Editor’s Note: Nick Seaver (@npseaver) kicks off the March-April special edition of Ethnography Matters, which will feature a number of researchers at the Intel Science and Technology Center for Social Computing on the forefront of exploring the cultures of hackers, makers, and engineers.
Nick’s post makes the case for the importance of “studying up“: doing ethnographies not only of disempowered groups, but of groups who wield power in society, perhaps even more than the ethnographers themselves.
Nick’s own research explores how people imagine and negotiate the relationship between cultural and technical domains, particularly in the organization, reproduction, and dissemination of sonic materials. His current project focuses on the development of algorithmic music recommendation systems. Nick is a PhD candidate in sociocultural anthropology at UC Irvine. Before coming to UCI, Nick researched the history of the player piano at MIT.
When people in the tech industry hear “ethnography,” they tend to think “user research.” Whether we’re talking about broad, multinational explorations or narrowly targeted interviews, ethnography has proven to be a fantastic way to bring outside voices in to the making of technology. As a growing collection of writing on Ethnography Matters attests, ethnography can help us better understand how technology fits into people’s everyday lives, how “users” turn technologies to unexpected ends, and how across the world, technologies get taken up or rejected in a diverse range of cultural contexts. Ethnography takes “users” and shows how they are people — creative, cultural, and contextual, rarely fitting into the small boxes that the term “user” provides for them.
But ethnography doesn’t have to be limited to “users.”
My ethnographic research is focused on the developers of technologies — specifically, people who design and build systems for music recommendation. These systems, like Pandora, Spotify, Songza, or Beats Music, suggest listening material to users, drawing on a mix of data sources, algorithms, and human curation. The people who build them are the typical audience for ethnographic user studies: they’re producing technology that works in an explicitly cultural domain, trying to model and profile a diverse range of users. But for the engineers, product managers, and researchers I work with, ethnography takes a backseat to other ways of knowing people: data mining, machine learning, and personal experience as a music listener are far more common sources of information.
Ethnographers with an interest in big data have worked hard to define what they do in relation to these other methods. Ethnography, they argue, provides thick, specific, contextualized understanding, which can complement and sometimes correct the findings of the more quantitative, formalized methods that dominate in tech companies. However, our understandings of what big data researchers actually do tend to lack the specificity and thickness we bring to our descriptions of users.
Just as ethnography is an excellent tool for showing how “users” are more complicated than one might have thought, it is also useful for understanding the processes through which technologies get built. By turning an ethnographic eye to the designers of technology — to their social and cultural lives, and even to their understandings of users — we can get a more nuanced picture of what goes on under the labels “big data” or “algorithms.” For outsiders interested in the cultural ramifications of technologies like recommender systems, this perspective is crucial for making informed critiques. For developers themselves, being the subject of ethnographic research provides a unique opportunity for reflection and self-evaluation.
Starbucks Listeners and Savants
Among music tech companies, it is very common to think about users in terms of how avidly they consume music. Here is one popular typology, as printed in David Jennings’ book Net, Blogs, and Rock ‘n’ Roll:
At the top are the “Savants” — very enthusiastic listeners who are willing to expend a lot of effort to find new music and, significantly for these companies, to spend a lot of money on the music they like. At the bottom are the “Indifferents,” who are the opposite: they do not particularly care about music, and they don’t want to spend much effort to find it. They are sometimes referred to as “Starbucks listeners” — these are the folks who would be perfectly happy picking up whatever compilation CD happened to be for sale at the Starbucks counter. For the developers of music recommender systems, these different attitudes toward listening pose a problem: most of the people who work at music tech companies are enthusiastic “savants,” while the bulk of the market is made up of “casuals” and “indifferents” who approach music in a very different way. There is a risk that savants will build systems that suit their own interests and ideas about music while neglecting those of the market at large. What designs would better suit less avid listeners?
By now, ethnographers are likely chomping at the bit: this is a textbook, perfect occasion for ethnography, not only to help understand what listeners beyond the company walls are like, but also to complicate this coarse typology. Instead of organizing listeners according to a single metric, ethnography could offer more insight into what the experience of listening to music is like beyond simple variables like “How much will I pay?” or “How many clicks until I get the music?” For designers, these insights could suggest more diverse and interesting directions for development.
But while this complicates our ideas about mainstream listeners, it ignores the details of the developers who supposedly build systems that fit their own preconceptions. If, at the end of the day, developers are different from the average user (whether or not they deserve the flattering title “savant”), then we really want to know what they are like: What are their preconceptions about culture? What experiences do they bring to bear on the systems they build? What auxiliary motives guide their technical decision-making? For outside critics and inside developers alike, the ethnography of technologists provides a useful counterpart to the ethnography of users. Perhaps most importantly, it helps us break down the distinction between these two groups: with the ethnographic gaze turned the other way, we can see the similarities and interrelations between the people commonly thought of as consumers and the people commonly thought of as producers of technology.
Studying up, revisited
In anthropology, this kind of ethnography is known as “studying up,” from an influential 1972 essay written by Laura Nader: “Up the Anthropologist — Perspectives Gained from Studying Up.” For most of its history, ethnography had been used to “study down” — people with power studied those without it. Anthropology was organized around the study of the “savage slot,” examining the lifeworlds of so-called primitive, small-scale, or savage society. These were people disempowered by colonialism, and anthropologists themselves often worked for colonial powers, using their research on local customs to assist colonial governments. Sociologists taking up ethnography also tended to study down, examining the lives of the urban poor or working class.
In its early history, the ethnographic method was built on this imbalanced power dynamic, which made it hard for the subjects of ethnography to refuse to be studied, and posed serious ethical problems regarding how ethnographers should use their findings. The unseemly parallel in the present moment is between “savages” and “users”: the term “user” designates a class of people disempowered in relation to technology, and treating people as users disempowers them.
To continue sketching out this very broad-stroke history: anthropologists had a crisis of conscience in the mid-twentieth century regarding their complicity with colonial projects, and they have spent much of their time since then trying to come to terms with the unsavory legacy of the ethnographic method. Today, many anthropologists use the rich knowledge and close friendships they develop during fieldwork to serve as advocates for their research participants; this anthropology is a far cry from exploitative studies of the past, such as research on social cohesion among tribal groups commissioned by colonial governments looking to break it down. The parallel in the present moment is with user researchers whose interest is not simply to package users up for technologists, but to complicate the idea of what “users” can be and to advocate on their behalf.
What Nader suggested in her essay was that anthropologists should also point their ethnographic eye up, to people who wielded power — corporations, the government, the wealthy, scientists, the police, white collar criminals, and so on. These studies would contribute to a more well-rounded anthropology and provide instructive challenges for ethnographic methods. The studier-up faces research problems that may not arise while studying down, like acquiring access to corporate offices or having to convince lawyers of the merits of your research project. While studying up, you realize that many standard elements of fieldwork seem strange when applied “out of place.”
For many researchers in science and technology studies, ethnographic studying up is their defining method. In the 1970s and 1980s, these researchers brought ethnography “home from the tropics” and studied in the labs of high energy physicists, plant biologists, and biochemists, producing a set of creative and perplexing ethnographies. Since these pioneering laboratory studies, STS-affiliated ethnographers have studied the development of technologies in corporations and academic research labs, drawing attention to under-appreciated aspects of science and technology: the influences of “cultural,” “social,” or “political” factors in settings often assumed to be purely rational, the sociocultural structures within high tech organizations, and the subjective dynamics that shape technical decision making.
Of course studying up is not a replacement for studying down, any more than ethnography is a replacement for all other methods of knowing people. Rather, as Laura Nader suggested, studying up adds a usefully contextualizing perspective to other forms of ethnography. As ethnographers examine how people live with algorithms, studies of those algorithms’ designers help complete our picture of the contemporary world, in which software sorting allows social theories to be ever more explicitly built into cultural infrastructures.
How does ethnography matter?
I went into this project thinking that my interlocutors and I approached the world from very different perspectives: I’m a humanities-trained guy, getting a PhD in the “most humanistic of the sciences,” while the people I work with in the field are typically engineers with advanced degrees in computer science. One of the most surprising findings from my research so far is how much disciplinary history we have in common.
The ethnographic method did not fall fully-formed from the sky, ready to use. It has a history, over which it has changed substantially as researchers have experimented with new practices and encountered new puzzles. Over the course of the twentieth century, these experiments often involved formal and quantitative methods (as in the “New Ethnography” of the 1970s). As I explore contemporary methods for understanding music and culture with computers, I am also looking to the history of ethnographic and anthropological research, and the parallels are striking. Multidimensional scaling — an analytical method extended by researchers in mathematical anthropology — is a common technique used at larger scale by academic researchers in music recommendation. Alan Lomax’s Cantometrics project, which sought to catalog and taxonomize music practices from around the world, looks a lot like Pandora Radio’s Music Genome Project avant la lettre.
As today, these methods were the subject of heated arguments about how best to capture cultural life: Clifford Geertz’s famous and influential essay “Thick Description” is substantially concerned with rebutting the New Ethnography (sometimes known as “ethnoscience”). Forty years earlier, Bronislaw Malinowski — the paragon of immersive anthropological ethnography — made a similar argument for “full-blooded description” against the practice of formally diagramming kin relations. Since its early days, our understandings of ethnography have been tangled up with our understandings of formal and quantitative methods.
If ethnographers have sometimes seemed like techies, as Tricia Wang wrote recently, “hard-core techies also look like ethnographers”: they pull together a diverse range of analytical resources, assemble their numbers in interpretive ways, and frequently draw qualitative conclusions from quantitative practices. The tools of ethnography can help us to appreciate the thicknesses and specificities of these other ways of knowing people, enabling more robust critique and facilitating cooperation. A better understanding of how computational methods work in practice can give us a better understanding of ethnography and how it matters.