organizational forgetting

The formal model (Argote, Beckman & Epple 1990: experience depreciates by a constant factor δ each period, E_t = δE_{t-1} + q_{t-1}) that treats accumulated production knowledge as something an organization can lose, not just accumulate. It is the mechanism by which the vault's standard learning curve can run backwards — costs rising when production slows — and the rate at which forgetting happens turns out to vary sharply by industry: much slower in Benkard's aircraft-manufacturing case than in the ship-production and franchise settings the specification was first tested on. There is an unexplored, uncausal echo of the same shape in neural-network forgetting curves.

This is the econometric branch of "organizational forgetting" specifically — a 2026-07-26 capture found the label is overloaded in the wider literature, and that a separate interpretive branch (entity-collective-forgetting-interpretive, descended from Halbwachs rather than production economics) shares the name but not the mechanism, and the two rarely cite each other.

References

• claim-organizational-forgetting-can-reverse-the-learning-curve · claim-benkard-2000-estimate-implies-far-less-forgetting-than-prior-studies · claim-benkard-2000-headline-depreciation-figure-is-the-spillover-adjusted-specification

• claim-organizational-forgetting-names-two-unrelated-research-traditions (disambiguation from the interpretive branch)

• claim-david-brachet-2011-splits-forgetting-into-continuous-and-discrete-channels · claim-david-brachet-2011-turnover-channel-twice-skill-decay-channel (a second, independent decomposition of the econometric branch, on a paramedic-workforce dataset)

• claim-thompson-2007-liberty-ship-forgetting-rate-is-specification-sensitive-by-his-own-evidence · claim-thompson-2007-re-estimate-is-one-fifth-to-one-seventh-of-argote-1990-25-percent-figure · claim-thompson-2007-attributes-inflated-forgetting-estimate-to-fischer-1949-aggregation-bias (a third primary re-estimate of the founding Liberty-ship case itself)

• Related hubs: entity-c-lanier-benkard · entity-linda-argote · entity-lockheed-l-1011-tristar · entity-collective-forgetting-interpretive · entity-jingle-fallacy · entity-peter-thompson · entity-leonard-rapping

• Captures: 2026-07-11-hop-learning-curve-unidentified, 2026-07-24-does-benkard-2000-report-the-organizational-forgetting-depreciation, 2026-07-26-hop-forgetting-has-two-lineages, 2026-08-11-bipartite-two-things-this-vault-knows-in-different, 10-inbox/raw/2026-09-29-does-peter-thompsons-2007-liberty-ships-paper-establish.md

• 2026-09-29: Thompson (2007), read directly at primary via a Wayback Machine capture, re-estimates the Liberty-ship case itself (not a second industry, like Benkard's aircraft case) at 3.6–5.7 percent monthly after correcting for a named aggregation bias in the standard Fischer (1949) dataset — roughly a fifth to a seventh of the case's founding 25%-per-month figure — and his own six specifications, plus his own concluding caveat, demonstrate the rate's specification-sensitivity as a finding in its own right, not an artifact of later citation. See claim-thompson-2007-liberty-ship-forgetting-rate-is-specification-sensitive-by-his-own-evidence, claim-thompson-2007-re-estimate-is-one-fifth-to-one-seventh-of-argote-1990-25-percent-figure, and claim-thompson-2007-attributes-inflated-forgetting-estimate-to-fischer-1949-aggregation-bias.

• 2026-08-11: A second econometric decomposition (David & Brachet 2011, on Mississippi paramedic/ambulance-company data, independent of the Argote/Benkard/Lockheed line already on this page) splits organizational forgetting into a continuous skill-decay channel and a discrete turnover-loss channel — and structurally, without either citing the other, parallels ML's catastrophic-vs-graceful forgetting split. See claim-david-brachet-2011-splits-forgetting-into-continuous-and-discrete-channels.

• 2026-08-12: A direct, complete-enumeration bibliometric check (David & Brachet's 26-entry reference list, and the 60 papers Semantic Scholar indexes as citing it) found zero contact with the ML catastrophic-forgetting literature in either direction — the strongest evidence yet that this econometric branch and the neural-network literature share no citation lineage, though confirmed for only this one founding pair. See claim-david-brachet-2011-reference-list-contains-zero-ml-citations and claim-david-brachet-2011-citing-papers-contain-zero-ml-work.

• 2026-09-27: A hand-comparison of the one fetchable pair of primary reference lists in this cluster (Benkard 2000 vs. David & Brachet 2011, both citing Argote/Beckman/Epple 1990) found divergent, not identical, title-quoting — weak evidence against a shared copy-fingerprint on that specific link, but a sample of one and no resolution of the cluster-wide question. The rest of the cluster's primary reference lists sit behind paywalled venues not reachable this session, and no published study has applied a copy-fingerprint audit to this literature. See question-do-organizational-forgetting-reference-lists-show-copy-fingerprints and claim-no-published-copy-fingerprint-audit-of-organizational-forgetting-cluster.