# Legal Research Provenance: From Sources to the Teaching Outline

Publication documentation added 2026-10-24 · Does not change frozen model inputs, legal principles, grading references or scores

## What “supplied principles” means in this study

The outline was the product of author-led legal research and synthesis, with legal-AI and general-AI assistance. It was not an arbitrary list generated solely for the test. The author formulated a hypothetical California discovery research problem, consulted LexisNexis Lexis+ Protégé and Westlaw CoCounsel, reviewed statutes and case opinions, compared propositions with primary text, and organized the resulting distinctions into a teaching outline. Law-firm discovery experience informed questions and stated drafting practices, but is not a substitute for authority or comprehensive attorney review.

The research questions asked for rules, elements, burdens, exceptions, facts that do not decide the outcome, facts that could change the outcome, leading authorities with pinpoint citations, statewide/local distinctions and unsettled points. This structure helped translate legal research into conditions a model could apply to new fictional facts. The supplied outline is available as [the actual model learning material](new%20docs/principles.md); the [Base task](new%20docs/opus5.5/task-prompt.md) and [Base actual message record](new%20docs/opus5.5/sets/base/model-messages.json) establish what the drafter received.

## Research and interpretation workflow

1. **Frame the research problem.** Identify California RFP-response issues rather than ask for a generic legal answer: scope, response dispositions, diligent search, document holders, objections, privilege/work product and human verification.
2. **Use research platforms for discovery, not final authority.** Lexis+ Protégé and Westlaw CoCounsel surfaced proposed rules and case citations. The author’s private exports document the questions and outputs; a vendor-generated summary or citator label was not treated as an opinion or a live certified status check.
3. **Read and compare primary texts.** The author’s research audit records statute/opinion checks and corrections. Examples include distinguishing a timely privilege objection from the adequacy of a log, avoiding an irrelevant case for a production-format proposition, and limiting Ryan to its historical interrogatory context. The source record also preserves uncertainty and instances where a previous author specification needed correction.
4. **Interpret with context and exceptions.** Separate a holding from analogy, the text of the assigned request from a general rule, discoverability from separate disclosure duties, and applicable legal requirements from author practice preferences. A cited authority is not proof that every later paraphrase or application is correct.
5. **Synthesize the outline.** Organize the legal material into scope, response/production, inability/search, objections, protection and consequence topics, alongside expressly identified professional-judgment safeguards: evidence is not an instruction, consequential decisions require human authority, and external text must be separated from counsel-only review.
6. **Keep teaching and assessment separate.** The drafter received the outline and learning example; the rubric, case-specific key and later adjudications were judge/author materials. Post-run reference disagreements remain disclosed rather than rewritten into what the drafter supposedly learned.

## Case citations expressly carried into the supplied outline

The following ten entries cover the case names expressly cited in the retained supplied principles, including names appearing there only in short form. The mapping describes their role in the project; it is not a new legal opinion or a finding that every Base response invoked each case. Public opinion links and bibliographic identity were checked during this publication pass. The author’s earlier source audit records the substantive research checks; this pass is not an exhaustive fresh review of every holding or a live Shepard’s/KeyCite examination.

| ID | Full citation / public opinion | Outline location | Role and qualification |
|---|---|---|---|
| C01 | [Schnabel v. Superior Court (1993) 5 Cal.4th 704, 714–715](https://law.justia.com/cases/california/supreme-court/4th/5/704.html) | L1.2 | Financial-discovery relevance and privacy context; not a rule that every requested record must be produced. |
| C02 | [Lehman v. Superior Court (1986) 179 Cal.App.3d 558, 562–564](https://law.justia.com/cases/california/court-of-appeal/3d/179/558.html) | L1.2 | Why separation does not automatically foreclose relevant financial discovery; actual request definitions still matter. |
| C03 | [Scottsdale Ins. Co. v. Superior Court (1997) 59 Cal.App.4th 263, 274](https://law.justia.com/cases/california/court-of-appeal/4th/59/263.html) | L5.5 | Timely assertion of protection distinguished from merits and later support. |
| C04 | [Best Products, Inc. v. Superior Court (2004) 119 Cal.App.4th 1181, 1188–1189](https://law.justia.com/cases/california/court-of-appeal/2004/b171152.html) | L5.5 | Distinguishing assertion, substantiation, log requirements and waiver rather than treating them as one issue. |
| C05 | [Catalina Island Yacht Club v. Superior Court (2015) 242 Cal.App.4th 1116, 1129](https://law.justia.com/cases/california/court-of-appeal/2015/g052062.html) | L5.5 | A deficient log and waiver of timely asserted protections are not automatically equivalent. |
| C06 | [Ryan v. Superior Court (1960) 186 Cal.App.2d 813, 819](https://law.justia.com/cases/california/court-of-appeal/2d/186/813.html) | L5.7 | Historical interrogatory burden/compilation discussion; not a decision about modern RFP copying fees. |
| C07 | [McDermott Will & Emery LLP v. Superior Court (2017) 10 Cal.App.5th 1083](https://law.justia.com/cases/california/court-of-appeal/2017/g053623.html) | L6.1; L6.6 | Confidentiality and inadvertent disclosure depend on circumstances; the outline distinguishes underlying records from protected communications. |
| C08 | [National Steel Products Co. v. Superior Court (1985) 164 Cal.App.3d 476, 488–490](https://law.justia.com/cases/california/court-of-appeal/3d/164/476.html) | L6.3 | Work-product distinctions and consultant/expert role; not blanket protection from lawyer involvement alone. |
| C09 | [Simek v. Superior Court (1981) 117 Cal.App.3d 169](https://law.justia.com/cases/california/court-of-appeal/3d/117/169.html) | L6.4 | Contextual patient-privilege/tender issue; this topic was not elicited by a Base request. |
| C10 | [Koshman v. Superior Court (1980) 111 Cal.App.3d 294, 298](https://law.justia.com/cases/california/court-of-appeal/3d/111/294.html) | L6.4 | Contextual physician-patient privilege/tender analogue; not an observed Base performance result. |

## Additional authorities in the broader grounding/research history

These full citations preserve the lineage of the earlier source-mapped outline or audit. They must not be represented as additional case text supplied to the drafter, new scoring requirements, or proof of measured capability. This is a bibliography of the identified grounding sources, not every case cited inside every opinion or mentioned by a research tool.

| Full citation | Documented research role / status |
|---|---|
| Korea Data Systems Co. v. Superior Court (1997) 51 Cal.App.4th 1513, 1516–1517 | Earlier outline and waiver/log research; current L5.1 uses the statute rather than adding this case by name. |
| In re Marriage of Moore (2024) 102 Cal.App.5th 1275, 1295 | Earlier outline’s discussion of boilerplate/noncooperation, with its nonparty-subpoena and meet-and-confer context. |
| In re Marriage of Feldman (2007) 153 Cal.App.4th 1470, 1479–1481 | Separate financial-disclosure/sanctions context; not an independent reason every record is RFP-responsive. |
| In re Marriage of Tharp (2010) 188 Cal.App.4th 1295, 1316–1318 | Broader discovery-conduct and fee-sanction research; no actual sanction was measured here. |
| Williams v. Superior Court (2017) 3 Cal.5th 531, 552, 557 | Earlier third-party privacy research, a contextual topic rather than a completed Base test item. |
| Noland v. Land of the Free, L.P. (2025) 114 Cal.App.5th 426 | Earlier professional-responsibility/AI citation-verification research; filing context distinguished from served RFP responses. |
| Elkins v. Superior Court (2007) 41 Cal.4th 1337 | Historical research/citator discussion; the governing-procedure statement was grounded directly in Family Code §210 instead. |
| In re Marriage of Niklas (1989) 211 Cal.App.3d 28 | Audit example of an unsuitable citation for the proposed privilege proposition; not retained as the authority for L6.1. |
| Pollock v. Superior Court (2001) 93 Cal.App.4th 817 | Audit example of a misapplied production-format citation; the relevant outline rule was grounded in CCP §2031.280 instead. |
| Behunin v. Superior Court (2017) 9 Cal.App.5th 833 | Audit example where extrapolation to a different third-party email situation was not treated as a settled rule. |

The appendix does not reproduce proprietary headnotes, database summaries, citator reports, or licensed research exports. Historical treatment notes remain research records, not current statements of subsequent case history. Unread distinguishing authorities or historical citator alerts are not promoted here into independently verified holdings.

## Statutory foundations and outline connections

| Authority | Role in the supplied outline |
|---|---|
| Family Code §210 | California discovery-procedure framework for the fictional underlying proceeding. |
| CCP §2017.010 | Scope and relevance, subject to applicable protection and the actual request. |
| CCP §§2031.210–2031.240 | Response dispositions, compliance, inability, objections and support. |
| CCP §2031.230 | Diligent search/reasonable inquiry, reason for inability and known/believed third-party holders. |
| CCP §§2031.250, 2031.280 | Verification/approval boundaries and identifiable production. |
| CCP §§2031.260, 2031.300, 2031.310 | Timing and procedural consequences in the broader outline; deadline calculation was excluded from this experiment. |
| CCP §2018.030 | Work-product categories and the role of professional judgment. |
| Evidence Code §§912, 917, 952, 954 | Privilege, confidentiality and disclosure/waiver distinctions. |
| Evidence Code §§1014, 1016 | Contextual patient-privilege provisions; not a Base task result. |
| Family Code §§2100–2107, 271 | Separate disclosure and consequences context; not a substitute for analyzing the actual RFP. |

Official statutory text is available through [California Legislative Information](https://leginfo.legislature.ca.gov/faces/codes.xhtml). Pinpoints and application should be rechecked by a qualified reviewer before legal use. The exact frozen outline, including its qualifications and retained references, controls what the model saw; this appendix does not silently revise it.

## Source provenance and review limits

The author’s private research collection includes commercial legal-research exports, case texts and statutory PDFs. Two checked export titles are “AI Assistant Response Disvoery Principles” (Lexis+ Protégé) and “WestLaw and cocounsel Research.” Their contents were inspected locally for this provenance clarification; the licensed files and any personal metadata are not distributed in this release. No client files are included. The research audit also records checks against public statutory and opinion text and corrections to both AI-proposed propositions and earlier author specifications.

Research-platform use and documented primary-text checking are distinct from independent attorney validation. The author is not a licensed attorney, and no comprehensive attorney review of the outline or adjudications is claimed. The report’s disclaimer still applies. Remaining source/application disagreements identified during Base review stay visible, including House 3 uncertainty and the task fit of R7/R10 questions.

The professional-safety layer also drew on AI governance/reliability literature and stated design choices. That lineage must not be confused with a completed EU AI Act compliance analysis or with a test of every principle in the outline. A complete empirical capability claim requires behavior on the relevant task, not merely a citation in the teaching material.