The Encyclopédie and the Renvois

In the AI Story

Diderot understood that cross-references were editorial judgments. He wrote in the Preliminary Discourse that the renvois were designed to indicate the close connections among human knowledge, and he acknowledged that readers must exercise judgment in following them — not every suggested connection would prove productive. The renvois were hypotheses about relationship, proposed by curators for evaluation by readers.

The Encyclopédie embodies an alternative to pure alphabetization that the entire history of reference works has explored. Alphabetical order is arbitrary and therefore transparent — the reader knows how it works without mastering any theory of knowledge. Systematic order embeds an interpretive framework and therefore teaches, but it demands mastery of its scheme. The renvois system attempted to have both: alphabetical arrangement for retrievability, cross-references for intellectual depth.

Large language models occupy a third position in this long tension. Their organization is neither alphabetical nor systematic nor even algorithmic in the search-engine sense. It is emergent: the product of statistical patterns in training data. The model's implicit cross-reference system is richer than Diderot's renvois in volume but poorer in intellectual curation. Diderot's connections reflected deliberate editorial judgment; the model's connections reflect the statistical frequency of co-occurrence in the training corpus, which is a measure of cultural habit rather than intellectual significance.

The contrast has practical consequences. A reader of the Encyclopédie could evaluate the renvois against her knowledge of the subjects, could see whether a given cross-reference was illuminating or misleading, and could supplement the curated connections with her own based on independent knowledge. A user of a language model cannot see the organizational principles, because they are not explicit. She must infer the structure from the outputs and evaluate each connection independently — a heavier evaluative burden that the hidden index problem names.

Origin

The Encyclopédie, ou dictionnaire raisonné des sciences, des arts et des métiers was edited primarily by Diderot and Jean le Rond d'Alembert and published in Paris between 1751 and 1772. It involved hundreds of contributors and became the central reference work of the French Enlightenment, repeatedly suppressed by religious and political authorities and repeatedly surviving through subscriber networks and clandestine publication.

Key Ideas

Curated connection. The renvois were deliberate editorial judgments about which relationships between domains mattered.

Network knowledge. The Encyclopédie's organizational innovation embodied the view that knowledge is a web of relationships, not a list of independent facts.

Transparent curation. The reader could evaluate the renvois because they were explicit; the curatorial judgment was visible and therefore critiquable.

Contrast with emergent organization. Large language models produce connections by statistical rather than intellectual process, and the opacity of the process undermines the reader's capacity for critical evaluation.

Judgment burden scales with opacity. The less visible the organizational scheme, the more evaluative work the reader must perform independently.