legal-ai/docs/legal-decision-lessons.md

Legal Decision Writing - Lessons Learned

Lessons extracted by comparing our planning/drafts against Dafna's published final versions.

Source

• Published decision: 01_Projects/כתיבת החלטות משפטיות/ערר 1180-1181 הכט/החלטה/הכט 1180-1181.pdf
• Our draft: 01_Projects/כתיבת החלטות משפטיות/ערר 1180-1181 הכט/החלטה/דיון-והכרעה-טיוטה.md
• Date: February 2026

What Our Draft Got Right

• Section numbering continuity (no resets)
• "להלן" definitions with bold formatting
• Overall arguments structure (appellants > local committee > permit applicants)
• Citation of relevant case law (שפר, הימנותא, דסטגר)
• Clear separation between parties' arguments

What the Published Version Changed

1. Discussion Section Structure

• Draft: 6 sub-headers (H2) breaking the discussion into topics
• Published: ZERO sub-headers. One continuous flow of numbered paragraphs
• Lesson: The discussion reads as a legal essay, not a structured outline

2. Citation Technique

• Draft: Each case cited in its own paragraph (7 separate paragraphs for the proprietary claims section)
• Published: One massive paragraph (~600 words) citing through ערר נגאח 1011-03-25, which itself consolidated all the case law
• Lesson: "Citation through consolidating decision" technique

3. Paragraph Length in Discussion

• Draft: Uniform 50-70 words per paragraph
• Published: Ranges from 20 to 600+ words. Key citation paragraphs are very long.
• Lesson: Don't fragment long legal arguments into tiny chunks

4. Opening Formula

• Draft: "לאחר שבחנו את טענות הצדדים... החלטנו שיש לדחות את הערר על הסף"
• Published: "לאחר שבחנו... החלטנו בשלב ראשון כי... אך יחד עם זאת ועל מנת לא לצאת בחסר... מצאנו להוסיף מספר הערות"
• Lesson: The opening promises both conclusion AND elaboration

5. Summary Section

• Draft: "סיכום והכרעה" with 5 items (א-ה)
• Published: "סיכום" with 6 items (א-ו), more specific language
• Lesson: Title is "סיכום", not "סיכום והכרעה" or "סוף דבר"

6. Transition Phrases (new ones discovered)

• "ועל מנת לא לצאת בחסר" - for obiter dicta
• "נציין כי טענות אלו נטענו בלשון רפה" - acknowledging weak claims
• "עינינו הרואות" - summary after long quote
• "נוסיף." - ultra-short transition (one word!)
• "אם כך, לעת הזו" - drawing conclusion from citations
• "למיטב הבנתנו" - cautious position on pending matter
• "נשלים ונציין" - last point before summary

7. New Case Law References (not in our draft)

• ערר (מרכז) 1011-03-25 נגאח עבד אל קאדר (consolidating decision on proprietary claims)
• עע"מ 3975/22 ב. קרן-נכסים (Supreme Court on proprietary feasibility - extensive quote)
• ערר 1071/25 מינץ (own previous decision)
• סעיף 71ב(א)(1) לחוק המקרקעין (majority required for common property changes)

8. Substantive Changes

• Added response from local committee on parking (columns) and tree (sections 29-30)
• Added "על החלטת רשות רישוי מיום 30.11.25" in opening (specific decision date)
• Changed expenses from "no order" to "appellants shall bear expenses"
• Added "ניתנה פה אחד" (unanimous decision)

Applied To

• Updated .claude/skills/legal-decision/SKILL.md - added Section 7 (Discussion methodology), updated Section 4 (transitions), updated Section 1.3 (structure), updated Section 2.1 (paragraph lengths), updated Section 6 (checklist), updated case law references

---

Lessons from בית הכרם 1126/25 + 1141/25

Source

• Final version (Draft 9): 04_Archive/ערר-1126-25-תמא-38-בית-הכרם/החלטה/בית הכרם-טיוטת החלטה-9.pdf
• Our planning: 04_Archive/ערר-1126-25-תמא-38-בית-הכרם/סטטוס-תכנון.md
• Date: March 2026
• Result: Partial acceptance (קבלה חלקית)

What Our Planning Got Right

• Overall result prediction (partial acceptance, same operational directives)
• Identification of key issues (parking, setback lines, preservation)
• Basic structure (background → arguments → discussion → summary)
• Content of parties' arguments section
• Citation through consolidating decision technique (ערר אדלר)

What the Final Version Changed — Critical Gaps

1. Threshold Question Skipped Entirely

• Planning: "שכבה 1 — קריטית: אין זכות ערר לפי ס' 152"
• Final: Zero discussion of right to appeal. Straight to substantive analysis.
• Lesson: The threshold question (6.1 in skill) is a STRATEGIC TOOL, not mandatory. When the case has strong substantive questions (parking, setback, preservation), Dafna prefers engaging with substance over procedural blocking. This also strengthens the decision against judicial review.

2. Concentric Circles Model Not Used

• Planning: 5 defined layers (threshold → merit → parking → setback → specific claims)
• Final: Different structure — context → tension mapping → issue-by-issue analysis → operational conclusions
• Lesson: Concentric circles fit REJECTED appeals (like הכט). For partial acceptance, Dafna uses flexible issue-by-issue analysis. The skill's 6.3 is one tool among several, not THE framework.

3. New Opening Type: "Tension Mapping"

• Final ס' 39: Lists 6 specific tensions in bullet points before analysis begins
• Pattern: "בערר דנן עולות שאלות כיצד והאם..." → bulleted list of tensions → "כל הנקודות לעיל עומדות לפנינו ולשם כך..."
• When: Partial acceptance or cases with multiple complex intersecting issues
• Not in skill: This is a new opening type distinct from "broad context" (rejected) or "direct conclusion" (accepted)

4. "Single Building" Weakens TAMA 38 Interest

• Final ס' 41: "עסקינן בחיזוק בית בודד ועל כן... לא קיים באופן מלא אינטרס חיזוק כזה המצדיק את אישור מלוא הזכויות"
• Not in skill or planning. New analytical pattern: when TAMA 38 applies to a single house (vs. large apartment building), earthquake protection interest is weaker → more cautious approval of rights, especially setback lines and parking.

5. Master Plan as "Shield" Against Ad-Hoc Planning Concern

• Final ס' 42: "קיימת תכנית אב אשר מקלה על בחינת הבקשה... החשש לאישור היתר מכח תכנית 10038 על מגרש בודד ללא ראיה כללית אינו קיים"
• Pattern: When a master plan exists → cite it to validate individual permit → conclusion: permit "integrates with existing comprehensive vision" rather than creating ad-hoc precedent.

6. Depth of Plan Provision Citations

• Planning: Expected general parking analysis
• Final: Extensive direct quotes from plan provisions (6.8(4), 6.8(9), traffic appendix, 5166b) — 400+ words of plan citations with interleaved analysis
• Lesson: For parking/infrastructure issues, Dafna goes very deep into plan provisions with direct quotes, not summaries.

7. Ultra-Minimal Summary for Partial Acceptance

• Planning: ~1,000 words, 8 reasons, expenses, warm closing
• Final: 3 short sections (ס' 84-86) — conclusion + 2 operational directives. No expenses. No warm closing.
• Lesson: In partial acceptance, all reasoning is already in the discussion. Summary = operational directives only.

8. Precedents — Planned vs. Actually Used

Planned but dropped	Added unexpectedly
חנין, נחמיאס (right to appeal)	ערר 1192/18 אילן (preservation + nuisance)
הלכת שפר (deviation from plan)	ערר מובשוביץ 1009-02-24 (urban renewal — ~400 word quote)
ערר כהן (no delaying permit)	ערר ארד 1156/18 (construction nuisance)
עניין שיק (neighborhood change)	ערר זוהר 1169/19 (same topic)

9. New Transition Phrases Discovered

• "הדברים משליכים על שיקול הדעת ב..." — linking finding to conclusion
• "רוצה לומר כי" — alternative phrasing/explanation
• "נוצר מצב בו" — presenting factual situation/problem
• "לכך נוסיף כי" — adding another layer
• "יש אולי להצר על כך ש..." — gentle critical remark
• "עם ההבנה לטענה זו של העוררים, אין בידנו לקבלה" — soft acknowledge-reject

Meta-Lesson

Our skill was "over-indexed" on one case type (הכט = rejected appeal). The concentric circles model, threshold question as mandatory, and warm closing were all patterns from that single case. Beit HaKerem (partial acceptance) reveals that Dafna's approach is more flexible than we captured. We now have two data points — need to distinguish between patterns that are universal vs. result-dependent.

Applied To

• Updated .claude/skills/legal-decision/SKILL.md — added partial acceptance track (7.2, 7.3, 8.4), caveats to 6.1/6.3, new analytical patterns (6.10, 6.11), new golden ratios (3.2), new transition phrases (5.2)

---

Lessons from קרית יערים-1 Structure Building (March 2026)

Source

• Structure draft: 01_Projects/כתיבת החלטות משפטיות/ערר קרית יערים-1/החלטה/החלטה-ערר-1130-25-מבנה.docx
• Reference decisions: בית הכרם (Dafna), ARAR-24-1078-44 (Arieli)
• Date: March 2026

10. "Neutral Background" Rule

• Problem: First draft put detailed 2017 district committee quotes ("נולד חטא", "חריג לסביבתו", "לא בדיוק המקום הזה") in the Background section.
• Chaim's correction: These are parties' arguments disguised as background. The background was "revealing cards" before the parties spoke — effectively summarizing the case before presenting it.
• Rule: Background (Block ו) = objective facts only. Test: "Does this sentence contain a direct quote from a party, or value/judgment words (חריג, חטא, בעייתי)?" If yes → belongs in Claims (Block ז) or Discussion (Block י), not Background. Prior decisions cited as dry fact ("rejected on date X") — reasoning, quotes, and interpretations appear only in claims/discussion.
• Applied to: SKILL.md section 11.2 Block ו, added "⚠️ כלל רקע ניטרלי"

11. New 12-Block Decision Structure

• Created formal 12-block structure based on analysis of Beit HaKerem + Arieli decisions
• Added mandatory "pre-discussion draft" step (Block 12 in SKILL.md)
• Created create-decision-structure.cjs script for generating structure DOCX
• Key innovation from Arieli: "ההליכים בפני ועדת הערר" as separate section (Block ח)
• "Judge Test": every block written as if administrative court judge reads cold

---

Lessons from Systematic Corpus Analysis (24 decisions, April 2026)

Source

• All 24 proofread decisions in /data/training/proofread/
• Full analysis: docs/corpus-analysis.md
• Date: April 2026

12. System Learned Style but Not Substantive Content

• Problem: Dafna reviewed Kiryat Yearim draft and noted missing planning discussion in block-yod
• Root cause: The block-yod prompt taught CREAC methodology and "answer all claims" but never said "in licensing cases, include comprehensive planning discussion"
• Fix: Content checklists added to lessons.py (CONTENT_CHECKLISTS), injected into block-yod prompt via {content_checklist}
• Applied to: lessons.py, block_writer.py

13. Corpus Composition — All Licensing, No Betterment Levy

• All 24 training decisions are licensing/construction (1xxx)
• Zero betterment levy (8xxx) decisions in corpus
• Not a current priority gap — focusing on licensing first

14. Planning Discussion Patterns in Licensing Decisions

• Always present when the appeal reaches substantive planning questions
• Never present when the appeal is purely jurisdictional or property-based
• Structure: broad planning context → direct plan provision citations (200-600 words) → application to specific case → planning conclusion
• Deepest planning: פרומר (pure plan interpretation), לבנון (height/building appendix), בית הכרם (multi-plan TAMA 38)
• No planning: טלי-אביב (property only), גבאי (jurisdiction only)

15. Five Appeal Subtypes Identified (Not Just Three)

Licensing appeals are not homogeneous — the discussion structure varies significantly:

1. Substantive licensing — full planning discussion + legal analysis (majority of cases)
2. Threshold/jurisdiction — legal analysis only, no planning
3. Property-focused — תימוכין קנייניים, minimal planning
4. TAMA 38 — balancing public interest + planning + neighbor impact
5. Deviant use (שימוש חורג) — deep plan interpretation across multiple plans

16. Chair Feedback System Established

• DB table chair_feedback records Dafna's comments on drafts
• Categories: missing_content, wrong_tone, wrong_structure, factual_error, style, other
• MCP tools + UI page for recording and reviewing feedback
• First entry: Kiryat Yearim — missing planning discussion (2026-04-12)

---

Lessons from External Expertise Research (April 2026)

Source

• Federal Judicial Center, Judicial Writing Manual (1991, 2nd ed. 2020)
• Bryan Garner, Legal Writing in Plain English (2001)
• Scalia & Garner, Making Your Case: The Art of Persuading Judges (2008)
• Richard Posner, How Judges Think (2008)
• Full texts stored in: docs/sources/

17. Methodology Document Created — Separating "How to Think" from "How to Write"

Problem: The system knew Dafna's STYLE (SKILL.md) and WHAT TOPICS to cover (content checklists), but had no formal methodology for HOW TO REASON through a decision — the analytical stages, when to balance, how to structure arguments, how to handle counterarguments.

Fix: Created docs/decision-methodology.md — a standalone analytical methodology document based on synthesis of all four external sources. 3,400 words, 12 sections, 10 guiding principles. Covers: pre-analysis, threshold questions, issue ordering, syllogistic structure (CREAC), balancing/proportionality, claims handling (steel-man, bundling), quotation technique (sandwich), factual findings vs. legal conclusions, disposition, writing techniques, analogy/precedent, editing checklist.

Key principle: Methodology is UNIVERSAL — it teaches how to think about any quasi-judicial decision. It does not contain case-specific content (parking, building lines, etc.). Case-specific content stays in the content checklists.

Applied to:

• docs/decision-methodology.md — new document
• lessons.py — new function get_methodology_summary() injected into block-yod prompt
• block_writer.py — new {methodology_guidance} placeholder in block-yod prompt
• .claude/agents/legal-writer.md — restructured block-yod workflow to follow methodology stages
• .claude/agents/legal-qa.md — new check #7 (methodology compliance)

18. "Answer All Claims" Made Flexible

Problem: The block-yod prompt hardcoded "answer every claim individually" and the QA check enforced it. But Dafna sometimes bundles weak claims, skips irrelevant ones, and focuses on what matters.

Fix:

• Block-yod prompt changed from "חובה לענות על כל אחת" to flexible handling: address substantive claims; bundle [bundle]; skip [skip]
• Chair can mark claims in chair_directions as bundle or skip
• QA check #3 updated to respect these markings
• Methodology teaches WHEN to address individually vs. bundle vs. skip (methodology §ו)

19. Source Library Established

Downloaded and converted to text 5 authoritative sources for the methodology:

• docs/sources/fjc-judicial-writing-manual-1991.txt (13,567 words)
• docs/sources/fjc-judicial-writing-manual-2nd-ed-2020.txt (15,912 words)
• docs/sources/garner-legal-writing-plain-english.txt (97,475 words)
• docs/sources/posner-how-judges-think.txt (156,789 words)
• docs/sources/scalia-garner-making-your-case.txt (54,683 words) Total: ~340,000 words of source material.

Intermediate extraction documents also saved:

• docs/fjc-principles-extraction.md — 38 principles from FJC
• docs/garner-methodology-extraction.md — ~50 principles from Garner/Scalia