Post-Agonistic Infrastructure: Justificatory Filtering and the Democratic Grammar of Aligned Language Models

This article argues that aligned large language models are not politically neutral infrastructures of discourse but are structurally predisposed toward deliberative democratic grammars over agonistic ones. It introduces the concept of justificatory filtering to describe how aligned systems govern which forms of political justification and claim-making are treated as answerable and worthy of assistance at the moment of composition. It further argues that the cumulative democratic effect of this mechanism is post-agonistic infrastructure. The manuscript includes a small-scale comparative probe across three commercial models and supplementary materials documenting prompts, responses, and coding categories. This version includes a corrected disclosure of AI tool use during the research and writing process. Version 3 adds a Russian-language version of the manuscript. The English manuscript and supplementary materials remain unchanged from Version 2.