Consumers can be suspicious of software second-guessing their tastes

Book interview: ‘Computing Taste’ by Nick Seaver

Image credit: Dreamstime

The gap between technology and culture might not be as wide as we think, says Nick Seaver in his analysis of how music recommender systems are produced.

When it comes to the relationship between technology and culture there’s an apparent contradiction in how taste – in the sense of our subjective human preferences – is perceived. Few would challenge that when it comes to fashion, art or music, what we choose to like all comes down to a matter of taste. In other words, it’s a matter of personal preference, no wrong answers, live and let live.

Move into the technology space and, as anthropologist Nick Seaver observes in his new book ‘Computing Taste’, we enter a different world. Humans rarely express taste preferences for the engineering that goes into jet engines, nuclear power stations or, for that matter, algorithms. And yet, algorithms are the cornerstone of how music recommender systems such as Spotify or Apple Music work. Back in the subjective world of the arts, consumers are suspicious of software that tells them what they should like. This is because when strings of binary maths codify, anticipate and to all outward appearances dictate what is our next tune, we feel insulted by machines.

As an anthropologist, Seaver’s job is to study human behaviour, and while ‘Computing Taste’ might focus fundamentally on the people and algorithms behind music recommender systems, it’s also more far-reaching in its scope. He’s examining the contradiction outlined earlier: that of the paradox inherent in the coexistence of technology and culture. “My central concern is to do with the broad common sense that taste is this ineffably human thing that can’t be accounted for – ‘there’s no accounting for taste’ and so on – while technology is primarily concerned with efficiency and rational design.” He explains that this starting position assumes that there is a “fundamental mismatch between these two domains”, while there is an industry whose sole purpose is to establish a meaningful bridge between the two.

Seaver admits that as an anthropologist he comes to the subject from the position of “trying to defend culture from people trying to do too much engineering to it”. This is added to his fascination with the idea that there’s a software industry that knows only too well “that people think that what they’re doing is impossible. The animating question behind my book is how to deal with that. How do you go about building a machine that is supposed to predict people’s taste, despite the fact that everybody knows that it shouldn’t work? Does working in this space change what we think about culture? Does it change what we think about technology? How do people deal with this paradoxical question?”

We read it for you

Computing Taste

It’s hard to fault their intentions because music recommender systems developers want to improve your listening experience. They want to broaden your horizons and help obscure musicians find audiences. But many consumers are critical, suspecting them of flattening culture into numbers and profiling users for commercial purposes. In ‘Computing Taste’, Nick Seaver describes how algorithm creators navigate this by delving into how recommenders understand their relationship with listeners, how scientists conceive of listening 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. You’ll come away from ‘Computing Taste’ realising that algorithms aren’t the enemy, ready to think again.

Put in its most straightforward terms, music recommender systems decide what we want to listen to next. But this is only one domain in which the technology is applied. While Seaver’s research was done in the field of music, recommender systems are used “all over the place” – from fintech to fashion – while related machine-learning technologies “are used even more broadly than that”. He accepts that such systems are regarded by the public as controversial, “but I would argue that their use in music is one of the less controversial applications. This is because personalisation in social media or the news is often associated with filter bubbles, or politically charged, whereas in my experience consumers tend to think of recommender systems in music as fairly innocuous. They might think of them as useless or insulting, but I don’t often hear them talked about as a problem in their own right.”

Seaver, who is assistant professor of anthropology at Tufts University and co-editor of ‘Towards an Anthropology of Data’, says that his background in media studies means that he’s always had an interest in the overlap between technology and music. One of his earlier projects was the history of the player piano, a self-playing instrument that uses pneumatic or electromechanical mechanisms to do the job of the human musician. “I came into this as a historian of historic automatic instruments, and I became interested in music because it is a domain that people think of as being very human and expressive. But it is extremely technological: there are instruments, recording apparatus, streaming services and so on. It’s seen as simultaneously a universal language that is shot through with all of this tech.” Which makes it the perfect case study for “how humanness and technology come together”. ‘Computing Taste’ is the result of a decade’s research on just that.

If you’re interested in algorithms, personalisation and recommender systems, says Seaver, music is a useful area of study “because the people working in that space are experimental, doing weird things and happy to talk about it. It doesn’t have the moral charge of studying, say, predictive policing, which is a domain that uses sometimes similar technology, but has much more serious outcomes. That’s a hard place to study because everyone involved knows that they’re on thin ice. Music isn’t a life or death situation, and so from a practical point of view it’s easier to get people to talk to you.”

Seaver accepts that there will be listeners whose perception of themselves is founded on their taste, who will find it beneath their dignity to allow this self-​image to be manipulated by black-box algorithms. While he accepts that there is an argument for seeing such consumers as luddites, Seaver asserts that “my job is to represent the people on ‘the other side’: the thoughts of the people building these things”. Crucially, and one of the big revelations of ‘Computing Taste’, these software developers “tend not to think of themselves as telling people what they should be listening to. They talk about what they do in terms of making suggestions and helping listeners to find things that they did not know about before.”

It may seem strange that an anthropologist by trade would be so sympathetic to the position of the algorithm nerds. But, as Seaver demonstrates over 200 pages of immaculately reasoned and (for an academic treatise) highly readable and (sometimes) even gently ironic book, these are the people creating access to a brave new world of music taste. There may be no accounting for taste, he says, but it may be less subjective than we once thought.

‘Computing Taste: Algorithms and the Makers of Music Recommendation’ by Nick Seaver is from the University of Chicago Press, £16


Taste and technology

Popular understandings of taste are neatly complementary with technology: if technology is the domain of necessary decisions, then taste is a human domain of arbitrary and subjective judgment. We can have taste in music but not in the mechanics of jet engines. Here, too, social scientists like to reverse things, emphasising the flexibility of engineering and the determinedness of taste. Technologies are shaped by much more than the dogged pursuit of efficiency, and our tastes are actually shaped by forces beyond our own control.

Most social scientists attribute this to Pierre Bourdieu, the sociologist whose book ‘Distinction’ provides the field’s common sense about what taste is and how it works. In brief: tastes are not caused by their objects but consequences of the social order. Fancy people like fancy things, common people like common things, and preferences are a function of status, which serve to reinforce social hierarchy. This ‘homology thesis’ is not the only argument Bourdieu makes, but for many social scientists, the homology between taste and status is Bourdieu’s theory of taste. If we use this theory to build a recommender system, we might end up with an unsettling design: assess users’ social positions and give them the cultural artifacts that correspond to their class fraction.

Critics of this theory suggest that it takes taste to be “an arbitrary election which has to be explained... by hidden social causes”, leaving an explanatory hole: how are tastes acquired in the first place?

Music offers a useful domain in which to examine these concerns because it embodies the paradoxical relationship between culture and technology from which many concerns about algorithms emerge.

Edited extract from ‘Computing Taste’ by Nick Seaver, reproduced with permission.

Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.

Recent articles