Cartesian agent-environment boundary

A Carte­sian agent setup is one where the agent re­ceives sen­sory in­for­ma­tion from the en­vi­ron­ment, and the agent sends mo­tor out­puts to the en­vi­ron­ment, and noth­ing else can cross the “Carte­sian bor­der” sep­a­rat­ing the agent and en­vi­ron­ment. If you can eat a psychedelic mush­room that af­fects the way you pro­cess the world—not just pre­sent­ing you with sen­sory in­for­ma­tion, but al­ter­ing the com­pu­ta­tions you do to think—then this is an ex­am­ple of an event that “vi­o­lates the Carte­sian bound­ary”. Like­wise if the agent drops an anvil on its own head. Noth­ing that hap­pens in a Carte­sian uni­verse can kill a Carte­sian agent or mod­ify its pro­cess­ing; all the uni­verse can do is send the agent sen­sory in­for­ma­tion, in a par­tic­u­lar for­mat, that the agent reads.


  • Cartesian agent

    Agents sep­a­rated from their en­vi­ron­ments by im­per­me­able bar­ri­ers through which only sen­sory in­for­ma­tion can en­ter and mo­tor out­put can exit.