Doctorow, “Metacrap” (2001)

What it is Cory Doctorow’s essay against voluntary metadata, version 1.0 dated 15 May 2001 and version 1.3 dated 26 August 2001 — months after the Scientific American article, though Doctorow never mentions it, the Semantic Web, or Berners-Lee. His stated target is broader: “explicit, human-generated metadata has enjoyed recent trendiness, especially in the world of XML.” His thesis is that a world of exhaustive, reliable metadata would be a utopia, and is also “a pipe-dream, founded on self-delusion, nerd hubris and hysterically inflated market opportunities.” He gives what he calls “at least seven insurmountable obstacles,” under the section title “seven straw-men.”

Key concepts The seven obstacles:

1. People lie. Metadata exists in a competitive world, and when poisoning the well pays, the well gets poisoned. His examples are spam subject lines and keyword-stuffed press releases.
2. People are lazy. Untitled files, subject-less email, MP3s with no track information despite a one-click lookup button.
3. People are stupid. His eBay example: search “plam” and find nine misspelled Palm Pilots, cheap, because misspelled listings never appear in correct searches.
4. Mission: Impossible — know thyself. “People are lousy observers of their own behaviors.” Nielsen’s log-books skewed heavily to Masterpiece Theater and Sesame Street; replacing the journals with set-top boxes “showed what the average American family was really watching.”
5. Schemas aren’t neutral. A maker of efficient washing machines ranks by energy and water use. A maker of feature-laden ones ranks by colour and programmability. Any hierarchy asserts that some axes matter more.
6. Metrics influence results. “Agreeing to a common yardstick … necessarily privileges the items that score high on that metric.” His example: MTV stopped showing videos because “Nielsen couldn’t generate ratings for three-minute mini-programs, and so MTV couldn’t demonstrate the value of advertising on its network.”
7. There is more than one way to describe something. Requiring one vocabulary “denudes the cognitive landscape.”

How you’d use it Use obstacles 1 to 4 as an incentive audit on any proposed ontology. Ask who enters the data, what they gain from entering it well, and what they gain from entering it badly. Use obstacles 5 to 7 as a design review: name the axes the model privileges and say out loud whose interests they serve.

LLM angle Language models change the economics of obstacles 2, 3, and 4 by extracting structure that nobody wanted to type. They do nothing about 1, 5, 6, or 7, because those are about incentives and power rather than effort.

Pitfalls & lessons Doctorow’s own conclusion is not that metadata is worthless. He endorses implicit metadata, citing Google’s use of link structure — data produced as a by-product of behaviour rather than declared. In his words, “this sort of observational metadata is far more reliable than the stuff that human beings create for the purposes of having their documents found.” That distinction, observed signal versus declared signal, is the durable part of the argument.

Verdict Worth reading in full; it is short, and it is the incentive argument every ontology proposal has to answer. Read it as an attack on hand-declared metadata generally, not as a reply to the Semantic Web article — it never mentions it.

Sources consulted

• http://www.well.com/~doctorow/metacrap.htm (original)
• https://chnm.gmu.edu/digitalhistory/links/pdf/preserving/8_17.pdf (archived copy consulted)
• research/firecrawl/crit-doctorow-metacrap.md