Book Recommendations 2.0?


Last week, the social reading website GoodReads.com announced it had acquired Discovereads.com to bolster its algorithmic book recommendation technology. The Discovereads recommendation service, which has some origins at Stanford, will draw upon the data of the GoodReads community’s database of 100 million book ratings from 4.6 million users. As Otis Chandler (GoodReads Founder & CEO) notes at the end of his announcement post, better book recommendations also mean better targeted book advertisements, and thus sales and profits. I’m curious about how Discovereads works and will have to check out how good its recommendation engine is, especially with my all-over-the-place taste in books.

The Friendship Algorithm


The New Yorker Book Bench blog  (“Algorithms: Better Than a Buddy?“) ruminates upon “the business of determining taste” and what the future of book recommendations might look like. While there are similarities to what Netflix and Zappos is doing in terms of recommendations based upon preferences and previous purchase history, it seems to me that the comparison with book recommendations can only extend so far (“Netflix is like a rather dimwitted but well-meaning robot-friend with whom it’s amusing to waste a little time”).

One of the main virtues of GoodReads.com is the community it offers, which is inherently different than the upselling that a Netflix or Amazon.com does with its recommendation engines. The New York Times Bits blog draws a similar distinction (“Need Advice on What to Read? Ask the Internet“):

For books, Amazon.com already has a robust recommendation system. But Mr. Chandler said Goodreads’s recommendations will be better because Amazon considers books a customer has browsed or bought, so buying a gift for a child could throw off the recommendations, for instance.”

Granted, the user population that rates books read within GoodReads is somewhat self-selecting, whereas Amazon is, for better or for worse, taking a more all-encompassing approach with how it recommends books and other products.*

I’m all for better book recommendations. There are lots of things I find appealing about an online book community, but is it the same thing as the shared social experience of having read the same book with someone else? At least it seems pretty clear which side of the ‘friends vs. algorithm’ discussion Book Bench falls on:

Whatever algorithm God put inside these two people is the right algorithm for me. Otherwise, though, I have to engage in a little pragmatic chaos: I have to listen to the opinions of a few buddies and a few good reviewers, and sort of wait to get wind of the general opinion (it’s mystical), and then I can decide whether to jump. And it’s a great system! It’s unbeatable, even by a clever machine.”

Could we be losing something in the quest for more and more efficiency? Can there be a formula for our taste in books? Our reading tastes change over time because, after all, we change over time. Do I always like every single book recommendation I get from friends? No. But I’d be sad to see those sort of imperfections disappear anytime soon. Evenact of communications such as a book recommendation — good or bad recommendations, observed or ignored — are always instructive in learning more about another person.

In that spirit, here’s a great new blog (Tandem Reading) from J. Raimo, devoted to friends reading books with other friends.

* This isn’t actually related, but it’s still funny. From McSweeney’s: “Amazon.com’s Recommendation Algorithm Applied to Life Events

Drop Me a Line, Let Me Know What You Think

© 2023 by Train of Thoughts. Proudly created with Wix.com