Let us tell you what you like

In 1996, in a remarkably prescient letter to Jorge Luis Borges, penned a decade after his death, Susan Sontag hypothesised a possible future of reading: ‘Soon, we are told, we will call up on ‘bookscreens’ any ‘text’ on demand, and will be able to change its appearance, ask questions of it, ‘interact’ with it. When books become “texts” that we “interact” with according to criteria of utility, the written word will have become simply another aspect of our advertising-driven televisual reality.’ Dystopic, perhaps, but not without truth. Recently, this idea of interaction with books in accordance with a ‘criteria of utility’ has been accelerated by the influence of digital preference algorithms and recommendations.

Last October, just in time for the Christmas shopping season, Penguin Random House unveiled its new online book recommendation platform, Flipper. ‘Simple and fun’, the tool allowed shoppers to avoid spending hours in bookshops, tentatively fingering hardbacks that would most likely end up in their family member’s garage by New Year’s Day. Now, you can simply enter a few details (‘I’m shopping for <an adult> who <likes to laugh>’), and ‘flip for the perfect book’ (in this case, The Ladybird Book of the Zombie Apocalypse). In exchange, Penguin generates a wealth of data that the publisher can use in the future.

Penguin’s program is the latest experiment in the publishing world’s foray into data-driven recommendation systems, or ‘taste engines’. Albert Hogan, the brains behind Flipper, was brutally frank about Penguin’s approach: ‘the days of talking to customers and readers and telling them what they’re reading, that’s gone. It’s more about surfacing ideas and trends and responding to those’.

Cutting, but who could blame him? The allure of preference algorithms has inundated all corners of cultural production, from television to furniture. Literature has not remained untouched. A site such as Goodreads (purchased by Amazon in 2013) analyses ‘20 billion data points to give suggestions tailored to your literary tastes’.

Caught in Amazon’s vice, which over the last decade has triumphed the use of unconstrained algorithms to create recommendations based on a user’s history and ratings, publishing houses have increasingly started to implement similar systems into their marketing strategies. Barnes & Noble, Simon & Schuster and HarperCollins all employ variations of recommendations based on preferences and user history. Your reading practices are predicted by your digital footprint.

The benefits of such an approach are tangible: authors are exposed to a much wider audience, readers can navigate the deluge of choice, and publishers can respond more accurately to discernible trends. But the downside is as equally obvious. Turning literature into a numbers game may make sense from a commercial perspective, but it runs the risk of defining reading, and taste, in a very limited and stratified way.

The implications of recommendation systems on readers’ ‘tastes’ reflects a broader shift in conceiving of how our tastes function on a social level. We have long since moved past Kant’s categorical aesthetics, or Hume’s oligarchy of ‘true critics’. But the notion of taste still serves an important role in defining social identity. Bourdieu’s Distinction has been equivocal in how we think about our preferences. If we accept his gnomic that ‘social subjects, classified by their classifications, distinguish themselves by the distinctions they make’, then the choices that we make in our reading habits should be understood as part of a communicative exchange with our social network. We understand our preferences through conversing, agreeing, and disagreeing, with friends, peer groups and critics.

This is undermined by the idea of ‘personal preferences’. Instead of ‘taste’, it is fashionable now to talk about ‘discovery’, ‘curation’ and ‘personalisation’. The Book-of-the-Month Club, launched in 1926, initially claimed that any title that could survive the ‘differing tastes’ and ‘judgment’ of its panellists was ‘bound to be an outstanding book’. Now, it promises a selection of books that ‘we know you’ll want to read’. The change is subtle, yet demonstrative: our reading practices are fashioned in a solipsistic bubble, with the influence of social interaction relegated backstage. We allow our digital shadow to make our aesthetic choices.

The influence of user history and preferences on literature runs deeper than first appears: data collection affects not only the choices we make in reading, but the content of what we actually read. Take Jellybooks, for example, a reader analytics company based in London that works with publishing houses. Like screen-testing movies, the company gives free e-books embedded with JavaScript to subscribers, which bounces back real-time information: how long people read for and at what time, how far they got, or why they stumbled for so long over chapter six.

Publishers can then tweak the manuscripts, or marketing strategy, accordingly. Some books have been scrapped completely, and others pushed forward to the front of the cue. Such information has suggested to some publishers the possibility of determining ‘bestsellers’ on a purely algorithmic level. La mort de l’auteur indeed, but by a thousand cuts.

But there is a brighter side: several programs have combined recommendation systems with a larger social context. Alexi, a subscription-based app supported by Pan McMillan, Europa and Bloomsbury, gives members access to the personal recommendations of writers such as Peter Carey, Ali Smith and Alice Sebold. Co-founder Andrew Kidd has stressed that ‘people are fixed in their own echo chamber and hearing their own opinion’. The app aims to help people break out of their own perpetual feedback loop promulgated by algorithm recommendation systems.

Reading is an inherently social practice, and communities are consistently involved in a dynamic process of aesthetic redefinition. Whilst the future of taste engines is promising, the current correlation between personal preferences and Sontag’s ‘criteria for utility’ is too close for comfort. Such technology makes for a tightly regulated market, with literature relegated to variations of bestselling formulas. But most portentously, it leaves us bridled by our past judgments, cemented in our digital history.


Image: Fractal / pixabay

This is part of a series responding to our recent Pitch Page query about reading strategies in an age of 130,000,000+ books, a topic that received an unusual amount of interest.

Read the other responses in the series, ‘Infinite book lists and other loathsome behaviours’ and, ‘Against literary evangelicals’.

Nicolas Liney

Nicolas Liney is the editorial intern at the International New York Times in London and graduate student in classics at the University of Oxford. His work has appeared in The Local, Brag Magazine and ARNA.

More by Nicolas Liney ›

Overland is a not-for-profit magazine with a proud history of supporting writers, and publishing ideas and voices often excluded from other places.

If you like this piece, or support Overland’s work in general, please subscribe or donate.

Related articles & Essays