Further Experiments in Agent-based Musical Composition
Photo by whistler1984 @Flickr.
Editor’s Note: WARNING: THE FOLLOWING POST IS INTERACTIVE!!! This week’s post is especially designed by one of our regulars, Andreas Duus Pape, to spark conversation and provoke debate on its comment page. I have directly solicited feedback and commentary from several top sound studies scholars, thinkers, artists, and musicians, who will be posting at various times throughout the day and the week–responding to Andreas, responding to each other, and responding to your comments. Look for commentary by Bill Bahng Boyer (NYU), Maile Colbert(Binaural/Nodar, Faculdade de Belas Artes da Universidade do Porto), Adriana Knouf(Cornell University), Primus Luta (AvantUrb, Concrète Sound System), Alejandro L. Madrid (University of Illinois at Chicago), Tara Rodgers (University of Maryland), Jonathan Skinner (ecopoetics), Jonathan Sterne (McGill University), Aaron Trammell (Rutgers University, Sounding Out!) and yours truly (Binghamton University, Sounding Out!). Full bios of our special respondents follow the post. We wholeheartedly wish to entice you this Monday to play. . .and listen. . .and then share your thoughts via the comment page. . .and play again. . .listen again. . .read the comments. . .and share more thoughts. . .yeah, just go ahead and loop that. –JSA, Editor-in-Chief
—
I’m a musician and an economist. Sometimes you will find me playing acoustic folk rock and blues on guitar, harmonica and voice. And at other times I will be at work, where I apply my expertise in game theory to the computer modeling of social phenomena. I create simulations of people interacting – such as how people decide which way to vote on an issue such as a tax levy, or how people learn to sort objects given to them in an experiment. In these simulations, the user can set up characteristics of the environment, such as the number of people and their individual goals. After things are set up, users watch these interactions unfold. The simulation is a little story, and one need only tweak the inputs to see how the story changes.
As a musician, I was curious if a program that generates social stories could be refashioned to generate musical pieces. I wanted to build a music-generation engine that the listener could tweak in order to get a different piece each time. But not just any tune – a piece with some flow, some story. I like that tension between randomness and structure. On one hand, I want every song to vary in unpredictable ways; on the other hand, I want to create music and not structureless noise.
I created a basic story of predators and prey, whimsically naming the prey “Peters,” represented by rabbits, and the predators “Wolves.” My simulation depicts a plain in the savannah with a green oasis. The prey seek the oasis and the predators seek the prey. Each character has its own goals and the closer they are to achieving them, the happier they are. Both predators and prey want to have stomachs full of food, so naturally they want to be close to their target (be it prey or oasis). As they travel through the savannah, they learn what choices (directions of movement) make them happier, and use this experience to guide them.
Photo by bantam10 @Flickr
So how does this story become music? To this question there are two answers: a technical one and an intuitive one. The intuitive answer is that in real life the story of predators and prey plays out geographically on the savannah, but musically this is a story that plays out over a sonic landscape. To elaborate, I abstracted the movement of the prey and predator on the geography of the plain into the musical geometry of a sonic landscape. The farther north an agent travels, the higher the pitch. And, the farther east an agent travels the longer the duration. In other words, as an agent travels to the northwest, she makes longer-lasting tones that are higher pitched. I also mapped happiness to volume, so that happy agents make louder tones. Finally, so that each agent would have a distinct voice as they traveled through this space, I chose different instruments for each agent.
In the video below I assigned the “church organ” sound to prey, and the “brass section” sound to predators.
Ultimately, there are some things that I like about this piece and others that I do not.
As a harmonica player, I improvise by creating and resolving tension. I think this piece does that well. The predator will pursue the prey into a quiet, low-pitch corner, creating a distant, rumbling sound – only to watch prey escape to the densely polyphonic northwest corner. There is an ebb and flow to this chase that I recognize from blues harmonica solos. In contrast to my experience as a harmonica player, however, I have found that some of the most compelling parts of the dynamics come from the layering of notes. The addition of notes yields a rich sonic texture, much like adding notes to a chord on an organ.
Unfortunately, for largely technical reasons, there is a lack of coherent rhythm and pacing. The programming platform (agent-based modeling software called NetLogo) is not designed to have the interface proceed in real-time. Basically, the overall speed of the piece can change as the processing load increases or decreases. I found that as agents learnt more about their surroundings (and more system resources are allocated to this “memory”), they became slower and slower. To fix this, I capped the size of their memory banks so that they would forget their oldest memories. The closest I have come to a rhythmic structure is by ordering the way that the agents play. This technique makes the piece have a call-and-response feel. If only the piece to had a coherent rhythm, then I could imagine playing harmonica along with it.
One last comment on pitch: while an earlier version of this piece mapped each step in space to a semitone, things sounded too mechanical. Even though this was the easiest and most intuitive decision from a technical standpoint, it was aesthetically lacking, so I have now integrated traditional musical scales. The minor scale, in my opinion, is the most interesting as it makes the predator/prey dynamic sound appropriately foreboding.
Photo by deivorytower @Flickr.
You can play this piece yourself. Simply go to this link with Java enabled in your browser (recommended: Google Chrome). Pressing “Setup” then “Go” will create your own run of the piece. As it is running, you can adjust the slider above the graphic window to change the speed. Press “Go” again to stop the model, adjust any parameters you wish and press “Setup” and “Go” again to see how the piece changes. Here are some parameters to try: instA and instB to change the instruments associated with prey and predators; PlayEveryXSteps to change the pace of the piece (higher results in a slower paced piece); Num-PackAs and Num-PackBs changes the number of prey and predators; the vertical PeterVol and WolfVol adjust the overall volume of prey and predators.
In regards to my version of “Peter and the Wolf,” I have a number of things that I’m curious about.
First, how does this relate to what you think of as music? Do you like listening to it? Which elements do you like and which do you dislike? For example, what do you think about about the tension and rhythm – do you agree the first works and that the second could be improved? Would you listen to this for enjoyments’ sake, and what would it take for this to be more than a novelty? What do you think about the narrative that drives the piece? I chose the predator and prey narrative, admittedly, on a whim. Do you think there might be some other narrative or agent specific goals that might better drive this piece? Is there any metaphor that might better describe this piece? As a listener do you enjoy the experience of being able to customize and configure the piece? What would you like to have control over that is missing here? Would you like more interaction with the piece or less interaction?
Finally, and perhaps most importantly, what do you think of the premise? Can simple electronic agents (albeit ones which interact socially) aspire to create music? Is there something promising in this act of simulation? Is music-making necessarily a human activity and is this kind of work destined to be artificial and uncanny?
Thanks for listening. I look forward to your thoughts.
“The Birth of Electronic Man.” Photo by xdxd_vs_xdxd @Flickr.
– – –
Andreas Duus Pape is an economist and a musician. As an economist, he studies microeconomic theory and game theory–that is, the analysis of strategy and the construction of models to understand social phenomena–and the theory of individual choice, including how to forecast the behavior of agents who construct models of social phenomena. As a musician, he plays folk in the tradition of Dylan and Guthrie, blues in the tradition of Williamson and McTell, and country in the tradition of Nelson and Cash. Pape is an assistant Professor in the department of Economics at Binghamton University and is a faculty member of the Collective Dynamics of Complex Systems (CoCo) Research Group.
– – –
Guest Respondents on the Comment Page (in alphabetical order)
Bill Bahng Boyer is a doctoral candidate in music at New York University who is completing a dissertation on public listening in the New York City subway system.
Maile Colbert is an intermedia artist with a concentration in sound and video, living and working between New York and Portugal. She is an associated artist at Binaural/Nodar.
N. Adriana Knouf is a Ph.D. candidate in information science at Cornell University.
Primus Luta is a writer and an artist exploring the intersection of technology and art; he maintains his own AvantUrb site and is a founding member of the live electronic music collective Concrète Sound System.
Alejandro L. Madrid is Associate Professor of Latin American and Latino Studies at the University of Illinois at Chicago and a cultural theorist and music scholar whose research focuses on the intersection of modernity, tradition, globalization, and ethnic identity in popular and art music, dance, and expressive culture from Mexico, the U.S.-Mexico border, and the circum-Caribbean.
Tara Rodgers is an Assistant Professor of Women’s Studies and a faculty fellow in the Digital Cultures & Creativity program at the University of Maryland. As Analog Tara, she has released electronic music on compilations such as the Le Tigre 12″ and Source Records/Germany, and exhibited sound art at venues including Eyebeam (NYC) and the Museum of Contemporary Canadian Art (Toronto).
Jonathan Skinner founded and edits the journal ecopoetics, which features creative-critical intersections between writing and ecology. Skinner also writes ecocriticism on contemporary poetry and poetics.
Jonathan Sterne teaches in the Department of Art History and Communication Studies and the History and Philosophy of Science Program at McGill University. His latest book, Mp3 The Meaning of a Format comes out this fall from Duke University Press.
Jennifer Stoever-Ackerman is co-founder, Editor-in-Chief and Guest Posts Editor for Sounding Out! She is also Assistant Professor of English at Binghamton University and a former Fellow at the Society for the Humanities at Cornell University (2011-2012).
Aaron Trammell is Multimedia Editor of Sounding Out! and a Ph.D. Candidate in Media and Communications at Rutgers University.
Share this:
Easy Listening: Spreading and the Role of the Ear in Debating
Photo courtesy of Sounding Out! intern, Sky Stage. All rights reserved.
As I write, my institution, the University of Denver, is gearing up to host the first of the 2012 Presidential Debates. The debate promises a polite and eloquent exchanging of reasons on matters of public concerns. This image of “the good man [sic] speaking well,” roughly represented by Obama and Romney, pervades popular culture representations of debate. While there are certainly some forms of debate that stress public reason, like Worlds style debate, most other competitive styles of debate, look and sound very different.
American “Policy Debate,” the type of debate I will explore today, is a high school and collegiate competitive activity where two-person teams argue the merits of governmental legislation. In every round there is an affirmative that advocates an action and a negative defending the status quo, or some alternative. The debate prompts, or resolutions, change yearly and produce research collections comparable to a master’s thesis. But, it is not the research that captures newcomers’ attention. No, it is the velocity that information travels. Indeed, highly successful debaters transform their voice into a high-speed information weapon, sometimes speaking up to 400 words per minute (wpm).
Nobody really knows who decided that speaking quicker than her/his opponents would be a viable strategy. Some allege that David Seikel and Mike Miller, a dominant team from the University of Houston in the late 1960s invented the strategy. However, Miller denies this fact, insisting that the practice of speaking quickly, or “spreading,” predated their success by a couple years. While it is difficult to pin down exactly when talking becomes spreading, studies indicate passive comprehension has a threshold of around 210 wpm. Simply put, spreading is a vocal practice that propagates more arguments than an opponent can rebut, forcing the opposing team into a strategic choice of conceding and/or inadequately responding to some positions.
Almost immediately after its introduction, spreading became a polarizing issue. While some alleged that speed undermined communicative argumentation, others applauded its ability to foster critical thinking. Rate of delivery became a site for what Douglas Ehninger called in 1958 the debate about debate: should debate be concerned with public reason or technical proficiency? Spreading figured prominently in these discussions because it circumscribed debate to the technical sphere, the quick pace precluding lay comprehension. As a site for cultivating what argumentation theorists Ron Greene and Darrin Hicks call “the ethical attributes required for democratic citizenship” (101), debate operates as an ideal problem space to inquire about the relationship between listening and judgment that underwrite argumentative exchanges.
Extracting meaning from a spread requires a trained ear that can delineate the nuance of tone, rhythm, code, and breathing, while translating these sounds into symbols that are then recorded onto a “flow” (a document where participants keep notes of the debate). This technical sphere of argumentation, then, requires what cultural historian Jonathan Sterne calls in The Audible Past: Cultural Origins of Sound Reproduction an “audile technique,” or a set of culturally defined listening practices. Here, spreading denotes not just rate of delivery, but encompasses a range of other practices that attend to the debate. Yoking velocity with technical proficiency, spreading requires an audile technique that attunes the ear to speed, while calibrating the viscera to quickly appraise each position espoused. Extending Sterne’s observations into an embodied theory of listening, I argue that spreading entrains visceral judgment that organizes expertise under the banner of “exchange-value,” instead of its veracity. Reflecting what Jodi Dean calls in Blog Theory: Feedback and Capture in the Circuits of Drive capitalism’s communicative style, I contend, audition in this case privileges flow over content.
Today, debaters are much faster than their 1960s counterparts. Take for example the championship round of this year’s (2012) National Debate Tournament (NDT). The 2012 topic debated the merits of the United States’ democracy assistance in the Middle East.
Finals – NDT 2012 Northwestern BK v Georgetown AM – 1AC + CX
In the speech linked to above, dubbed the first affirmative constructive (1AC), the speaker’s first words are easily audible, but as the seconds expire, the affirmative’s words increase in velocity; syllables collide and contort. The speaker pushes through, taking a split second to take a short, fast breath (called a clutch) before he begins his next word. To the seasoned ear, the voice is instructive; a slower, deeper tone signifying an assertion. also known as a “tag,” and a faster tempo indicating evidence. Out of this buzz, debaters’ are able to discern distinct phrases, like “Arab Spring,” “economic down turn,” and “nuclear war.” In 9 minutes the speaker outlined his advocacy, consisting of multiple scenarios for planetary extinction.
During the 1AC, the negative team is carefully listening, gathering the affirmative’s evidence, and translating the speech into symbolic shorthand, termed flowing. In addition to its purely functional use—to keep track of positions—the flow also provides third party adjudicator(s) a document to evaluate the debate.
Image of flow notes taken during a debate
After a period of questioning, the first negative speaker, linked to below, elaborates his positions: a procedural concern that the affirmative is outside the confines of the resolution; two philosophical objections (or critiques); two alternative courses of action that resolve the affirmatives exigencies, while conserving the president’s political capital, so Obama can push for Jackson-Vanik legislation; and finally direct responses to the affirmative position.
Finals – NDT 2012 Northwestern BK v Georgetown AM – 1NC + CX
Coupled with the “burden of rejoinder,” which dictates a conceded argument is a true argument, speed becomes a strategic weapon. The more positions, the more likely one will be missed, granting the other team an advantage. For example, failure to address the negative’s concerns about the Jackson-Vanik legislation means that the affirmative position would result in a US-Russian nuclear war culminating in extinction and probably losing the affirmative the debate. Even less dire claims exert force in the round. For instance, missing the relatively innocuous claim that Obama does not have political capital abrogates the negative’s whole Jackson-Vanik position. After all, what would be the point of Obama conserving capital, if he has none to begin with?
Debating thus requires a refined ear capable of quickly determining a argument’s merit, while allotting the proper amount of attention. If a claim is deemed irrelevant, the debater can focus on a more relevant position, keeping her/his ear piqued for beginning of the next position.
Debate team members listening to the flow at a 2009 Tournament at University of the Pacific; Image by Flickr User Inkyhack
This confluence of speed, evidence, time constraints, and a burden of rejoinder cultivates an instrumental relationship to expertise. Media theorist Jayson Harsin suggests in “The Rumour Bomb: Theorising the Convergence of New and Old Trends in Mediated US Politics” that these “conditions encourage a relationship of viewer to text (slogan, soundbite, fragment) which is essentially fiduciary, based on trust, not critical understanding”(101). Indeed, time constraints coupled with a proliferation of positions, produces a sound to listener relationship, where the veracity is assumed and significance is dictated by strategies, not the least of which is vocal. This is because there is simply not enough time to evaluate the credibility of each piece of evidence read. Patrick Speice and Jim Lyle, two debate coaches, explain that “Debaters have to be focused on the arguments being offered, have to be able to understand them very quickly, and they have to be able to discern which arguments are of the greatest significance for the round.” Evidence, then, is reorganized along a strategic continuum—the more likely an argument may help or hinder a team’s chances of winning, the more important it becomes. This can precipitate a “race to the margins,” where one side tries to find an obscure argument to beat back more probable analysis. This is not to say that one kind of argument is better than the other. But, the rationality used to organize the evidence relegates veracity to the epiphenomenal. This fosters an epistemic leveling, indexing expertise according to its exchange-value.
The repetitive nature of debate inscribes these listening habits into the body, shaping future interaction. Writing on the pious ear, cultural anthropologist Charles Hirschkind explains in The Ethical Soundscape: Cassette Sermons and Islamic Counterpublics, “patterns of sensory attunement configured through continuous practice of such a listening constitute an intensifying perceptual background for […]ethical agency and public reason”(28). Similarly, the practice of listening to debate rounds, along with the movement of the hands, quickness of thought, and so on are burned into an affective substrate, defining the horizons of future action. That is, through the repetition of debates, audile technique is entrained into motor memory, residing just below the conscious register of thought, acting as a potentiality. Most often, this listening technique is activated in the next debate round, the spread’s distinctive sound generating an anamnestic effect, activating the policy debate ear.
In sum, spreading proffers an audile technique concerned with the exchange-value of evidence, where the “best claim” is the one that wins the round. It entrains a technical mode of audition, aiding the debater in making quick decisions. This listening style underwrites a knowledge economy, where expertise becomes just another commodity that is bought, sold, and traded to support pre-formulated opinions. “Under communicative capitalism,” Jodi Dean writes, “an excess of polls, surveys, and assessments circulates, undercutting not only the efficacy of any particular poll or survey but the conditions of possibility for knowledge and credibility as such. There is always another survey, done by another group or association with whatever bias and whatever methodology, displacing whatever information one thought one had” (103).
—
Justin Eckstein is a doctoral candidate and former director of debate at the University of Denver. His work explores the intersection of listening, affect, and argumentation. Justin’s writing has appeared in Argumentation & Advocacy, Relevant Rhetoric, and Argumentation in Context. Currently, he is writing his dissertation on the micropolitics of podcasting in the post-deliberative moment.
Share this:
Current Top Posts
April 2021
January 2020
November 2019
October 2019
August 2019
April 2014
November 2017
March 2015
December 2017
October 2015
Sounding Out! 
Recent Comments