barthel:
Since taste is socially constructed, a far better indication of whether we’ll like a movie isn’t whether the movie has a happy or sad ending, but whether other people who like the movies we like also like that movie. That’s why Amazon’s “people who viewed this also viewed” recommendation system is still the best one yet devised. It doesn’t seem interested in why someone searched for two books on the same day, just that the connection was made. And because we’re probably looking for books in the same way that other person was looking for books, Amazon can predict our search before we make it. Taste is social—and social phenomena are far harder to break down than what Netflix is doing with movies.
I agree with this, Taste is social, existing between people and not in isolation. People don’t just like what they like in a vacuum; they like what they can talk to others about or what will make them seem smarter or cooler to others or whatever. But as Felix Salmon points out, Netflix has to work to convince users that they don’t want to watch the quality stuff people talk about but the second-rate stuff the company can license but users are less likely to recommend.
What Netflix accomplishes by trying to atomizing taste in the ways Madrigal details is to give users a different to game to play with their identity that is nonsocial, that consists of manipulating decontextualized toy blocks into effigies of their identity — catering to the dream (if anyone really wants this) of a nonsocial self that can be built and entertainingly toyed without the risk of others’ judgment. Algorithms stand in for the approving other in the circuit of self-production.














