From Library Babel Fish, July 9, 2015
Joshua Kim honored me by paying attention when I was one of those annoying people tugging his elbow to say “hey, you have to read this book!” (The book in question is Data and Goliath by Bruce Schneier. By the way you, too, have to read this book.)
The question he raises about how we discover what to read is interesting and points out how books aren’t quite like other consumer goods. I don’t generally go around telling people “you have to wear these shoes” even if I am happy with a pair I bought. But when I really, really like a book, I am likely to mention it to people who I think will like it, too. Why is that? There are so many books out there, and it’s hard to know what to read next unless you have trusted sources. So far, it has been hard for algorithms to replicate the wisdom of a small in-real-life crowd or even for the wisdom of large crowds to be numerically calculated without tears. Let’s take Amazon as an example.
Amazon started selling books because they are fairly easy to source and ship, there’s a lot of them, there’s a substantial installed consumer base, they were able to partner with the book distributor Ingram to start with a lot of data about books, and the book industry uses an international system of unique identifying numbers that make inventory control easier than for, say, shoes. But Amazon always meant to be a major retailer of goods of all kinds. Their “review” system can be a signifier of how a lot of readers feel about a book or how divergent opinions might be, but it also can say “it arrived a week late and the package was damp. I’ll never buy from this seller again,” which is a review of a service, not a book.
Apart from the mixed messages, there have been lots of problems with the rating and reviewing of books on Amazon. Ten years ago, a glitch caused reviewers’ real names to appear on the Canadian site, revealing authors who praised their own books and made mean-spirited remarks about other writers’ books. Authors have complained of being bullied by mean reviewers and readers have complained of being bullied by demanding authors. Authors have had public meltdowns and stalked their critics online and off. Most recently, Amazon has incensed some reviewers by removing their comments, saying they are friends with the author. How they know this relationship exists is a trade secret, and both the lack of transparency and the appearance of surveillance has caused consternation. Goodreads, a book social network owned by Amazon, has gone through similar drama, and their sheer scale – Goodreads currently has over 40 million members who have listed 1.1 billion books and written 43 million reviews – mean that a change can affect a lot of people, all of whom feel some sense of connection to the platform.
There are at least a couple of issues entangled in this state of affairs. One is that it’s very hard to design a system that can’t be gamed somehow, particularly when the system is designed around gaining attention. This is one of the reasons Google keeps switching up its search algorithm. People whose job is SEO (search engine optimization) figure out a method to bump a site toward the top of search results and Google designs a work around. The second is that people who devote hours to creating content using proprietary platforms feel a sense of investment and ownership, so feel betrayed when the platform changes and their content is gone.
A recent opinion piece in The New York Times on the Reddit protests – volunteer moderators of popular threads were unhappy when an employee they liked was abruptly fired – prompted a lot of puzzled comments. Why should they have any input into a company’s personnel matters? And isn’t it just stupid to spend hours of time contributing content to a website without getting paid?
I think these responses miss the strange nature of the social web. Some things people do for love, not money. A platform needs active members. Members who are active are going to resist decisions they disagree with. And the platform itself promotes a sense of belonging that can morph into a community of resistance.
All of which is a way of saying that finding the next book to read is complicated. You can sign on with BookVibe (which harvests books mentioned by the people you follow on Twitter and tells you which books are trending). You can join Goodreads groups that discuss your favorite genres. You can try out LibraryThing, which offers automated recommendations based on subject headings and tags, member suggestions based on those with similar reading taste, and even “unsuggestions” – books that are most unlike the ones in your collection. You can check out the incredible number of book lists out there – Book Riot, for example, will keep you busy with lists about various topics or different genres and even a top ten list of top 100 lists – how meta. Or you can hang out, in person or online, with people who read the kind of book you like and compare notes.
At root, the best book recommendation algorithms are based on information provided by people who take the time to tug your elbow to say “you’ve got to read this book” online or in person. Sometimes they’re right. Or, of course, you could see if your local public librarians or favorite booksellers, who won’t sell your data to a third party, have ideas for you. They usually do.