Ranking systems almost never expose the "knobs" to users.

· Bits and Bobs 6/16/25
  • Ranking systems almost never expose the "knobs" to users.
    • There are a number of reasons for this.
    • First, there's the tyranny of the marginal user–if only 1% of users would have twiddled the knobs, then they're not worth building because it might distract and annoy everyone else.
    • Second, they lock in whatever ranking ontology happens to be used right now.
      • That makes it harder to change internal ranking signals in ways that don't fit within that exposed ontology.
      • Ranking systems often have fluid experimentation throughout all components that are hard to fit in a user-facing ontology.
      • If you change the ontology, how do you migrate users' previous settings?
    • Third, you might reasonably think, "the user might set things now based on current low quality in some sub-domain, but if we improve that sub-domain, the user might have inadvertently boxed themselves out of a better experience."
    • Fourth, exposing too many knobs might give insight into internal signals, making them easier for creators to game.
    • But finally and most importantly, consumer systems have an emergent gravitational pull towards engagement maxing.
    • A ranking system would never ship a toggle to "make me more bored."
    • This is probably the logic behind OpenAI refusing to make the memory feature user controllable.
    • Also, they think they can get away with it without too much uproar…
    • … and they're probably right.

More on this topic

From other episodes