Context misfit

A term coined by Abdulhamid, Nyairo, Ankrah & O'Neill (arXiv:2610.31344, 2026) for recurring generative-AI output mismatches along three dimensions: local institutions and place, language practices, and representational fidelity. The paper frames it as a mechanistic account of how known categories of geographic, linguistic, and representational bias surface as everyday breakdown-and-repair in client-facing small-business work, documenting a five-part repair repertoire (rewriting, constraint layering, source substitution, tool switching, disengagement). First encountered in this vault 2026-10-24, in a single unrefereed preprint; watching for recurrence before promoting to a full hub.

References

• claim-generative-ai-context-misfit-three-dimensions-repair-repertoire

• 2026-10-31: A targeted corroboration search found no independent paper using the term "context misfit" or reproducing its three-dimension taxonomy or five-part repair repertoire; neighbouring LLM terms ("context confusion," "context-memory conflict," "misleading context") name within-session reasoning failures, not the localization/representation output mismatch described here. Absence of evidence after search, not evidence of absence — the term may simply be too new. Still single-source; see question-corroborate-abdulhamid-2026-generative-ai-workable-at-work-single-source.