Shirky, “Ontology Is Overrated” (2005)

What it is Clay Shirky argues that designing categories in advance to cover cases that have not happened yet is badly overrated in digital systems. His own wording is that the strategy “is both widely used and badly overrated in terms of its value in the digital world.” The essay is “a heavily edited concatenation” of two 2005 talks: “Ontology Is Overrated” at O’Reilly ETech in March and “Folksonomies & Tags” at IMCExpo in April. He is precise about his target: not classification in general, but ontological classification, which organises entities by their essences and possible relations, and assumes a new item’s logical place already exists before the item does.

Key concepts

• The periodic table as the best case. It organises by proton count and gets enormous predictive value. Even so, “noble gas” is a frozen accident: helium is no more essentially a gas than mercury is essentially a liquid, but the chemists could not get it cold enough to see that. If a near-perfect scheme over physical essences still has context errors, softer domains have no chance.
• Library catalogues optimise for shelves, not concepts. Dewey’s 200s split religion into nine subdivisions, eight of them Christian or Christian-adjacent, with everything else in “290 Other religions.” Library of Congress History puts DA Great Britain and DC France at the same level as DR Balkan Peninsula, DS Asia and DT Africa: “these are all the top-level categories — all of these things are presented as being co-equal.” What is being optimised is books on a shelf. “The essence of a book isn’t the ideas it contains. The essence of a book is ‘book.’”
• There is no shelf. Digital items have no physical constraint forcing one item into one place, so the constraint that produced hierarchical classification is gone.
• The parable of the ontologist. Yahoo, free of physical constraints, hired a professional ontologist and added the shelf back: Books and Literature appears in Entertainment with an “@” marking that it is not really there. Yahoo chose to assert what is real rather than let items sit in many places at once.
• The alternative. Links and tags. Free-form labelling looks like a recipe for disaster, but large messy data sets yield surprising value.

How you’d use it Shirky offers not a test but what he calls “a partial list of characteristics,” in two groups. Domain: small corpus, formal categories, stable entities, restricted entities, clear edges. Participants: expert catalogers, an authoritative source of judgment, coordinated users, expert users. “The more of those characteristics that are true, the better a fit ontology is likely to be.” His worked positive examples are the periodic table and the DSM-IV, “a classic example of an classification scheme that works because of these characteristics”; his negative list is the mirror image — large corpus, no formal categories, unstable entities, no clear edges, uncoordinated and amateur users. Inverting the list into a go/no-go screen for a project is my framing, not his.

LLM angle none stated

Pitfalls & lessons The essay is regularly overread as “ontologies are useless.” Shirky attacks a priori categorisation applied to unbounded, uncoordinated, user-facing corpora. He does not deny that the periodic table works. Dave McComb of Semantic Arts published a direct rebuttal in 2015 to Shirky’s earlier syllogism essay of 7 November 2003, arguing that “this entire article is a syllogism” and conceding that “we still have a long way to go to staunch the critics.”

Verdict Worth reading in full. It is the sharpest statement of when formal classification does not pay, and its characteristic list is directly reusable as a project screen — provided you notice that Shirky offers it as description, not as a decision procedure.

Sources consulted

• http://shirky.com/essays/ontology-is-overrated-categories-links-and-tags/
• http://shirky.com/essays/the-semantic-web-syllogism-and-worldview/
• https://www.semanticarts.com/shirky-syllogism-and-the-semantic-web/
• research/firecrawl/crit-shirky-overrated.md
• research/firecrawl/crit-shirky-syllogism.md
• research/firecrawl/crit-semanticarts-reply.md