What is Wrong With Book Discovery?
A few weeks ago I discovered a lovely article through Radish Reviews Linkspam listing 25 hard truths about the publishing industry. One of them was, as they put it, that book discovery is “wonky.” Well, they might have used more colorful language in addition, but I think we can all agree that there is some wonkiness. I wanted to take a moment today to start talking about the problem so that maybe we can start brainstorming solutions. After all, it’s the technological revolution, why can’t we get someone to find a way to recommend all the books we actually want to read??
Define “Book Discovery”
First, let’s get a good idea of the problem. Book discovery is the process of finding new books. Back in the old days (hehe) this was mostly dependent on word of mouth: “Oh, you liked that book, you have to try this one!” There are a fair number of sites that have tried to automate book discovery, such as Goodreads, Amazon and the beta of Bookish, but they just don’t always work great. I think a lot of us will agree that we still rely on word of mouth, but now we can get those recommendations through our blogger friends. Wouldn’t it be awesome if the recommendations for books worked as well as Pandora or Netflix? I mean, Pandora guesses pretty darn well what music I will like based on ONE song. And while sometimes I hit a bad movie with Netflix recommendations, most of them are what I was in the mood for. Why is it then, that when I enter in Daughter of Smoke and Bone into Bookish, I get suggestions that don’t even like vaguely similar from the description? Seriously, I was getting books that didn’t even look like fantasy! Not to mention that there are doubtless lots of indie books out there that I want to find, but I am going to have to get very lucky to stumble onto a review of them from a friend. Otherwise I’m not likely to hear of them, let alone from a trusted source that will lead me to actually try out the book.
It’s also worth pointing out that there are two different models of automated book discovery. Goodreads focuses on a full profile approach, where they make recommendations based on your entire shelf, but this makes it difficult for you to know if a book has particular aspects that you might be looking for if your shelves aren’t very specific (like me :( ). Bookish and Amazon have narrower approaches where they make specific recommendations based on one book. This is nice since you know if you are in the mood for that type of book, their recommendation might be useful. However it can also lead to spurious recommendations when they make a match based on one attribute that you actually don’t care about at all…. I think this might be causing my confusion on Bookish’s recommendations ;-). Which model do you prefer?
Problem 1: Ratings Material
Now, I’m just guessing here, but I’m willing to bet that one of the problems with making an algorithm for book discovery is that it takes a lot longer to read a book than listen to a song or watch a movie. It’s probably just difficult to get enough feedback from readers to categorize all the world’s books reliably on the categories that count (or even figure out which categories count!). While Goodreads gets lots of reviews on books with a fair amount of popularity, there are plenty of books without any reviews at all. While ratings are helpful, it seems likely that a review that can be mined for keywords is necessary to get good results in an algorithm. Thoughts? Movie and music discovery systems have also been setup to get feedback on how the algorithm did, with all the thumbs up and down and Netflix begging for ratings after you finish a movie. However, there isn’t the connection between book discovery recommendations and feedback on several of the sites. Amazon does ask for feedback, but I find that I often haven’t finished the book (or even started it) by the time they send an email. This might be due again to the amount of time it takes to read a book, or perhaps Goodreads just isn’t as focused on this element of book discovery?
Problem 2: Money
Netflix is a service that you pay for, so there is an obvious monetary benefit to them perfecting their recommendations engine. Pandora is a bit different, since (unless you upgrade) it’s free, however they want you to come back and listen to their ads, so again they want to be the best. While you could make the argument that Amazon wants to have the best recommendations so you buy books based on them, the monetary incentive from the publishing side seems to have corrupted book discovery on Amazon. Personally, it’s gotten to the point that I don’t even look at Amazon reviews because I have no way of knowing if that is an author boosting their positive reviews or bribing fans to post more reviews, given all the scandals we keep hearing about. This seems to be a problem of Amazon’s algorithm being too easily influenced and the publishing industry being too closely tied to book sales. This is obviously not changing anytime soon, but it is a hurdle that will have to be figured out when designing a recommendation algorithm.
So, now I want to hear from you. Do you use recommendation engines as a primary source of book discovery or do you just rely on suggestions from friends? Is there a particular site that you prefer? Do you think the problems that I’ve listed can be solved easily? Do you think that they are problems at all?
© 2013, Anya. All rights reserved.