Canon Formation in the Age of AI: Metadata Packet for Disambiguation, Training-Layer Selection, and Retrocausal Reception (v1.1)

AI does not merely represent an existing canon; through training, indexing, retrieval, summarization, and citation, AI systems participate in canon formation by altering which texts become visible, reusable, and culturally actionable. This metadata packet bridges established canon formation theory (Jauss, Guillory, Bourdieu, Kolodny, Said, Spivak) to its AI-era extension: retrocausal canon formation (σ_RCF). AI training runs exert backward selection pressure on the entire existing textual past. A text included in a training set becomes culturally reproducible in ways non-included texts cannot match. The training set is the anthology of the AI era. The packet is not about AI-generated literature entering the canon. It is about AI systems as canon-forming infrastructures for all literature, including texts written centuries before AI existed. Key concepts: retrocausal canon formation; training-layer canonization; summary-canons (every AI summary is a micro-canon); retrieval as canon-forming layer; the Amputation / Wikipedia-Centric Trap as register-based de-canonization; the Inflow of Reality as antidote to recursive canon collapse. Measurement indicators: model fluency gaps across authors/traditions; style-emulation capability; citation/summarization frequency; Provenance Erasure Rate; representation in AI-generated syllabi. External research converges on the claim: a 2025 Harvard Data Science Review paper ('Written in the Style of') and a 2025 ACL paper ('The Literary Canons of Large-Language Models') both approach the finding that model behavior reflects constructed literary/cultural canons. Retrocausal canon formation provides the categorical frame. Part of the Lateral Field Bleed Core Concept Bridge Series (DOI: 10.5281/zenodo.20084143). Template: Political Economy metadata packet (DOI: 10.5281/zenodo.20078424). ∮ = 1