enculturation is related to black box tinkering in the sense that it involves learning how the algorithms detect (or not) copyrighted material. Only, instead of learning how the machines do, or do not, detect uploaded material in order to hold private companies accountable, machinic enculturation involves a rhetorical sensibility of effective expression to the mechanisms that private companies utilize for private enforcement of public law.² While copyright bots seem to pose constraints on artistic expression, it is important to recognize that they are also generative, spawning newer forms of expression. Put differently, art finds a way, despite the bots, and leaves us in awe when it does.

The aesthetics of machinic enculturation: speaking past machines, for humans

It takes little effort to appreciate that with digital technologies come distinct aesthetics. In listening to electronic dance music or hip hop one immediately recognizes that these particular sounds smack of “digital signatures,” or the coloring of media content by the mediatization of digital production technologies, such as glitch, auto-tune, delay, and so on (Brøvig-Hanssen and Danielsen 2016, 7). In such cases, where at least some of the qualities to be appreciated are those left by the technologies themselves, there are also specific cases wherein the aesthetic is entangled with machinic creation. Accompanying the performances of machines who run on randomization software or machine learning systems to post imaginative poetry (e.g., @nabgbot 2019) or create their own musical content (AIVA 2018) is a sort of machinic aesthetic. McCormack and Dorin (2001) describe such media experiences as ones of the “computational sublime,” or the “feelings of pleasure and fear in the viewer [or listener] of a process realized in a computing machine” (12). That is, when we hand over creative control to a machine, its products do not cease being art media. Rather, they gain a unique aesthetic, which floats in that cloudy estuary between human artifice and natural processes, where media “just happens.”

Agency, as shared between humans and machines in the form of “intra-actions” (Barad 2007), is a conceptualization useful for pushing on the notion that when humans use machines they are simply static tools, and rather, to acknowledge that they are active agents with whom we collaborate to create outcomes. The negotiation of agential outcomes with machines has been studied within the grander (ethically charged) metaphor of “hospitableness” (Brown 2015) wherein some online spaces, and the machines that occupy them, can be accommodating (or not) to different modes of expression. Machines themselves have also been studied concerning the “machine aesthetic” (Brummett 1999) that accompanies contrivances when we take them seriously as objects of aesthetic value. In each of these veins of thought, machines are imagined as co-collaborators or as objects of aesthetic value in and of themselves. We wish to push this further by discussing machines not merely as shapers of human expression for other humans, or merely as objects of experience, but also as audience members whose “ears” are considered in contemporary media composition.

As such, we are interested to describe digital signatures that represent communication practices adapted to machinic listeners. If copyright can be automated to regulate the types of creativity that are engendered on a given platform, we think it is also helpful to recognize that the automated enforcement of copyright can also generate creativity by forcing human users to consider their (machinic) audiences. The results are media compositions, colored by a unique aesthetic, entangled with the phenomenon of machinic enculturation, a term that we will operationalize now.

When one learns to speak with particular regard for who will hear what is said, and for the context in which that utterance is to be heard, one can say that person has been communicatively enculturated. Learning how to make hit songs, or popular articles, or videos that get large numbers of views is an exercise in enculturation, requiring the producer to consider how the content and style of their media will be perceived by their audiences. Responding to the expectations of one’s audience, in other words, is to represent communicative enculturation. As the coming analysis will show, there are media production practices that are responding to the perceived expectations of the machines that are listening to mashup compositions. Machinic enculturation is a means by which to name the phenomenon of adapting one’s media practice to the perceived expectations of machinic audiences.

The class of rhetorical sensibility we identify in this paper as endemic to machinic enculturation is similar to Hong’s (2016) concept of machinic sensibility. Hong, in his discussion of the phenomenon of the quantified-self, identifies a set of skills and practices of perceiving data about the self, encouraged by the ability of machines to sense data independently of the human. Where machinic sensibility is concerned with the habits of knowing self as they are entangled with the sensory data of our devices (“Who does the data say I am?”), machinic enculturation is distinctly focused on the sorts of communicative adaptations meant to forward expression to machinic audiences (“How do I speak to audiences, at least partly constituted by machines?”).³

As scholars whose perspectives on media are informed by the rhetorical tradition, we find it tremendously productive to think about media practices as functioning to influence by successfully negotiating the audiences and constraints that exist in a particular public. More specifically, we will be analyzing human expression as responses to “rhetorical situations,” where communication can be analyzed by triangulating audiences, constraints, and reasons for engaging communication in a given case (Bitzer 1968). Only, instead of imagining human-only audiences, we are carving space to incorporate the machinic agents that also constitute online audiences. Where navigating constraints on expression would involve consideration of the sources of invention available to a producer it is useful to assess copyright bots as targets of that invention. That is, constraints are navigated as instances of “do” or “do not.” “Can I add reverb to this sample?” Audiences, on the other hand, are navigated as matters of “hedge” or “hedge not,” requiring qualification of one’s expressions to ensure that they “land” well. “Should I say it outright?” It is in this sense that the generative nature of copyright bots in mashup production manifests. Among the targets considered by artists as they navigate constraints to compose their expressions are machines, influencing how their expressions are hedged, generating new musical forms. In recognizing machinic enculturation, we can begin to analyze the expressive dexterity of compositions that respond to their constraints, purposes, and (machinic) audiences. And, with such analysis, we can recognize the emergent aesthetic represented by the digital signatures of such compositions.⁴

Mashups—pieces of sampled audio composed from (often unauthorized) media sources—can possess an exciting maverick ethos, which Gunkel (2016) compellingly illustrates in his discussion of the work of mashup artist, Danger Mouse.

When Danger Mouse cut up and recombined the Beatles and Jay-Z, he did so without seeking the prior approval or permission of either party. In fact, part of what made Danger Mouse’s The Grey Album so interesting and seemingly “dangerous” was the fact that it deliberately violated the integrity of the source material, taking it over, reconfiguring it in new and interesting ways, and making it do and say things it was not initially designed for. (166)

This “dangerous” aesthetic is enhanced in some cases where the features of a given mashup represent the digital traces of machinic enculturation, and by consequence, an indignance for the bots: “Oh, bots, all I do, I do in spite of you.” Conversely, in other cases, machinic enculturation can represent a more pragmatic sense of cleverness: “Wow, I can’t believe they pulled that off!” As such, our aesthetic framework is one founded on the unique persona—the ethos—that emerges from compositions, colored by stylistics meant to hedge expressions amid rhetorical situations inhabited by (machinic) audiences.

Drawing on rhetorical theory to access the aesthetics of machinic enculturation

The appeal of compositional choices in traditional sample-based media can be accessed by comparing the original and the mashed version to discover the four basic operations of rhetoric: omission, transposition, addition, and transmutation (Coleman 2013). That is, by looking at the original pieces of media used in a given mashup, the analyst can locate moments where media were added to one another, pieces of media were removed, the original ordering of media pieces rearranged, and/or moments where original pieces of media are modulated by effects (e.g., delay or pitch-shifting). Doing so helps the analyst discover and appreciate the compositional character of a given mashup. In the current case, we will be focusing primarily on stylistics that represent adaptive media practices meant to respond to the presence of machinic “ears.” In the very same way that we might appreciate the cleverness of a political critique that “dog whistles” past a censorship-happy tyrannical government to nonetheless forward public expressions, we too can begin to appreciate the rhetorical prowess of artistic expressions that stylistically maneuver (machinic) audiences. The something “more,” residual to the stylistics of a given composition can be accessed by applying the basic operations of rhetoric.

The basic operations of rhetoric were first articulated in the ancient Roman text, Rhetorica ad Herennium. Alongside the basic operations, the text also outlines a stylistic framework. Though there are over 40 individual figures of diction described in Rhetorica ad Herennium, each foreseeably with its own contribution to understanding public expressions, marked by machinic enculturation, three figures emerged as particularly relevant to our analysis: antithesis, apostrophe, and reasoning by question and answer. Antithesis is a style marked by combinations and contrasts of ideas against one another (usually involving ideas that might not logically fit together) (Anonymous 1954, 283). Apostrophe is the expression of indignation or grief at a specific person, place, or thing (Anonymous 1954, 283–285). Reasoning by question and answer is a style of diction in which one interjects answers to their own questions (Anonymous 1954, 285–289).

Examples will be elaborated in the following analysis, but to sum, the four basic operations of rhetoric (omission, transposition, addition, and transmutation) aid in discovering the compositional character of mashups as their source media are cut up, taken away, added, and transformed, including those that are aimed at negotiating (machinic) audiences. The figures of diction (antithesis, apostrophe, and reasoning by question and answer), on the other hand, offer a framework for interpreting the meanings of those compositional characteristics, and more specifically, the digital signatures aimed at the “ears” of machines as they shape the ethos of a given mashup.

The following analysis draws on public discussions of mashup production techniques and user profiles on SoundCloud to identify some practices of machinic enculturation. Guided by these same fora, we identify an example of a mashup, which can be “read” as machinic enculturation. By way of musicological examination of the stylistics of the mashup’s composition via reference to the source media incorporated into the piece, we throw light on some aesthetics that can attend compositions, produced for (machinic) audiences.

Machinic enculturation and mashup music production

In a Reddit forum discussing how to avoid having sample-based music tracks removed from SoundCloud for copyright infringement, one discussant shares their production process, prior to uploading their files:

I always take my final remix audio file, and downpitch the whole track by 4 to 9 cents and they never catch it hell i have whole dj sets with many very popular edm tracks in them and i do the same thing to them to avoid soundcloud from removing them. you cant even tell they were tweaked at all fam (mrmauzer 2016)

It is debatable as to whether or not such moves actually skirt algorithmic detection or not. Some argue that such production practices might have worked in the earlier days of content identification, but are largely futile nowadays (e.g., see Douglas 2018). Regardless, the comment illustrates a clear example of machinic enculturation: media practice adapted to the presence of machines among one’s audiences. The author is talking about production moves not meant to add clarity to the recording, or to enhance the sonic appeal of the music for human hearers—these production moves are explicitly for the machines. While the example is clearly one of machinic enculturation, it still might not represent an aesthetically interesting one. For that, we turn to a video from Zen World, a popular electronic dance production channel on YouTube.

In a video, entitled, “Stop Your Remixes from Being Taken Down,” the Zen World vlogger conjectures several “tips” for producers to avoid detection when they post their mixes to SoundCloud. Using samples from the funk band, Kool & The Gang, and the electronic music group, The Prodigy, the video illustrates the implementation of the compositional “tips” for editing samples of copyrighted music in ways that are difficult for the bots to detect.

1. 30–60 seconds of silence [at the beginning of the track] (Zen World 2017, 02: 37)

2. Don’t add bootleg in the title [of the track posting] (Zen World 2017, 03; 10)

3. Remix creatively instead of using [original sample files] as is

   1. [This involves, pitching up or down, equalization changes, converting samples from stereo to mono, transposing rhythmic elements, multiband compression] (Zen World 2017, 03: 27)

4. Don’t remix popular songs, mashup unpopular songs (Zen World 2017, 05: 03)

5. Don’t use more than 8 bars of a master stem [the original sample file] (Zen World 2017, 05: 24)

The phenomenon of machinic enculturation, because it often involves the specialized implementation of the production techniques of traditional sample-based music, is difficult to analyze in the “wild.” Even when a track represents what can be inferred as digital signatures of machinic enculturation, it is difficult to verify simply on the formal merits of a given track. As such, our analysis is augmented by a tracing of public discussion between mashup artists and their approaches to copyright bots in an online forum. Drawing from the example tracks shared between the producers, we locate a case of mashup composition that can be studied as an example of machinic enculturation.⁵

The online forum we consulted in our analysis is at Serato.com on a discussion, entitled, “How to stop your mixes from getting kicked off soundcloud.” The forum houses a discussion that has been growing since 2012, and is still open today. On it, mashup artists share techniques for mixing and affecting their tracks to keep from drawing the ire of the copyright bots (Capo Status, 2012). Many of the points of discussion resemble that of the “tips’’ listed above. One user, in particular, shares some advice while pointing to the work of another producer as an example. In answer to some previous posts about the importance of the introduction of a given track, the user geeunot (2014) states that “put[ting] a drop with lots of stuff going on for 20 secs seems to work. If you search up DJ Primetyme on soundcloud, he has a ton of stuff with those drops.” Upon some investigation into the list of tracks on DJ Primetyme’s SoundCloud account, we found a track that implements the compositional approach described by geeunot. Beyond including the “drop” in the introduction, the track also implements samples from a well-known artist, making it an interesting example to read as machinic enculturation. Because we are interested in probing the aesthetics entangled with the discourses of new cultural practices more so than calling out an artist and what they may (or may not) be doing with regard to copyright, we will analyze the composition as an illustrative example more so than a definitive case of machinic enculturation.

Mashing Drake’s “In My Feelings” for feelingless machines

On DJ Primetyme’s SoundCloud page is a track, entitled, “Drake—In My Feelings—DJ Primetyme Mashup” (Primetyme Official 2018). The track mashes original instrumentation and samples with Drake’s “In My Feelings,” a hip hop song that maintained status as a “Billboard Hot 100” for two consecutive months in 2018 (Trust 2018). The top hit is known for its catchy New Orleans bounce rhythm and famous accompanying “Kiki Challenge,” a viral internet phenomenon, wherein people would play the song and video record themselves dancing next to a driverless car, rolling through traffic (Hussein 2018; see also, Lyons 2018). Primetyme’s mashup of the song is a breakbeat rendition of the New Orleans bounce rhythm of the original. The piece starts with a breakbeat intro, painted in the aural likeness to sounds one would expect from an early model analog drum machine, such as a Roland “808.”⁶ The breakbeat noticeably omits nods to the source track, but eventually is accompanied by vocal samples that rhythmically chant, “Hey!” alongside a bass-heavy synthesizer melody. About thirty seconds into the song, a sample of the main refrain from “In My Feelings,” is introduced (original track available at Drake 2018). The sample itself is pitched up and filtered, ostensibly with a high pass filter, giving the vocal sound a telephonic quality, augmented by a reverb effect added to the sample. Despite the overall rhythm of the mashup itself having a sort of “in your face” character, the affected, “In My Feelings” sample adds an atmospheric quality to the whole track, as if one were listening to dance music on a space station and Drake was just outside, contributing his voice to the performance.

Stylistic antithesis: it is impossible; but there it is

The transmutation of the source track in the mashup offers a unique experience with the original song. But, further, when considered in light of the discussion of the “tips” for mixing for copyright bots discussed in the previous section, it also emerges as an experience of machinic enculturation. In the same way that one might be in awe of a graffiti artist, who has somehow figured a way to ascend a building to create a mural in the middle of the night without being spotted, here is a mashup composition, posted to a public-facing site, which incorporates a well-known, recognizable sample from a song assuredly being hunted by copyright bots. Such an experience is one of antithesis, in which the experience is colored by a paradox: “According to physics and the law, this should not exist; but because of physics and the law, here it is.” Adapting one’s paintings to city walls and highway signs are examples of enculturation aimed at the technologies of community, which, in and of themselves, are integral to the artistic experience of graffiti murals. We are in awe of such art pieces to forward expression, despite inhospitable circumstances, just as much as we can be captivated or lost in the particular content of a given piece. The case of this mashup is the same. Woven into the very experience one has with the catchy mashup, is a sense of awe, spoken through an antithesis, realized in the acknowledgment that this art has successfully achieved expression to a public comprised of humans and machines.

Stylistic apostrophe: art in spite of machines

Later, toward the middle of the mashup, punctuated by staccato air horn samples, the song deconstructs into a dubstep breakdown.⁷ “Wobble bass” transitions to a guttural, grinding bassline that sits under a choppy sample of the vocal line of the City Girls (guest artists on “In My Feelings”), replete with the same telephonic quality as the earlier Drake sample. The additions of the horn sounds and the “dirty” sounding wobbly bassline add an indignant, but confident mood to the transition. In particular, the samples of City Girls are fore-fronted in a sort of taunting rhythmic manner that builds to a crescendo, as if to shout: “Yes, these samples are being used!” One could infer that the targets of this apostrophe—or, expression of indignation—is Drake, the City Girls, or the labels that represent them. However, among those targets to be considered are the copyright bots. “Oh, copyright bots, what I do, I do in spite of you.” The listener is enticed to take on feelings of righteous indignation, potentially toward humans, but also, and perhaps more profoundly, toward the machines that work for them. Listening to the song is a triumph for poiesis—artistic creation—in the face of potentially disapproving (machinic) audiences.

Reasoning by question and answer: sampled call and (human) response

Nearer to the end of the mashup, are a series of calls and responses, fashioned from samples of both Drake’s and the City Girls’ vocals from “In My Feelings.” Both are pitched very high and cut at syllabic intervals from the original track. The result is a series of melodic, but lyrically impenetrable sounds that slightly resemble the original essence of the source vocals, but which are otherwise difficult to identify as such. Transposing the syllabic content of the vocal samples and contrasting them in such a manner as to pose a question with one type of sample, and to answer with the other, allows the listener to contrastively discover that, despite being drastically edited, these are in fact samples taken from Drake and the City Girls. This performance of reasoning by question and answer adds an almost conversational quality to the track, fluently spoken through the samples that compose it. By consequence, the splitting of melodic rhythms is a means by which to incorporate recognizable samples from the original track while also adapting them to “speak past” the copyright bots. The experience is one that is riveting to a select (non-machinic) section of the audience. Demonstrating a pragmatic cleverness, the composition of the samples forces questions on the audience (e.g., Who are these samples from?), while providing answers (e.g., Drake and the City Girls), but those answers are distinctly hedged toward one section of that audience (people), while also hedging them away from the other (the bots). Just as the song began, it shifts back to breakbeat, transitioning into an outro that revisits the starting theme.

Conclusion

Mashups are being produced for (machinic) audiences, involving specific techniques shaping the sounds of those compositions, meant for the “ears” of machines. We named this phenomenon of adapting one’s expressive practice to machine audiences as machinic enculturation. By looking specifically for compositional omissions, transmutations, additions, and transpositions relevant to persuading machinic listeners, we illustrated that some sample-based music can be colored by an aesthetic, entangled with machinic enculturation. That is, one of the things to be appreciated about the compositional character of some mashups is their rhetorical prowess of expression to machines, for humans. The contribution of this article is not so much in the fact that machines can make media for persons, or even in instances where machines share creative agency with persons to create media in a sort of “cyborg” shared agency between human and machine (Gunkel 2000) as they produce media. Rather, it is from the other direction, wherein machines are approached as consumers of media. Machinic enculturation, in the specific case of mashup production, can be conceptualized as a form of algorithmic thinking, centered on the discovery and integration of procedures of musical composition (Collins 2018), derived from the perceived expectations of machinic audiences.

Reading DJ Primetyme’s mashup of Drake’s “In My Feelings” with machines in mind as audience members, and specifically through the frames of antithesis, apostrophe, and reasoning by question and answer—exposed the aesthetic value of compositional practices adapted to machinic audiences. The application of rhetorical theory to mashups in this article complements the work of others, which investigates specific tactics used by artists to engage rhetorics of remix to wager parody and political critique (e.g., Brøvig-Hanssen and Sinnreich 2019; Kuhn 2012). Specifically, the framework expands conceptualizations of audience by clearing space for acknowledging the machines that are now increasingly listening to the work of artists, demanding rhetorical response.

And, to that end, locating sites of human-to-machine expression is a means of discovering insight regarding the “emergent bias” of computer systems (Friedman and Nissenbaum 1996, 335). The systematic, and debatably “fair,” application of computer systems to enforce copyright, reveals the biases of copyright bots, which often prioritize copyright holders over persons attempting public expressions. Our study reifies the need to complicate the simple metaphor of the singular media platform—e.g., SoundCloud—by locating the “extra” software that runs over the top of a given platform (what Geiger (2014) names “bespoke code”). Sure, SoundCloud has constraints built into its platform. But trafficking that platform are third party services, with their own expectations about what constitutes legitimate artistic expression. To this end, it is analytically useful to approach some technologies as audiences rather than constraints. Using SoundCloud as a platform for hosting music is different from adapting one’s media composition practices to the biases of the third-party bots that listen to that music.

Approaching technologies as audiences, of course, is not simply a matter of the audience. As with any audience, there are textured arrays of expectations because there are textured arrays of bespoke code running on any given platform. For example, third-party code integrations are emerging in the form of services designed to identify and flag copyrighted material not punitively, but rather for sharing royalties. On the face, these services approach copyright and music production from a middling stance, encouraging intertextual public expression, while also navigating commitments to capitalistic models of art-making. For example, one such service, MixBANK, ostensibly encourages producers to compose mixes in ways that make samples more explicitly detectable, so its bots can perform automated copyright clearance. Upon clearance, mixes are then distributed to online music streaming platforms, including SoundCloud (Turner 2019). Through this process, MixBANK promises monetization of the (re)mixes for both content creators and copyright holders (Stassen 2019).⁸ The profit-sharing model of MixBANK introduces a motivation for producers to demonstrate machinic enculturation not to convince the bots their content is not worth flagging, but rather to make one’s samples more obvious to the machines. Even still, while producers might modulate their musical samples to adhere to the bots’ expectations, they are still practicing textured negotiations of art-making, which nonetheless sometimes call for speaking past machines, for humans. For instance, in a forum post, on Reddit, entitled, “How to put your mashups legally available on itunes, Apple music, Spotify and TIDAL,” a user posts a query regarding how to compose for the MixBANK bots, alongside automated feedback they received from the service: “The content owner of this segment created a rule that prevents the content from his catalog from being used for more than 75% of its original duration” (quoted in Hijimmylin 2018, emphasis in original). A fellow user responds with a tactic they use to thwart bots on YouTube: “cut each individual bar of the song and rearranged [sic] it in a random (but fitting) order” (GravitasMusic 2018). By way of compositional transposition, the producers are attempting to meet some of the bots’ expectations (e.g., by “showing” the sample), while also practicing strategic ambiguity to dodge others (e.g., by obfuscating the extent to which the source material is being sampled).

Though we have focused on art-making, we are hopeful that scholars from across the academy will consider taking up the phenomenon of machinic enculturation in contexts beyond art, not only to help “check” the automations with which we increasingly live, but also, to explore modes of expression that are effective, clever, and beautiful in their persuading of (machinic) audiences.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Notes on contributors

Miles C. Coleman

Miles C. Coleman is an Assistant Professor of Communication Studies. His research focuses on the values that exist in otherwise “objective” phenomena and the ethical implications of new media.

Mark Anthoney

Mark Anthoney is a Lead Instructional Designer. His research in media studies and library and information science concentrate on the interplay between emerging cultural practices and new media technologies.

Notes

1. It is in this way that such automations exacerbate problems such as “copyright trolling,” or the aggressive litigation of copyright infringement, often solely for monetary gain through “settlements priced just low enough that it is less expensive for the defendant to pay rather than to defend the claim, regardless of the claim’s merits” (Sag and Haskell 2018, 571; also, Melendez 2018).

2. Our pursuit of the (re)shapings of music practices, wrought by copyright law, finds inspiration from the innovative work of scholars participating in the MASHED research project at University of Oslo, directed by Ragnhild Brøvig-Hanssen: https://www.uio.no/ritmo/english/projects/mashed/.

3. Although our specific case study of machinic enculturation is one of art-making, the term, foreseeably could be applied to other realms of communication too (e.g., political or scientific communication).

4. Important to espousing our assumptions of analysis is the point that mashups are collaborations, explosions of creation, mishmashes of nature and human ingenuity, and, as such, for the analyst, it is not necessarily productive to search for the work of a single author. Such an approach unnecessarily constrain analysis while feeding into the problematic assumptions of ownership as being had by one person (which is seemingly antithetical to the spirit of remix culture) (e.g., see Logie 2015). Therefore, in the likeness of other scholars (Inacio da Silva 2015, 102), we will not be examining feats of genius on the part of mashup artists. Rather, we will be pursuing the persona that emerges from a given piece of music. Enmeshed with the compositional structure of every piece of mashup music, is an emerging ethos. For our analysis, who the actual producers are does not matter. However, integral to our analysis is the performative persona and the meanings that attend the character of a given mashup composition.

5. This approach to the analysis also underscores that machinic enculturation is a phenomenon shaped by public discourses, and, as such, its aesthetics are accessible not simply via the formal qualities of a given mashup, but rather by tapping into the collective conversations of producers as they make music for machines.

6. Focused on the introduction as an important element for adapting past the copyright bots on SoundCloud, is Dj Rehab (2014), who writes: “Take the time to make an intro. Period, it works. I have 50 mixes up from the top 40 to my own mashup megamixes. If you really want your music heard, Soundcloud is the way” (further examples available at Probably Chris 2019).

7. Mighty Dragon Sounds (2014) is a user who suggests composition choices that have to do less with creating an original sounding introduction, but rather with adding sounds at specific intervals throughout the song. “I drop a whole bunch of Airhorn and bomb drop samples at the beginning middle and end of the song ….They tried to stop me from uploading an few beats that I had created ….I guess the Bomb Drops an Airhorn and Reggae Sound Effects screw with the auto detection software they have running” (further examples at Mighty Dragon 2019).

8. Mixcloud.com is yet another example. This platform does not use bots to flag and remove content as much as it uses bots to identify particular source tracks in mixes so as to pay royalties to copyright holders (Mixcloud 2019).