Cover of Computing Taste
Computing Taste Algorithms and the Makers of Music Recommendation
by Nick Seaver
University of Chicago Press, 2022
Cloth: 978-0-226-70226-1 | Paper: 978-0-226-82297-6 | Electronic: 978-0-226-82296-9
DOI: 10.7208/chicago/9780226822969.001.0001

AVAILABLE FROM

University of Chicago Press (cloth, paper, ebook)
Barnes & Noble Nook
Google Play
Kobo
OverDrive
ABOUT THIS BOOKAUTHOR BIOGRAPHYREVIEWSTABLE OF CONTENTS

ABOUT THIS BOOK

Meet the people who design the algorithms that capture our musical tastes.
 
The people who make music recommender systems have lofty goals: they want to broaden listeners’ horizons and help obscure musicians find audiences, taking advantage of the enormous catalogs offered by companies like Spotify, Apple Music, and Pandora. But for their critics, recommender systems seem to embody all the potential harms of algorithms: they flatten culture into numbers, they normalize ever-broadening data collection, and they profile their users for commercial ends. Drawing on years of ethnographic fieldwork, anthropologist Nick Seaver describes how the makers of music recommendation navigate these tensions: how product managers understand their relationship with the users they want to help and to capture; how scientists conceive of listening itself as a kind of data processing; and how engineers imagine the geography of the world of music as a space they care for and control.
 
Computing Taste rehumanizes the algorithmic systems that shape our world, drawing attention to the people who build and maintain them. In this vividly theorized book, Seaver brings the thinking of programmers into conversation with the discipline of anthropology, opening up the cultural world of computation in a wide-ranging exploration that travels from cosmology to calculation, myth to machine learning, and captivation to care.

AUTHOR BIOGRAPHY

Nick Seaver is assistant professor of anthropology at Tufts University. He is coeditor of Towards an Anthropology of Data.

REVIEWS

“Computing Taste tells a fresh story in the increasingly crowded scholarship on artificial intelligence and culture. It will be immensely useful for those outside of computer science and engineering who want to understand how people think and work in the AI industry.”
— Jonathan Sterne, author of "Diminished Faculties," "MP3," and "The Audible Past"

“Seaver’s nimble account of how contemporary music recommendation systems are conceived and crafted takes readers beyond easy oppositions of humans and algorithms to explore the captivating dynamics of taste and technics, hearing and computing, guidance and coercion.”
— Natasha Dow Schüll, author of "Addiction by Design: Machine Gambling in Las Vegas"

"Seaver’s exquisite and essential book brings us into an expert community aspiring to find the delicate balance between caring for and controlling the sprawling phenomenon of taste. The ethnographically engaging Computing Taste offers a complex rendering of the makers of music recommendation systems who believe that algorithms can predict and shape musical taste while also wrestling with the reductive absurdity of such a claim. Seaver’s theoretical creativity both pushes critical studies of technology in new directions and makes this book a joy to read.”
— Lisa Messeri, author of "Placing Outer Space: An Earthly Ethnography of Other Worlds"

“Who are the programmers writing the music recommendation recipes that structure so many of our auditory habits in these digital days? How do these new taste makers script listeners into the musical multiverses their algorithms create? Seaver brilliantly tunes us to the cadences of these people’s works and lives, decoding the mix of cosmologies, capital, and computation that channel how and what we hear today.”
— Stefan Helmreich, author "Sounding the Limits of Life: Essays in the Anthropology of Biology and Beyond"

“Perhaps there’s no accounting for taste, but as Seaver demonstrates in Computing Taste, his resonant and resourceful ethnography of music recommendation algorithms, musical taste can indeed be counted and coded. By listening to the sociotechnical dynamics of that translation process—the means by which aesthetic, subjective, social, and situational choices are transcribed into human-orchestrated algorithms—Seaver helps us appreciate not only the myriad harmonic parts that music and machines play in our personal and social lives, but also the many modes and contexts in which we listen.”
— Shannon Mattern, author of "A City Is Not a Computer: Other Urban Intelligences"

"Streaming music services are the norm today, but people don't often think about how they work or how they recommend the next song. Seaver peeks behind the musical curtain in this book about the humans behind the algorithms. . . . Music lovers and those who like books about artificial intelligence will enjoy Seaver's deep dive into the culture, data, and science of music recommendation systems. Computing Taste offers insight into algorithmic music recommendations that's entertaining and easily digestible."
— Natalie Browning, Library Journal

"Artists and music journalists have been coining genres for decades, based on sounds shared between artists. This new era for genre is derived from listener data and labelled by engineers who, Seaver says, never expected to become authorities on the matter. This speaks to the contradiction at the heart of Computing Taste: it’s both easier and harder to pinpoint a person’s music taste than you might expect. It all depends on what you think taste is. Spotify can tell us how many times we loop a favorite song, make reasonable assumptions about the genres that speak to us, and deduce from GPS data what we might want to hear in the gym as opposed to the office. But Seaver stresses that a key anthropological question remains unwrapped: why do people love the songs that they do?"
— Katie Hawthorne, Guardian

"A useful deep dive into precisely how these systems are built, the people who build them, their goals and aspirations, and much more."
— Bill Marx, Arts Fuse

"I would recommend this book if you identify with the following phrase, which is taken from Seaver’s interview with one music company engineer: ‘I’m plagued by the idea that there’s something I haven’t heard yet.’ Music nerds will especially appreciate that Seaver proposes definitions for topics that are hard to describe, like taste and genre. They will enjoy identifying their habits in Chapter 3, ‘What are Listeners Like?’ and learning how music engineers define obscure subgenres like ‘shiver pop.’ . . . The peeks into these veiled companies are almost reminiscent of spy novels. If you’re interested in start-up culture and liked The Social Network, there’s something for you in this book. Throughout Computing Taste, Seaver comments on the balancing act between artificial intelligence and human expertise. He says the title ‘is meant to index that tension’—to probe how technological systems can coexist with something as personal as music taste."
— Hope Karnopp, Daily Cardinal

"The gap between technology and culture might not be as wide as we think, says Seaver in his analysis of how music recommender systems are produced. . . . You’ll come away from Computing Taste realizing that algorithms aren’t the enemy, ready to think again.“
— Engineering and Technology

"Recommendations now 'drive close to half of all users’ streams', according to Spotify’s co-president Gustav Söderström. In Computing Taste, an ethnography of the data scientists and product managers working in 'the world of music recommendation', Seaver gives an account of the way this sort of technology operates. The job of his interviewees, who tend to work for private companies hired by streaming services, is to help their clients 'answer an apparently simple question: what’s next?'"
— London Review of Books

"Computing Taste: Algorithms and the Makers of Music Recommendation is a pleasure to read. It is well-written, with nice turns of phrase. I commend it to anyone interested in how media works in the 21st century."
— Metascience

"The central premise uniting these theories is that we can’t really tell an algorithm who we are; we have to show it. Platforms used to offer recommendations based on clear user inputs (consider that Netflix used to ask you to rate a movie out of five stars); now things have gotten murkier as our behavior is tracked and collated in complex, opaque ways. Consumers have learned to adjust their actions to get the content they want, according to Nick Seaver, an anthropology professor at Tufts University and the author of Computing Taste: Algorithms and the Makers of Music Recommendation. 'You were much more in control of how you represented yourself under those [earlier] systems,' Seaver told me. Now our behavior—even the embarrassing kind—generates our unique media world."
— Atlantic

"Computing Taste is a valuable resource for scholars in music and anthropology at a time of increasing need of interdisciplinary dialogue between information technology and the humanities at large."
— Notes

TABLE OF CONTENTS


DOI: 10.7208/chicago/9780226822969.003.0009
[recommender systems;music recommendation;taste;start-ups]
This prologue sets the stage for the book to come, introducing the offices of Whisper, a pseudonymous music recommendation company. It introduces several themes: the relationship between corporate structure and software, the omnipresence of music listening in the workplace, and the variety of theories about taste held by the makers of music recommendation.


DOI: 10.7208/chicago/9780226822969.003.0001
[taste;algorithms;sociotechnical systems;ethnography;access;recommender systems;music recommendation]
This chapter introduces the central themes of the book, beginning with how people working in the music technology industry in the early 2010s understood the relationship between humanityand technology. Despite a popular discourse that pit humansagainst algorithms, these people understood themselves to be working in hybrid sociotechnical systems. This hybridity has consequences both for how we might understand taste in a world of technical mediators and for how we might gain access to andstudy algorithmic systems.


DOI: 10.7208/chicago/9780226822969.003.0002
[information overload;myth;cosmology;recommender systems;music recommendation]
This chapter offers a brief history of the origins of modern recommender systems in the mid-1990s, describing the dominant explanation for why these systems should exist: they are tools for mitigating information overload. The chapter questions the obviousness of this claim, analyzing stories of overload not as factual claims, but as myths. These myths provide a window intothe cosmology of algorithmic recommendation—the world in which their makers understand themselves to work.


DOI: 10.7208/chicago/9780226822969.003.0003
[recommender systems;metrics;trapping;persuasive design;captivation;music recommendation]
This chapter describes a shift in what it meansfor a recommender system to "work," following changes in dominant metrics of success. The developers of recommender systems have come to understand their goal as enticing users into enduring usage. The chapter argues that this shift emerged as part of a broader transformation in the business models and technical infrastructures of digital media services.Drawing on work in the anthropology of animal trapping, the chapter theorizes recommender systems as a kind of trap, with models of their prey embedded in their technical designs. This understanding of algorithmic recommendation challenges popular dichotomies between persuasive and coercive design.


DOI: 10.7208/chicago/9780226822969.003.0004
[recommender systems;context;music listeners;music recommendation]
Where the previous two chapters present visions of listeners as overloaded people in need of help or as prey to be captured, this chapter pursues another understanding of the listening subject: the "lean-back" listener, whose preferences are defined by context. For the makers of recommender systems, "context" is a capacious label that applies to many kinds of data, from location to weather to device configuration. Context-aware recommender systems acquire all sorts of data, figuring listeners as fragmentary and potentially incoherent subjects.


DOI: 10.7208/chicago/9780226822969.003.0005
[machine listening;interpretation;sound;music;quantification]
This chapter addresses a concern that has been surprisingly absent from most music recommenders: how music sounds. Machine listening techniques, used for bringing audio data in to algorithmic recommendation, treat musical sound as a kind of signal. This understanding of music is predicated on the idea that music and mathematics—and hearing and counting more generally—are fundamentally linked. Where humanistic critics often worry about the reductive effects of quantification, the chapter describes how the connection between numbers and sound also enables a set of interpretive listening practices that are central to algorithmic work.


DOI: 10.7208/chicago/9780226822969.003.0006
[spatialization;recommender systems;genre;machine learning;music recommendation]
This chapter describes the dominant understanding of musical variation among the makers of music recommendation: music can be analyzed as occupying a "music space," where it is distributed by similarity. Algorithmic recommendation is commonly imagined as occurring within spaces like these, where similar items are "neighbors" and plausible recommendations are nearby. These spaces are treated by machine learningpractitioners as intuitive models of difference, but also as uncanny computational objects that escape human spatial intuition. Understanding musical variation as essentially spatial changes the meaning of familiar terms like genre and preference, rendering them as clusters or regions to be computationally mapped and explored.


DOI: 10.7208/chicago/9780226822969.003.0007
[recommender systems;metaphor;machine learning;control;music recommendation]
Building on the discussion of the music space in the previous chapter, this chapter examines a set of spatial metaphors commonly used by the makers of music recommendation: pastoral metaphors, which figure technical workers as gardeners or park rangers who tend to the music space and the listeners who travel within it. Many critics argue that metaphors like these naturalize the work of machine learning, mystifying how it actually works and eliding the role of human labor. The chapter offers a different interpretation, suggesting that developers find pastoral metaphors useful because they describe an ambivalent and circumscribed form of control.Analyzing these metaphors helps us interpret how the makers of music recommendation think about their power and responsibility, in relation to the objects of their labor and to music more generally.


DOI: 10.7208/chicago/9780226822969.003.0008
[recommender systems;power;music recommendation]
This epilogue steps back from the argument of the book, presenting an interview conducted with one of the central figures in the field of music recommendation several years after the conclusion of ethnographic fieldwork and his departure from the industry. It reflects on the souring of the optimistic early visions of recommender systems, the increasing power granted to algorithmic recommendation, and the ethical status of these technologies' makers.

Book published by University of Chicago Press, 2022. Page maintained by BiblioVault.