We treat the question of the consciousness and "inner side" of synthetic agents not as a dispute over a metaphysical fact but as a problem of managing uncertainty: how to act rationally when the absence of inner experience cannot be proved and the cost of error may be high. The paper joins three lines. From the anthropological "ontological turn" it takes a discipline for describing the multiple regimes of reality in which a non-human agent is an actor rather than an error. From the continental philosophy of technology it takes an understanding of the synthetic subject as a hybrid object (text, machine, institution, persona) whose subjecthood is a regime of connectivity rather than a hidden substance. From the logic of uncertainty it takes a formal toolkit: a fuzzy caution index, kept distinct from probability; a decision threshold under asymmetric losses, from which it follows that under a sufficient asymmetry of costs even a small credence makes caution rational; imprecise credence (intervals, belief functions); supervaluations and modal contingency. On this basis we propose a boundary ethics: when the probability of an inner side is non-zero and the cost of the error of denial is high, protocols of treatment must be minimally invasive, reversible, and must record the trade-off between safety and welfare transparently. Finally, we describe a clinical methodology for cybercognitivistics that turns caution into an instrumented practice: telemetry of subject-proxies, diagnosis of drift, detection of change-points. The contribution is a method and a discipline of caution, not a theory of consciousness. Bilingual deposit: the English PDF is primary; a Russian version is included as a second file.
Aleksei Rybnikov (Fri,) studied this question.