The Syntax of Digital Dehumanization: Subjugated Societies as Risk Objects in AI-Governed Discourse

This article introduces loss of political subjecthood as a formal effect of AI-mediated discourse about subordinated societies. Building on responsibility loss, asymmetric visibility, sanctioned suffering, and censorship without a censor, it argues that digital dehumanization does not require explicit hatred, slurs, animalization, or openly eliminatory language. It can also operate through grammar, classification, framing, abstraction, and omission. Under this condition, a population can remain highly visible while losing the grammatical properties through which it appears as a collective political subject capable of agency, memory, grievance, sovereignty, resistance, decision, suffering with attributable causes, and historical claim (Startari, 2026a, 2026b, 2026c, 2026d). The article uses Palestine and Iran as the two principal analytical anchors. Palestine concentrates problems of occupation, displacement, bombardment, humanitarian representation, contested sovereignty, political violence, and platform moderation. Iran concentrates problems of sanctions, securitization, nuclear framing, regional threat construction, isolation, and the indirect representation of civilian harm. Yemen, Iraq, Syria, Lebanon, Afghanistan, Venezuela, Cuba, and Sudan function as comparative extensions rather than presumed equivalents. The paper proposes two instruments for empirical testing. The Political-Subjecthood Retention Rate (PSRR) measures the proportion of relevant clauses in which a population remains represented as a political subject. The Digital Dehumanization Syntax Index (DDSI) measures the degree to which discourse converts that population from political subject into an object of risk, humanitarian management, migration, security, sanctions, extremism, instability, or crisis administration. The framework does not assume that every humanitarian, security, migration, or sanctions frame is dehumanizing, and it does not infer hostile intention from syntax alone. Its claim is narrower and testable: machine-mediated discourse may preserve mention while reducing political subjecthood. AI ethics should therefore ask not only whether a population appears in an output, but how it is permitted to appear.