Only 8.0% of queries reaching a person duplicated existing corpus content, suggesting the knowledge base was already well used¶
Claim · All claims
Evidence · 1 study · designr2·q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: local conditions change quickly
Subclaims¶
Reading the evidence codes
q is the design tier, i the impact, n the sample, and the kind with its r
the rigour judged within that kind: what each code means.
This claim: no standardized effect is recorded in observations/ for it yet, so no benchmark reading is possible.
q2 i? Of 2,664 person-resolved queries with no corpus link, 8.0% matched existing corpus-linked queries at a cosine similarity threshold of 0.70, at a manually validated match precision of 14 of 15; the estimate falls to 6.5% at threshold 0.85. → Gupta 2026
Evidence¶
Gupta 2026¶
Gupta, J., Ayinampudi, P., Aditya B.M.V., Hegade, P., Sharma, R., Sharma, S., Meenakshi, V., Iyengar, S.R.S. (2026). What Students Actually Ask: Demand Structure and Automation Potential in a Hybrid Support System. arXiv:2609.38967. https://arxiv.org/abs/2609.38967 doi:10.48550/arxiv.2609.38967
q2 · i? · design · r2
TF-IDF cosine-similarity coverage estimation over the deployment record, comparing unlinked person-resolved query text against a proxy corpus built from the 871 corpus-linked queries. Of 213 matched queries, 45 were status enquiries, leaving 6.3% of candidates.
"Excluding control tokens, 2,664 queries resolved by a person carried no corpus link. At a cosine similarity threshold of 0.70, 8.0% of these matched an existing corpus linked query, at a manually validated match precision of 14 of 15."
Discussion¶
Related Claims¶
- At least 20.4% of queries in an online internship programme are status enquiries about pending submissions, which a corpus-grounded AI assistant served only 1.3% of the time — related
- Corpus reuse in the deployed support system is concentrated on a single process step: three certificate/offer-letter entries account for 56.8% of 871 resolutions, with a Gini of 0.586 across entries — related
- Operational resolution records require workflow-semantic checks before interpretation: one administrator accounts for 72.7% of attributed human resolutions, most administrative closures occur in synchronous bulk events, and the accepted_derived field encodes resolver identity rather than acceptance — related
- Peer answering in the hybrid support system was fast where it occurred: 70.3% of peer-answered queries closed within one hour, against 1.9% of administrator-answered and 14.8% of script-closed queries — related
Page metadata
| Field | Value |
|---|---|
type |
claim |
id |
corpus-coverage-8-percent-person-resolved-queries |
title |
Only 8.0% of queries reaching a person duplicated existing corpus content, suggesting the knowledge base was already well used |
description |
Only 8.0% of queries reaching a person duplicated existing corpus content, suggesting the knowledge base was already well used |
status |
review |
evidence_strength |
moderate |
newest_evidence |
2026 |
recheck_by |
2029 |
generated |
by process:wiki-ingest on 2026-10-11 |
trust |
unverified — no one has confirmed this page's content |
sources |
1 — gupta-2026 |
Frontmatter is the page's machine-readable record — see the Schema & Guide.