wait. perhaps common knowledge is only possible AT ALL because of entanglement / mutual simulation / virtual bargaining
Common knowledge, trust, and mutual simulation
27 tweets · August 2023 · 148 likes · 7 retweets · read on Twitter
there's a kind of magic that happens when somehow the infinite regress of "who knows who knows what" clicks into "we know it"
😮 and "trust" is the name of this magic!
or "being in sync", as RV puts it in one of my favorite tweets
there's a question I have though, around the is-ought thing... there's clearly some difference between seeing together the world and being on the same team—oh, hang on. got it I think.
yeah okay the difference is whether our common-knowledge is just about the world beyond us, vs about ourselves and each other do you know what matters to me, and I to you? do you know my affordances, and I yours?
in the preconquest / pre-snake-cult era, we didn't use recursion to solve game theory (we didn't have it!) we just lived in each other transparently and honestly, and thus knew what we wanted, and we chose cooperation together when the option existed
but sometimes we want different things! real fighting is costly—animals have some sort of mutual simulation that means that in-species fighting often ends with submission rather than injury or death
relatedly, my impression is that long-ago people often resolved conflict by just spreading out further; this would be a natural equilibrium when it was possible (until it wasn't because we were everywhere; probably one source of the pressure that tipped 10kya)
key difference: animals simulate each other's behavior humans simulate other beings' *experience* in first person "what is it like to BE my prey?" trippy question. x.com/ChasingMyself/…
where was I? common knowledge requires mutual simulation in some sense, obviously. but the point is that the byzantine generals problem only happens if the generals are insufficiently intertwingled
in the absence of deep trust/entanglement, agreements are fragile, based on some amount of pretending / fake trust, and so we need to keep reinforcing them
non-naive trust is made out of relentless honesty and curiosity, which allows discovering together the degree of win-win that is possible
by play-dancing with a person (or god or egregore) I can learn to trust whatever causal forces in the universe link us, which allows us to choose win-wins together 🔵
but to the extent that the capacity for play is broken, it's hard to even find a place from which to learn the only gods that feel plausible can be made of utter capricious chaos x.com/HephaistosF/st…
and yet, despite appearances, we're not separate even as we feel separate and my suffering matters to all the gods worth cooperating with; they're all there, already
and your suffering matters to me, even if I'm not experiencing it in first person much or at all, given my current capacity and everything else
the whole is whole; god is becoming (swap "omnipresent" for "omnibenevolent" below; typo)
and I trust we can collectively find each other in this dance
my synthesis of some dialogue happening in the replies:
in my late-night haze posting this, I failed to link the 2 MAIN SOURCES that contributed to the nonobviousness of my insight 1. this LW post, which I have issues with but the "playing with omega with monopoly money" is super relevant lesswrong.com/posts/PcfHSSAM…
2. Virtual Bargaining paper, which includes lots of excellent simple concrete realistic embodied examples (not contrived extreme bizarre heady game theory scenarios) seems it already mentions infinite regress, but I hadn't applied that to common knowledge
@AndrewCutler13 DM'd me saying he agreed (with this thread) > that the social fabric (including common knowledge) is woven recursively via modeling one another but I'm sorta saying the opposite—that by default it's something OTHER than many-step recursion
I don't quite know how to mathematically characterize the structure of common-knowledge ...and it seems to run into homunculus problems if I try, because even just having the me-&-you in my head virtually negotiate, well, that could have its own infinite regress internally!
ah, it recurses but not *infinitely* because it grounds out in ummm... a sense of "I can tell for myself". ahhh and having both negotiating parties in one head makes that WAY easier to access, because sensory reality is the same, meaning-making has common basis...
an alternative answer is just: - it's still recursive but it runs fast in world-model emulation, not slow in cognitive LeftHemisphere game theory land - it eventually reaches a level of probability where action feels better than inaction, so you act (obviating logical certainty)
very excellent thread that takes a totally different angle on the role of mutual simulation in trust-building that in particular it's relevant that we simulate how each other would act under different power differentials