Algorithmic monoculture

The economic/systemic framing, originating with Kleinberg & Raghavan (2021), for the risk that many independent decision-makers relying on similar algorithms produce correlated outcomes and systemic exclusion — distinct from any one algorithm's accuracy. First met by this vault on 2026-09-14 as the broader term under which entity-citation-monoculture sits ("not yet its own vault entity" at the time); promoted to its own hub on 2026-10-31 once a second, independent cluster — entity-kenny-peng and entity-nikhil-garg's LLM error-correlation work — showed the same framework recurring outside the matching-markets literature it started in. Matters to this vault as the shared theoretical ancestor for two distinct empirical bodies of work it now tracks: correlated decisions in matching markets, and correlated errors across LLMs.

References

• claim-correlated-errors-in-llms-exceed-chance-and-rise-with-accuracy
• claim-generative-process-diversity-predicts-lower-correlated-failure
• entity-citation-monoculture
• Capture: 10-inbox/raw/2026-10-31-does-ai-safetyinterpretability-literature-have-a-specific-name.md