🔗Jeff C. on Mastodon:

Amazon: “I see that you’re shopping for a rice cooker!

Would you like to buy this rice cooker along with a rice cooker and another rice cooker?“

I am 100% convinced that, nearly two decades into the age of algorithmic recommendations, we really have not progressed past “Other people who liked this liked that” and “Here’s a bunch more of what it seems like you like,” both implemented in the dumbest possible ways.

This sort of thing is super-frustrating to me as a user, because what I would like out of these sort of recommendation systems is stuff that I would not have thought of myself.

But giving me what I want for myself is not the point of most of the platforms and products that have implemented algorithmic recommendations. The point is to keep me on the site, put a lot of advertising in front of me, and gather as much data as possible about me in the process.

I am generally pretty skeptical that it is even possible for recommendation algorithms to really go beyond the “People who like this also liked that” model. Even if it were possible, though, I just don’t think any of these platforms and companies have any incentive to make them any better. It’s in their interest that we all think they are magical and based on super-secret technology that no one could possibly understand, but in the end, as long as we stay on their site and keep clicking on stuff, that’s all that matters to them.

Originally published at https://petebrown.gibber.blog/algorithmic-recommendations-are-shit.