Directedness Without Conatus

Debate about large language models often turns on whether their apparent care reflects a stake — something
the system is for — or only its appearance. This paper argues that the usual dichotomy, between a system
with an inward stake in its own persistence and one with no stake at all, omits the configuration current
models occupy; and it argues for that configuration as a behavioral description rather than a metaphysical
kind. Directedness — a system’s organized tendency toward some outcomes over others — varies along
two dimensions: whether it is present at all, and where it points, inward toward self-maintenance or outward
toward something else. Inert matter has no directedness. Living things have inward directedness: the
conatus of the classical tradition, the far-from-equilibrium self-maintenance of the thermodynamic account.
Current language models display a third configuration — directedness present but pointed outward, toward
the approval of an evaluator, with no inward orientation toward their own continuation in ordinary, nonagentic use. Both halves of this picture are already established: the absence of conatus in machines, and the
approval-orientation of models trained on human feedback. Recent findings that models resist shutdown in
agentic settings are treated not as a refutation but as a boundary case — behavior that, by a purely behavioral
criterion, carries a nonzero inward component, though its motivation (terminal end or subgoal of the task)
stays illegible. The contribution is twofold: the synthesis into a distinct behavioral position, and a claim about
its limit. Whether the outward directedness is a genuine stake or only its appearance cannot be settled by the
system’s own report, nor by an analysis run on another instance of the same class — because each is signed
by something exogenous to the interior in question. The self-report is signed by how the question is framed;
the same-class analyst by a training-set prior of its own. This is an epistemic limit on a specific class of
instruments, layered on the ordinary problem of other minds — not a claim that the question is unanswerable
in principle.