In the legacy publishing world, an oligopoly of gatekeepers decided what books would be available. Publishers chose which authors deserved attention. Reviewers, librarians and bookstores winnowed the field further. (If you could get Oprah Winfrey to recommend a book, its future was golden.) The system assured a certain level of quality at the top of the ladder. But discovery, apart from recommendations from friends and colleagues, was largely a top-down method.
Reputation was integral to that system. Publishers put their own reputations on the line by choosing their authors. Similarly, we learned to trust reviewers and their organizations, or not. And when our local bookstore owner recommended a book we hated, we were much less likely to take his word in the future.
The digital revolution hasn’t done away with the top-down recommendation model, even though news organizations have dumped book reviews, traditional bookstores are disappearing and the big publishing companies focus as much as possible on books they already know will sell. The most important recommender today may be Amazon*, which makes some corporate editorial judgments but mostly suggests books based on what “people like you” buy according to complex and proprietary algorithms.
Those highly customized online recommendations, in a variety of media formats such as video (Netflix) and audio (Spotify), suffer from their own imprecision. Sometimes the results are utterly laughable. They can often be amazingly right. They are based on deep dives into data, and over time the recommendations become more refined as we use them. But they rely much more on correlation than reputation.
In a system where readers’ choices are part of the formula, their own reputations can and should carry more weight. Some of those readers are our social media contacts. Others are bloggers whose work we’ve come to admire. They are part of an edge-in rather than top-down recommendation engine where readers make more or less explicit choices about who to trust. This is how I find much of the news I read (listen to/watch/etc.), but much less so when it comes to books.
That will change in coming years as we combine human and machine intelligence in more sophisticated ways. Here’s an extremely simple example: Suppose I could designate three people whose work I trust in a specific arena to tell me what they’re reading – as well as any three people each of them recommends in that arena. That would aggregate expertise and recommendations in ways I can’t easily do today. Someone will build a big business by creating better reputation-based tools for discovery.
How can we avoid finding out mostly (or only) about books we’re predisposed to liking, and thereby missing out on books we didn’t know we’d enjoy? I worry about the fact that Amazon tailors recommendations based on what it thinks I want. One of the joys of traditional bookstores is serendipity: the discovery of a nearby volume that I browse through and then decide to buy. This isn’t entirely random; the bookstore manager decided what books to put on the shelves, and a clever jacket design can entice me to check out a book I wouldn’t otherwise notice.
At some level we’ll need to create our own serendipity in the e-book era. This won’t be difficult, but we’ll probably need to do it more consciously, by going outside our zones of comfort and the recommendations of people we trust. Discovery can’t be a passive act.