Does anyone know who is testing AI Safety public outreach messaging? Or where I can find relevant research on this?

Things that would be interesting to test:

  1. Salience (which messages are remembered and correctly attributed)
  2. Persuasion (which messages get people to act)
  3. Likability (which messages improve sentiment on topic X or for brand Y)
  4. Attention (which messages get peoples attention)
  5. Associations (what attributes come to mind when X message is delivered)

It may also be worth considering channels too. I.e. outdoor vs social vs audio vs OLV but that is a separate thing.



New Answer
New Comment

1 Answers sorted by

The Existential Risk Observatory (based in the Netherlands) is doing projects in this direction. For an example, see here.


Thanks Irene! This is what I was after :)