picture of sin

Sina’s PKM

Catered to Your Future Self

[ Film analysis ]

Paper: Catered to Your Future Self: Netflix’s “Predictive Personalization” and the Mathematization of Taste
Author: Neta Alexander

Discussion on taste.

“It’s a matter of taste” did not meant it was entirely subjective.

Pierre Bourdieu: taste is not changing. Taste is classification of the classifier. Showing is the distinguisher is distinguished. It positions someone in relation to others.

Taste Machine: industry tries to shape taste in order to lift perception of something. This happened to cinema.

Enter netflix and the rise of individualism. Netflix's algorithm, Cinematch, is used to provide individual preferences. Choice is the new standard.

The myth of personalization

The golden role of this new economy can be found in the tautological slogan of a 2014 campaign for Motorola’s “Moto X” smartphone: Choose Choice . But what exactly do we choose, when we choose choice? Who—and what—is being excluded from this choice-based utopia?

Cinematch uses clustering and association rules, so naturally it's not a completely individual experience. Nobody wants to share their movie picks but is constantly being affected by other people's movie picks.

they confuse the “You” in “Recommended for You” with a unique, complex
individual rather than with a group of strangers who all happened to have made similar choices

Hired screenwriters to tag movies for ML. Over 76k microgenres. Netflix prize announced. Goal: Maximize time on platform.

Catering to your present self makes you live in a forever “now” (not in a good being present kind of way). Cinematch is no longer recommending, it's taking over.

But programmers aren't unethical, long complications (white papers, research, coding, testing, deploying, sharding, etc.) simply blur the ethical concerns.

This is because modularity manages complexity.

Endless choice is actually censorship in disguise.

All black boxes hide assumptions, and in Netflix's case that assumption is “remain in my comfort zone”.

The gut feeling is dead; long live the metadata.

We are no longer serendipitously exposed to unique films.

There is a discrepancy between the perceived “digital utopia” of endless choice and the reality of a shifting, tweaking, recommendation/taste-shaper algorithm.

You’re not chosing choice if your choice is limited to a few billion dollar enterprises.

Pages that link here:

taste
Various ideas on taste We see taste as meaning individual preference, but that's only a modern interpretation

recommendation takeover
[ Category - Essays ][ TODO - Expand ] Any widely spread enough recommendation algorithm, if used enough, will stop catering to the user's taste and start shaping their taste instead

modularity manages complexity
[ TODO - Expand ] The more modular a piece of code is, the better it can be managed

list of film analysis papers I've read
Paper Link Paywalled read online read online See also my paper reading template

Maximize time on platform
Goal of naive or evil organizations training ML to boost metrics

Current progress:

Loading... articles read 🤓

clear progress? 🚨