Theme 11 of 13 · 9 ideas

The political economy. A filter that follows rules can be tuned, tuning has an owner, and the owner has a business model. This theme covers how that plays out, and a map of the ways a filter can be turned against its purpose.

The same ground, told as a story: Chapter 10: Political Economy of Attention, Capture Taxonomy.

The ideas

Each one is a short page. Walk them in this order, or start with whichever catches your eye.

  1. Engagement-maximization as a captured equilibrium (mechanism): A feed that sells your time keeps showing whatever holds you longest, and nobody has to intend any harm for that to happen.
  2. The Orwell/Huxley axis: A society can lose the truth by having it banned, or by having it buried under endless entertainment, and the second is harder to fight.
  3. Cost-shifting from producers to consumers: When publishing became free, the work of checking what is true did not vanish; it moved onto readers, who have no extra time for it.
  4. Creation cheap / review expensive: Making content has become far quicker than checking it, so the unchecked pile grows faster than anyone can sort it.
  5. Out-competition of institutional carriers in attention markets: Careful journalism and scholarship cost money to make while feed content costs platforms nothing, so depth loses the audience without anyone ever attacking it.
  6. Self-capture vs. external capture: An institution can be bent by outsiders gaming its rules or by its own way of making money, and the second is harder to undo.
  7. Capture taxonomy (substrate × source × target asymmetry): Anything that filters what people see or learn can be turned against its purpose; the questions are which filter, by whom, and how fixable.
  8. Consumer-key vs. surface capture: Rigging what people are taught does deeper, longer-lasting damage than rigging what they are shown, because you cannot easily unteach a person.
  9. Capture asymmetry (training vs. preservation): A corrupted library can be rebuilt from other sources, but people trained to read wrongly are slow to retrain, so corrupted teaching is usually worse.

← Groups, splits and scale · All themes · AI in the chain →