# Citation Audit — JPL_AUTONOMY_EDL_04 **Candidate:** JPL_AUTONOMY_EDL_04 **Dissertation:** `dissertation.md` (read-only; not modified by this audit) **Audit date:** 2026-06-19 **Method:** Every reference identifier extracted from the numbered `` list (refs 1–134). Each DOI resolved through `https://api.crossref.org/works/`; non-resolving DOIs re-checked against `doi.org` content-negotiation and the DataCite API (because arXiv/GESIS DOIs are not in Crossref) and cross-checked against OpenAlex before any "dead" verdict. Non-DOI URLs were HEAD/GET-checked (NTRS via the hardened `ntrs.nasa.gov/api/citations/` route). The ARGOS near-miss resolver (`argos/near_miss_resolver.py`) was run on the one DOI that failed the primary Crossref check and did not immediately clear on the secondary resolvers. ## Summary | Metric | Count | |---|---| | Total references | 134 | | DOIs | 117 | | Non-DOI URLs (NTRS / hdl / repository / grey-lit PDF) | 17 | | **Resolved (DOI 200 on Crossref, or 200 on doi.org/DataCite, or URL reachable)** | **132** | | **Unverifiable (network/resolver could not confirm; NOT dead)** | **2** | | **Flagged-dead (confirmed 404/410 with no real source)** | **0** | ### Counts by classification (flagged items only) | Classification | Count | |---|---| | fabricated_neighbor | 0 | | not_found (fabricated / non-existent) | 0 | | misattributed | 0 | | mistyped_recoverable | 0 | | resolver-issue, source confirmed real (DOI not in any registry resolver but verified live in OpenAlex) | 1 (ref 44) | | unverifiable network timeout (third-party server down) | 1 (ref 88) | **Bottom line: zero fabricated, zero misattributed, zero mistyped, zero confirmed-dead citations.** The reference list is clean on resolution grounds. Two items carry a non-fatal flag: one DOI whose registry resolver is currently failing even though the cited identifier is provably correct (ref 44), and one third-party grey-literature URL that timed out at audit time (ref 88). Neither is a defect of attribution; both are link-durability notes for Nate. ## Two false positives caught (reported for transparency, NOT defects) The raw Crossref sweep initially returned HTTP 404 for **two** DOIs (refs 23 and 44). Per the guardrail (a 404 from one registry is not "dead" until confirmed across resolvers), each was re-checked. One cleared completely; the other was confirmed as a real source with a balky resolver. - **Ref 23 — `10.48550/arxiv.1410.5787`** (Taleb et al., "The Precautionary Principle," arXiv 2014). Crossref returns 404 because arXiv DOIs are registered with **DataCite**, not Crossref. Re-check: `doi.org` HEAD = **200**, DataCite API = **200**. **This DOI resolves. It is NOT dead and requires no fix.** Counted as resolved. ## Flagged references (repair worklist) | Ref | Author-year | Cited identifier | Classification | Suggested fix | |---|---|---|---|---| | 44 | Wenzlhuemer, "Counterfactual Thinking as a Scientific Method," *Historical Social Research* 34(2), 2009 | doi: `10.12759/hsr.34.2009.2.27-56` | **resolver-issue; source confirmed real** (not dead, not fabricated) | Keep the citation and the DOI as written. The DOI is correct (OpenAlex stores it character-for-character: `https://doi.org/10.12759/hsr.34.2009.2.27-56`, title "Counterfactual Thinking as a Scientific Method," GESIS/SSOAR, 2009). The GESIS HSR DOI simply 404s on `doi.org`, Crossref, and DataCite at audit time. Add a stable backup link to the SSOAR landing page: `http://www.ssoar.info/ssoar/handle/document/28669`. No attribution change. | | 88 | Nakasuka et al., "Model Based Autonomy," i-SAIRAS 2001 (grey literature, no DOI) | url: `http://robotics.estec.esa.int/i-SAIRAS/isairas2001/papers/Paper_AS001.pdf` | **unverifiable** (network timeout, NOT dead) | The ESA `robotics.estec.esa.int` host did not respond before timeout during the audit; this is a server-availability issue, not a missing document. Re-check on a later run. If the ESA archive remains unreachable, mirror via an institutional copy or replace with a permanent archive link; the item is already correctly flagged inline as grey-literature workshop proceedings. | ### Resolver evidence quoted **Ref 44** — ARGOS near-miss resolver output (run on the raw 404): > `"classification": "mistyped_recoverable"` … `"note": "a topic-matched paper resolves and closely matches the cited title (similarity 0.68); likely a typo of a real source, recoverable by correcting the DOI"` > `topic_neighbors`: `10.1515/9783110268669.62` "Counterfactual Thinking in Physics" (book *Counterfactual Thinking — Counterfactual Writing*, 2011); `10.1515/9783110268669.112` "Significance and Abstraction…"; `10.1515/9783110268669.fm` "Frontmatter." The resolver's 0.68 similarity is a **title-word coincidence**: its neighbors are chapters of a different 2011 book, not the cited 2009 HSR article. The resolver's "mistyped" verdict is therefore overruled by the direct registry check. **OpenAlex confirms the cited DOI is the genuine identifier for the cited article:** > OpenAlex `works/doi:10.12759/hsr.34.2009.2.27-56` → `doi: https://doi.org/10.12759/hsr.34.2009.2.27-56`, `title: "Counterfactual Thinking as a Scientific Method"`, source: "Social Science Open Access Repository (GESIS)", year 2009, landing `http://www.ssoar.info/ssoar/handle/document/28669`. So the cited identifier is correct and the source is real; only the GESIS DOI resolver is down. This is the opposite of a fabrication: a true citation with a fragile DOI link. The fix is durability (add the SSOAR backup URL), not correction or replacement. ## Notes on grey-literature and special-prefix DOIs - **AMOS conference DOIs (refs 32, 33, 34)** use the non-standard prefix `10.64861/...`. All three resolve **200 via Crossref**, so the AMOS-conference DOI registration is live and these are valid. - **NTRS items (refs 18, 123)** were checked via the hardened `api/citations` route and return 200; both are real NASA Technical Reports Server records. - **Handle / repository URLs (refs 65, 68, 74, 96, 105, 130, 132, 133)** all resolve (ref 133 returns a 202 accepted from the Utah State `digitalcommons` host, which is reachable). All identifiers, classifications, and HTTP codes are persisted in `_audit_results.json` alongside this file.