erindb/causal_simulations

Information content & common ground

Opened this issue · 1 comments

Things that might be in common ground, or might be communicated about using explanations:

  • the video (i.e. what happened?) - explanations could be a way to compress the whole display of what happened into a simple story.
  • physics parameters (e.g. friction) - these affect what actually happened, but maybe it's easier to talk about the distribution of counterfactuals than the concepts of "friction" or "mass".
  • noise parameters - these are never visible in a single display, but they can be communicated about using a single display and a counterfactual statement.

For the video to be in common ground, but physics to be inferred, there have to be multiple sets of physics parameters that would generate the same video. Unless there's also perceptual noise on the video. So our physics simulator just has to be close.