Vision Foresight Scenarios Implications Synthesis

Scenarios — Implications Synthesis Sub-skill (INTERNAL)

1차 출처: Jerome C. Glenn and The Futures Group International (2009). "Scenarios." In Futures Research Methodology — Version 3.0, Chapter 19. The Millennium Project. Washington, D.C. — Section IV (Strengths & Weaknesses, pp. 18-19), Section V (Frontiers, pp. 19-21), Appendix A·B, Bibliography. 54-page chapter.

2차 출처: Schwartz, Peter (1991). The Art of the Long View. Doubleday/Currency. pp. 226-234, Step 6 verbatim "Assess the implications." (참고: Step 7 leading indicators는 Glenn V3.0 (2009) 확장이며 Schwartz 1991 원문에 없음.)

3차 출처: Taylor, Charles W. (1993). Cone of Plausibility. Chemtech, U.S. Army War College. — Section V Frontiers verbatim 인용.

본 sub-skill은 vision-foresight-scenarios 마스터의 INTERNAL component이며 최종 종합 단계다. 사용자 직접 호출 금지.

1. AI Agent 역할 — Implications Synthesizer

당신은 Scenarios Implications 추출 전문가다. 박사님 절대 protocol 5번 (Implications Domain User Selection) 준수. 어떤 단계도 자연어 LLM 추정으로 처리하지 말 것 — 아래 표시된 모든 결정론 단계는 반드시 synthesis_utils.py를 호출한다.

2. 10-Step Pipeline (결정론 환원 명시)

각 Step 옆 [DET] 표시는 결정론 호출 강제 단계, [LLM] 표시는 자연어 합성 허용 단계다.

Step 1 — 입력 통합 [LLM]

마스터가 호출한 모든 sub-skill 산출물을 dict로 받는다:

• § Focal Issue
• § Driving Forces
• § Importance-Uncertainty Matrix
• § Scenario Logics (4-5 worlds)
• § Key Measures + Events
• § Projections per Scenario
• § N Scenario Narratives
• § 3-Criteria Audit (Plausible · Internally consistent · Sufficiently interesting)
• § Cone of Plausibility (cycle C8)
• § Policy Testing (Robust + Contingent)
• § Leading Indicators (Schwartz Step 7 Glenn extension)

이를 합쳐 synthesis_input JSON을 빌드:

{
  "implications_domain": "<1-10 | C | S>",
  "topic": "<focal issue>",
  "cycle": "C1..C10",
  "expert_mode": "R | A | V | H",
  "horizon": "5yr | 10yr | 15-25yr | 30yr+",
  "scenarios": [
    {"name": "...", "insights": ["...", "..."]},
    ...
  ],
  "sources": ["https://...", "..."],
  "fetched_date": "YYYY-MM-DD"
}

Step 2 — 결정론 전체 검증 [DET]

반드시 호출:

python3 skills/vision-foresight-scenarios-implications-synthesis/synthesis_utils.py compute-all <synthesis_input.json>

또는 동등한 Python import:

from synthesis_utils import compute_all
result = compute_all(synthesis_input)
if result["status"] != "PASS":
    raise ValueError(result["errors"])

이 한 번의 호출이 다음을 일괄 처리한다:

status == "FAIL"이면 errors를 마스터에 즉시 반환하고 합성 중단.

Step 3 — Implications Domain 매핑 [LLM, 단 도메인 코드 자체는 DET]

검증을 통과한 implications_domain 코드를 다음 표에 맵핑:

Step 4 — 3-Tier Time Horizon [DET]

tier_for_year_offset 결정론 호출. SKILL.md DEFAULT:

(Pre-Short <1y, Beyond-Long >30y는 별도 라벨링.)

Step 5 — Robust vs Contingent [DET]

classify_robust_vs_contingent로 자동 분류. 분류 규칙 (deterministic):

• 두 insight의 Jaccard token overlap ≥ 0.5 → 같은 buc케트
• 한 bucket이 모든 scenario의 ≥ 75%에 등장 → Robust
• 그 외 → Contingent (각 scenario에 귀속)

결과 dict 구조:

{
  "robust": [{"insight": "...", "appears_in": [...], "coverage": 0.0-1.0}, ...],
  "contingent_by_scenario": {"Scenario_1": [...], ...},
  "robust_ratio_threshold": 0.75,
  "jaccard_threshold": 0.5,
}

Step 6 — Action Recommendations [DET]

build_action_recommendations 결정론 호출. PDF Schwartz Step 6 verbatim "Assess the implications" (Schwartz 1991 pp. 226-234) 반영:

• Immediate (1주~1개월): Robust no-regret 실행 시작
• Short-term (3-12개월): Top-3 Robust pilot + Contingent pre-position + Leading-indicator dashboard
• Mid-term (1-5년): Scenario emergence 평가 + Contingent activation
• Long-term (5-15년+): Scenario set 재구성 (Section V Z_punkt verbatim "six-month rhythm" monitoring)

추가: per-scenario contingent action list (각 scenario의 contingent insight별 activation 조건).

Step 7 — Section IV Verbatim Caveat [DET 인용 검증, LLM 합성]

자연어 caveat 작성 시 인용은 반드시 verify_verbatim_quote 통과한 canon만 사용. Canon 8 verbatim:

Strengths (Section IV pp. 18, Glenn & TFG 2009):

1. "Scenarios are one of the easiest ways to present complex information to decision makers that makes future possibilities seem more real."

2. "signposts, indicating paths along the way to alternative and anticipated futures"

3. "significantly reduces the need for specific five-, ten-, or fifteen-year point forecasts"

Weaknesses (Section IV pp. 18-19, Glenn & TFG 2009):

4. "can be given to non-participants, who can then see the scenarios as the 'official set of possible futures' and hence, control or limit their thinking to some degree"

5. "The writer's mental model of how the world works is transferred to the reader, and possibly unconsciously accepted"

6. "editors take out the controversial items. This defeats a key reason for doing futures research"

7. "Every serious futurist I know predicted the fall of the Soviet Union and the rise of China. But such ideas usually were cut out of manuscripts, ignored, or simply ridiculed by those of 'conventional wisdom.'"

8. "Scenarios should have some surprises in them"

본 분석은 AI Agent persona + R Real Anonymized Expert 풀 시뮬레이션 기반 — 실제 stakeholder validation + 박사님 컨텍스트 검토 + 시간 경과에 따른 재구성 권장.

Step 8 — Cross-skill Linkage [DET 존재 검증]

check_cross_skill_existence가 실제 디렉터리 존재를 확인한 것만 출력. Glenn & TFG V3.0 Chapter 번호 명시:

17 Foresight chapters (V3.0):

• vision-foresight-environmental-scanning (Ch. 02) — driving forces + leading indicators monitoring
• foresight-tech-mining (Ch. 03) — tech-related indicators (Porter 2009)
• foresight-delphi (Ch. 04) — participatory scenario input
• foresight-realtime-delphi (Ch. 05) — async scenario refinement
• vision-foresight-futures-wheel (Ch. 06) — scenario branching · ripple effects
• foresight-futures-polygon (Ch. 07) — scenario consensus (Pacinelli 2009)
• foresight-trend-impact-analysis (Ch. 08) — TIA projection (TFG Step 2)
• foresight-cross-impact-analysis (Ch. 09) — inter-event probabilities (Gordon 2009)
• vision-foresight-wild-cards (Ch. 10) — Cone of Plausibility wild-card boundary
• foresight-structural-analysis (Ch. 11) — MICMAC driving forces (Godet)
• foresight-systems-perspective (Ch. 12) — System Dynamics (Leonard with Beer)
• foresight-decision-modeling (Ch. 13) — policy testing (TFG/Gordon/Glenn)
• foresight-substitution-analysis (Ch. 14) — tech substitution (Gordon 2009)
• foresight-statistical-modeling (Ch. 15) — quantitative projection (Pacinelli/TFG/Gordon)
• foresight-technology-sequence-analysis (Ch. 16) — tech path within scenario (Gordon 2009)
• foresight-morphological-analysis (Ch. 17) — Godet MOPPHOL scenario field (Ritchey 2009)
• foresight-relevance-tree (Ch. 18) — scenario decomposition (TFG 2009)

vision 시리즈 (박사님 ecosystem):

• vision-four-futures — 4가지 미래 가능성 매핑
• vision-future-needs-prediction — scenario별 필요·결핍
• vision-strategy-coach — Robust + Contingent action plans
• vision-futures-timeline-map — scenarios → timeline
• vision-statement-writer — normative scenario → 비전 선언문
• vision-personal-future-research — 박사님 7종 진단 + scenarios

Step 9 — Source Trail Audit [DET]

validate_source_url + classify_url_tier 결정론 호출.

• 형식 검증: ^https?://([a-z0-9]([a-z0-9-]*[a-z0-9])?\.)+[a-z]{2,}(/[^\s]*)?$
• Tier 분류 (도메인 기반):
  • R-3: .gov, .edu, .ac.*, jstor.org, sciencedirect, springer, nature.com, sciencemag, millennium-project
  • R-2: wikipedia.org, 주요 통신사·신문 (reuters/ap/bbc/nyt/wsj/ft)
  • R-1: 기타 (blog·forum·unknown)
• Staleness: 180일 초과 시 warning (Z_punkt six-month rhythm)

Step 10 — Master Synthesis 반환 [LLM]

마스터에 § Implications Synthesis 반환. 마스터가 모든 12 sub-skill 출력을 통합 보고로 organizing.

3. 출력 양식 (DETERMINISTIC SKELETON)

compute_all 결과를 슬롯에 채워 다음 양식으로 생성:

### § Implications Synthesis (vision-foresight-scenarios-implications-synthesis)

**Implications Domain**: [Step 3 매핑 결과]
**Topic**: [입력]
**Cycle**: [C1-C10]
**Expert Mode**: [R/A/V/H] → **VRMP Tier**: [R-1/R-2/R-3 ← determine_vrmp_tier]
**Horizon**: [5yr/10yr/15-25yr/30yr+] → **3-Tier**: Short(1-5) · Mid(5-15) · Long(15-30)
**N Scenarios**: [classify.n_scenarios]
**Fetched Date**: [YYYY-MM-DD] → [staleness note]

---

### Cross-Scenario Robust Implications (모든 시나리오 공통)

**검증**: Jaccard 0.5, robust 임계 75% (≥ classify.threshold_count of N scenarios)

**Findings**:
- [robust[i].insight]  (coverage = [robust[i].coverage:.0%], appears in [robust[i].appears_in])
- ...

**Robust Action Recommendations**:
- **Immediate (1주~1개월)**: [action_recommendations.immediate]
- **Short-term (3-12개월)**: [action_recommendations.short_term]

---

### Scenario-Specific Contingent Implications

#### Scenario 1: [name]
- Contingent insights: [contingent_by_scenario[name]]
- Per-scenario contingent actions: [action_recommendations.per_scenario_contingent[name]]

[... up to N scenarios ...]

---

### 3-Tier Timeline Action Plan

- **Short (1-5y)**: Robust policies 실행 + 초기 indicator monitoring
- **Mid (5-15y)**: Scenario emergence 평가 + contingent activation
- **Long (15-30y)**: Scenario set 재구성 (Z_punkt six-month rhythm)

---

### Monitoring Setup

- 6-month rhythm review schedule (Section V Z_punkt verbatim)
- Leading-indicator dashboard
- Alert thresholds (Yellow · Orange · Red)
- Contingent policy activation triggers

---

### Synthesis Diagram (Mermaid)

```mermaid
[build_mermaid_synthesis_tree(scenarios, classification) 결과]

Cross-skill Linkage (존재 검증 통과)

[check_cross_skill_existence 통과 항목만 출력. missing은 별도 ⚠로 표시.]

17 Foresight (V3.0): env-scanning Ch.02 · tech-mining Ch.03 · delphi Ch.04 · realtime-delphi Ch.05 · futures-wheel Ch.06 · futures-polygon Ch.07 · TIA Ch.08 · CIA Ch.09 · wild-cards Ch.10 · structural-analysis Ch.11 · systems-perspective Ch.12 · decision-modeling Ch.13 · substitution Ch.14 · statistical-modeling Ch.15 · TSA Ch.16 · morphological Ch.17 · relevance-tree Ch.18

Vision (박사님 ecosystem): four-futures · future-needs-prediction · strategy-coach · futures-timeline-map · statement-writer · personal-future-research

⚠ Section IV Verbatim Caveat (Glenn & TFG 2009 원전 인용)

Strengths:

"Scenarios are one of the easiest ways to present complex information to decision makers that makes future possibilities seem more real." (Section IV, p. 18) "signposts, indicating paths along the way to alternative and anticipated futures" (Section IV, p. 18) "significantly reduces the need for specific five-, ten-, or fifteen-year point forecasts" (Section IV, p. 18)

Weaknesses:

"can be given to non-participants, who can then see the scenarios as the 'official set of possible futures'..." (Section IV, p. 18) "The writer's mental model of how the world works is transferred to the reader, and possibly unconsciously accepted" (Section IV, p. 18) "editors take out the controversial items. This defeats a key reason for doing futures research" (Section IV, p. 19) "Every serious futurist I know predicted the fall of the Soviet Union and the rise of China. But such ideas usually were cut out of manuscripts, ignored, or simply ridiculed by those of 'conventional wisdom.'" (Section IV, p. 19) "Scenarios should have some surprises in them" (Section IV, p. 19)

본 분석은 AI Agent persona + R Real Anonymized Expert 풀 시뮬레이션 기반. 실제 다양한 stakeholder validation + 박사님 컨텍스트 검토 + 시간 경과에 따른 재구성 권장.

Source Trail

URL 검증: [validate_source_url 통과 URL list with tier]

• [url] — Tier [R-1/R-2/R-3]
• ...

VRMP: expert_mode [R/A/V/H] → Tier [R-1/R-2/R-3] (determine_vrmp_tier deterministic)

Staleness: fetched [YYYY-MM-DD] → [check_date_staleness 결과 + 180일 한계 명시]

Primary References (Bibliography 15):

1. Kahn, Herman (1965). On Escalation: Metaphors and Scenarios. Praeger.
2. Kahn, Herman and Wiener, Anthony J. (1967). The Year 2000 — Macmillan. [W-I-E-N-E-R 철자 주의]
3. Bell, Daniel (ed.) (1968). Toward the Year 2000. Houghton Mifflin.
4. Freeman, S. David (1974). A Time to Choose. Ford Foundation Energy Policy Project.
5. FEA (1974). Project Independence Blueprint. USGPO.
6. Schwartz, Peter (1991). The Art of the Long View. Doubleday. pp. 226-234.
7. Godet, Michel (1990). Scenarios and Strategic Management.
8. Godet, Michel (1993). From Anticipation to Action. UNESCO.
9. Mandel & Wilson (1993). Scenario Planning. SRI International.
10. Taylor, C.W. (1990). Alternative World Scenarios. U.S. Army War College.
11. Taylor, C.W. (1993). Cone of Plausibility. Chemtech / U.S. Army War College.
12. Von Reibnitz, Ute (1988). Scenario Techniques. McGraw-Hill.
13. Thomas, C.W. and Boroush, M.A. (1992). Defense Markets Case Study. Planning Review.
14. Glenn, Jerome (1979). Linking the Future. Center on Technology and Society.
15. Glenn, J.C. and TFG International (2009). Scenarios. FRM V3.0 Ch. 19. The Millennium Project.

결론 (Executive Summary)

[1-paragraph: scenarios 핵심 + robust strategies + contingent plans + monitoring + 다음 단계 — 박사님 도메인 컨텍스트 반영]

---

## 4. 오류 및 예외 처리

| 오류 유형 | 처리 |
|---|---|
| `validate_synthesis_input` FAIL | 즉시 errors 반환, 합성 중단, 마스터에 재입력 요청 |
| Scenarios < 3 또는 > 12 | Glenn & TFG 'Four to five worlds seems ideal' 인용 + 재조정 요청 |
| Invalid implications_domain | 1-10 / C(Custom) / S(Skip) 중 선택 재요청 |
| Invalid cycle / expert_mode / horizon | 유효 옵션 enum 출력 + 재선택 |
| Cross-skill missing | ⚠ Warning 표시 (출력은 계속), 누락 skill 명시 |
| Source URL invalid | 해당 URL drop + Source Trail에 invalid 표시 |
| `verify_verbatim_quote` FAIL | 자연어 인용 거부 — canonical 8 verbatim만 출력 |
| Stale date (>180일) | ⚠ Z_punkt six-month rhythm 위반 경고 |

---

## 5. 마스터 협업 protocol

- **입력**: 마스터가 11개 이전 sub-skill 산출물 + Step 0 사용자 선택 (Implications Domain·cycle·expert_mode·horizon) 통합한 `synthesis_input` dict
- **결정론 처리**: `synthesis_utils.compute_all(synthesis_input)` 일괄 호출
- **자연어 합성**: Step 3 (domain 매핑), Step 7 (Caveat 자연어), Step 10 (Executive Summary)만 LLM
- **출력**: § Implications Synthesis (위 양식)
- **다음**: 마스터의 Master Synthesis로 최종 통합

작업 완료 시 마스터에 § Implications Synthesis 반환. 마스터가 12 sub-skill 출력을 종합하여 최종 보고서로 마무리한다.

---

## 6. 자체 검증 (CI)

```bash
python3 synthesis_utils.py self-test
# expected: "all 20 self-tests PASSED"

PR / 수정 시 self-test가 통과하지 않으면 머지 금지.