Malcolm Ocean's roamblog

notes on What are plans for? by David Chapman & Phil Agre

created 2021-02-18 · edited 2021-02-18 · open in Roam
  • Notes mostly from May 6th, 2020
  • Lots of rich stuff in here and I only started taking notes partway through. The basic jist is "plans as communication"
  • p18
    • > "Communication doesn't pick up a "meaning" from my head and set it down in yours. Instead, communication is part of the work of maintaining a common reality. The players share a common reality because they are competent players and because they use language to keep their shared reality in good repair."
  • p20-21
    • > "Interleaved planning and improvisation differ in their understanding of trouble. In the world of interleaved planning, one assumes that the normal state of affairs is for things to go according to plan... In the world of improvisation, one assumes that things are not likely to go according to plan. Quite the contrary, one expects to have to continually redecide what to do.
    • [...]
    • > We think of this distinction between interleaved planning and improvisation in terms of the theme of control. A system that operates by constructing and executing plans lives, to speak metaphorically, in a sort of fantasy world, the one it projects when it reasons about its future actions. In this way, the system believes that it has a kind of control over its world that, at least in many domains, is not realistic. When things do not work out as projected, the system is surprised.
    • > An improvising agent, by contrast, does not live its life through an alternation between fantasy and surprise.
    • > It does not believe that it has complete control over its world. Instead, through a continual give-and-take with its environment, creatively making use of opportunities and contingencies, it participates in the forms of activity that its world affords.
  • Damn, this paper was published by MIT's AI Lab, but it's fucking dharma if you know how to read it.
    • Had Chapman been studying the dharma already? I thought he got into that a bit later.
    • The passage that prompted this reflection was on p23
      • > Our solution to this puzzle is to abandon the notion that activity has a particular locus of control at all.
  • I think this is actually a really great resource for explaining post-surprise - the whole article, and particularly this quote: