Sunday, 11 October 2015

If Amazon Behaved Like A Friend

I must have bought over one hundred books from Amazon since its inception - but, alas, I don't know what is up with Amazon's book recommendation algorithm. Apart from the obvious low-hanging fruit it seems to be pretty mediocre at suggesting books for me based on past purchases.

When you're round someone's house their bookshelves are a great way to gauge information about them - you can get a fairly good idea of their education, interests, passions, tastes, beliefs and background - and more importantly here, you can use those observations to suggest books for them with consummate ease.

This leads me to believe either that Amazon's data mining is not as proficient as it could be, or that it's another good example of how there is, and will always be, a significant qualitative discontinuity between computers and the intuitively perceptive abilities of the human brain.

I had an idea a while ago about how interesting a book recommendation of *opposites* or *thematic alternatives* would be for, say, one day a week, where through data mining sellers don't suggest all the same kind of books, but deliberately suggested radically different ones to diversify your tastes and experiences.

So instead of saying, "Hey I see you bought Steven Pinker, Jared Diamond and Malcolm Gladwell, why not try Matt Ridley and Steven J Dubner and Oliver Sacks?" - they'd instead say "Hey I see you bought Jackie Collins, Ricky Gervais and Andy McNab, why not try Soren Kierkegaard, Charlotte Bronte and Thomas Aquinas?"
If Amazon behaved like a friend, and really wanted the best for you rather than simply trying to sell you more of what you've already bought, it would throw up a few of those opposites and thematic alternatives to help you keep your thinking fresh and diverse.