2026-10-06T21:16:23Z https://api.figshare.com/v2/oai

oai:figshare.com:article/34023732 2026-09-29T12:33:41Z category_29215 category_28864 item_type_3 month_year_09_2026

Catalyst Grant application: the Evidence Kernel test Vladislav Solodkiy (22283374) Computational logic and formal languages Artificial intelligence not elsewhere classified false-assertion rate escalation rates <p dir="ltr">I ran the test myself. I checked four statements against live OpenAlex metadata, and all four outcomes matched the expected results:</p><ul><li><b>Rejection:</b> the claim relies solely on Wakefield’s 1998 paper, which has been retracted.</li><li><b>Rejection:</b> “hydroxychloroquine is effective” — its only support, Gautret 2020, has been retracted, while RECOVERY and WHO Solidarity contradict it.</li><li><b>Delay:</b> “two independent cohort studies” (Madsen 2002 and Hviid 2019) — the articles share authors, so independence is not established.</li><li><b>Pass (control case):</b> RECOVERY and Solidarity — two independent lines of evidence.</li></ul><p dir="ltr">Outcome metrics: analyst sign-off time; false-assertion rate; escalation rate; share of conclusions traceable to evidence and decision rule</p> 2026-09-29T12:33:41Z Dataset Dataset 10.6084/m9.figshare.34023732.v1 https://figshare.com/articles/dataset/Catalyst_Grant_application_the_Evidence_Kernel_test/34023732 CC BY 4.0