Tag Archives: prediction

Prediction Algorithms: You Might Also Like…

Amazon is a fairly poor predictor of what I might like to read next. For some reason, their algorithms just don’t work well on me. I am trying to think of a time when Amazon suggested a book, and I thought, Yes, that is exactly what I need.

I’m thinking about this today for two reasons: first, because I’m in one of those in-between states, where I can’t quite figure out what to read next; and second, because of an Amazon email that’s been sitting in my inbox since yesterday with a subject: “Discovery your next read.”

The email was well-timed, what with me at sea between books, so of course I took a look at it. The list offered ten possibilities broken into five groups. They are as follows:

Recommended for you

  • Bird by Bird: Some Instructions on Writing and Life by Anne Lamott. I suspect this is because I recently read The Destiny Thief: Essays on Writing, Writers and Life by Richard Russo. Okay, maybe this is a fair recommendation, but there is a little luck involved here, as I will explain shortly.
  • The Pioneers by David McCullough. This would be a great recommendation, right up my alley, if not for the fact that I have already read it.

Based on your reading

  • Bridge of Sighs by Richard Russo. This is a better recommendation than the Anne Lamott because I just finished reading Empire Falls. See, I went from reading about Russo to wanting to read his writing, not more about writing. That’s the problem with the Lamott recommendation–well, one of them, anyway.
  • Plain Text: The Poetics of Computation by Dennis Tenen. I suspect this is because I am partway through a fantastic book called Track Changes by Matthew Kirschenbaum, a history of the word processor. The only problem is, I’ve stalled on that book, not because it is bad–it is fantastic. But I am looking for something else at the moment. That makes Plain Text interesting, but not right for the moment.

Inspired by your wishlist

  • The Joy of x: A Guided Tour of Math from One to Infinity by Steven Strogatz. It does sound interesting, but no, not now.
  • Sync: How Order Emerges from Chaos in the Universe, Nature, and Daily Life by Steven Strogatz. Order certainly emerged from chaos here, where I got not one but two books by Strogatz, someone I’ve never heard of. What is on my wishlist that inspired these recommendations?

For you in biographies and memoirs

  • Ten Innings at Wrigley by Kevin Cook. Okay, I have this audiobook, and it is downloaded to my phone, which is what I do for books that I plan on reading in the near future. This is a good prediction and recommendation, and I’ll give Amazon credit for this one.
  • For the Good of the Game by Bud Selig. Interesting that both recommendations are related to baseball. This would be a good recommendation as well, yet once again, I have already read this book. Indeed, I enjoyed it so much I sent Selig a note, and even got a response from him!–Although it is possible the response I got was a form letter.

For you on Amazon charts

  • Where the Crawdad Sings by Delia Owens. I’m sure it’s a great book, but it’s not up there on my radar anywhere.
  • The New Girl by Daniel Silva. Not sure why Amazon would recommend the 19th book in a series when I haven’t read the first 18.

Okay, so if I were generous, I’d say that generally speaking, Amazon made 3 good recommendations: The Pioneers, Ten Innings at Wrigley, and For the Good of the Game. The problem is that I have already read two of those, so in practice, Amazon made only one good recommendation. One is better than none, I suppose, but it doesn’t encourage me to take their recommendations seriously.

There are two problems, as I see it:

First, Amazon doesn’t seem to know which books I have read and which I haven’t. I mark books “Finished” on Goodreads, which Amazon owns, so they have access to that data, and could, in theory, use that to eliminate recommendations and replace them with others. Moreover, I have finished 365 audiobooks on Audible, which Amazon also owns, and from which, they should be able to tell what I have finished and what I haven’t. That seems like a simple problem to fix.

The second problem is more complicated. Predictions work better, I suspect, for readers who read primarily within a set genre or two. But what of an eclectic reader, someone who reads, say, a classic collection of sportswriter interviews, and follows that up with a Hollywood memoir, after which he reads a book about NASA engineers, and then just for kicks, a book on the White House chiefs of staff. I suspect Amazon’s algorithms are good at saying, “If you liked the Kingkiller Chronicles, then you should try…” But how good are they at making recommendations for someone like me, whose whimsy is often guided by the butterfly effect of reading? Given that series of four books I j just listed, what direction does an algorithm take?

I empathize with Amazon’s prediction bots at times like these, when I am floating on an ocean with no interesting books in sight. Today, just to read something I started The Great American Sports Page: A Century of Classic Columns from Ring Lardner to Sally Jenkins. Maybe this one will take.