Retrieval-Augmented Generation (RAG)

Definition

Retrieval-Augmented Generation (RAG) is an AI technique that grounds language-model responses in a specific corpus: a system first retrieves relevant passages from an archive (via semantic search over vector embeddings) and then generates a natural-language answer from them. In Benjamin’s work, RAG is not an abstract method but an enabling technology for a “paradise of knowledge” — the layer that makes a knowledge commons queryable, turning static archives into living systems that communities can ask questions of in their own language.

Democratizing Expertise

Benjamin’s primary frame is civic: RAG democratizes access to expertise by making archives queryable rather than locked behind specialist mediation. His recurring example is Heenal Rajani’s Open Council project, which archived roughly 1,500 city council meetings from London, Ontario over ten years and made them queryable through a RAG chatbot that surfaces insights across transcripts, minutes, and press reports simultaneously. The vision he draws from it — “If every city had this, collective insights could talk together” — positions RAG as infrastructure for collective intelligence across place-based communities, where natural-language questions replace the need for librarians or specialists to broker access.

Infrastructure for the Knowledge Commons

RAG sits within Benjamin’s broader knowledge commoning workflow. Elder interviews and practitioner knowledge are captured, transcribed, and processed by agents that extract patterns and tag entities, then ingested into shared repositories (Obsidian vaults, GitHub). RAG provides the queryable interface on top of this substrate, and connects to his interest in federated indexing and agent-mediated translation: agents not only federate knowledge between bioregions but help communities query it in their own taxonomy. This is also temporal — indigenous knowledge systems embed timing into what surfaces when, and semantic search makes such context-sensitive retrieval possible. Within the Knowledge Organizational Infrastructure (the COI protocol’s “mycelial substrate” for federated, holonically nested, sovereignty-preserving knowledge), RAG is the access layer that keeps embodied wisdom continuously accessible and updateable rather than frozen.

Commons Theory and Digital Un-Enclosure

The deeper grounding is Eleanor Ostrom’s “Knowledge as a Commons”: knowledge is non-rivalrous, subject to a tragedy of under-contribution, and requires governance to sustain. RAG functions as infrastructure that both preserves and democratizes this commons. It also appears in Benjamin’s post-capitalist vision, where AI drives marginal costs toward zero and enables digital un-enclosure — communities can build queryable knowledge infrastructure locally, reducing dependence on proprietary platforms and centralized gatekeepers of expertise.

In Benjamin’s Work

RAG is treated most directly in Knowledge Commons (via the Open Council example) and as the access layer of the Knowledge Organizational Infrastructure. Its post-capitalist framing is articulated in how-and-why-ai-will-accelerate-capitalism-and-then-destroy-it, and its role in place-based knowledge infrastructure resonates with the-infrastructure-of-belonging and the bioregional knowledge commons.