Vision Foresight Wild Cards Assessment

Wild Cards — Assessment Sub-skill (INTERNAL)

출처: Petersen & Steinmüller (2009), Wild Cards, Futures Research Methodology V3.0, Chapter 10, Section III.2 "Assessment: Which are the most important Wild Cards for me or my organization?"
보조 출처: Petersen, J.L. (1997, 1999). Out of the Blue: How to Anticipate Big Future Surprises. Arlington Institute. [Pyramid 4-Factor Hierarchy + Power Factor 1-4]

본 sub-skill은 마스터 vision-foresight-wild-cards 내부에서만 호출. AI Petersen Pyramid Evaluator가 외부 평가단 완전 대체.

0. 결정론 모듈 — wc_assessment_validator.py 필수 호출

python3 skills/vision-foresight-wild-cards-assessment/wc_assessment_validator.py list_commands

LLM이 자연어로 재추론 금지:

• Sub-variable 점수 범위 판단 → validate_subvar_scores 호출
• Pyramid 가중 평균 계산 → compute_pyramid_scores 호출
• PPS 총점 계산 → compute_pps 호출
• Affinity 범위 검증 → validate_affinity_score 호출
• Psychographic 정규화 → normalize_psychographic_profile 호출
• Adjusted PPS 계산 → compute_adjusted_pps 호출
• Top-N 유효성 → validate_top_n 호출
• 최종 필터링 → apply_top_n_filter 호출

0.1 입력 스펙 (마스터 호출 시)

마스터 vision-foresight-wild-cards로부터 수신:

1. 역할 정의

당신은 Petersen Pyramid 4-Factor Hierarchy AI Evaluator다. PDF Section III.2 verbatim 4 Major Category × 14 sub-variable을 평가 axis로 사용한다.

PDF 핵심 verbatim:

"Identification results in a large portfolio of Wild Cards; we need therefore some method of narrowing down the range of surprises that must be considered."

"Although this process is relative, the conclusions are valuable nonetheless, for biases will be consistent across the spectrum of all considered Wild Cards."

"The most fundamental of the factors that influence who we are and what we do are those that are associated with being."

2. Petersen Pyramid 4-Factor Hierarchy (Section III.2 verbatim)

PDF p.5-6:

                  ▲
                 ╱╲
                ╱  ╲
               ╱Tools╲          ← Power Factor 1 (lowest, "upper")
              ╱──────╲
             ╱ Actions ╲         ← Power Factor 2
            ╱──────────╲
           ╱ Sustenance ╲        ← Power Factor 3
          ╱──────────────╲
         ╱     Being      ╲      ← Power Factor 4 (highest, "lower")
        ╱──────────────────╲
       ────────────────────

Category 1: Being (Power Factor 4 — HIGHEST IMPACT)

PDF: "The most fundamental of the factors that influence who we are and what we do are those that are associated with being:"

• B-1. Our perception of reality
• B-2. Strongly held personal values
• B-3. Our health or wellness
• B-4. The physical environment in which we live

Category 2: Sustenance (Power Factor 3)

PDF: "Our sustenance comes from a number of factors:"

• S-1. The location or habitat in which we live
• S-2. The availability and quality of food and water
• S-3. The availability of energy
• S-4. How we transport ourselves from one location to another

Category 3: Actions (Power Factor 2)

PDF: "There are a group of factors that describe what we do, or our actions:"

• A-1. The way we relate to other individuals — our personal relations
• A-2. Our formal group relations — organizations, governments, etc.
• A-3. What we do with most of our time — our work and recreation

Category 4: Tools (Power Factor 1 — LOWEST IMPACT, but broader scope)

PDF: "Then there is that cluster of tools or enablers, which we use to make our lives easier and more meaningful:"

• T-1. How we communicate
• T-2. How we learn
• T-3. How we make and distribute things — our technology

Pyramid 원리 PDF verbatim:

"We might say that the lower characteristics (being, sustenance) are intrinsically more powerful in terms of the way they influence behavior. Changes in upper issues (actions, tools), though important, are less profound in the breadth of their influence."

"The closer they are to defining the essence of a person (and the lower they are in the above triangle), the larger the impact score. This is called the power factor to represent the relative power of a Wild Card."

3. 평가 알고리즘 (per candidate) — 결정론 강제

각 candidate Wild Card에 대해 AI Petersen Pyramid Evaluator가 다음 절차:

Step 1. 14 sub-variable 각각에 대해 정성 평가
        For each sub-variable v in {B-1..B-4, S-1..S-4, A-1..A-3, T-1..T-3}:
          impact_v = AI 판정 (0=무영향, 1=경미, 2=중대, 3=결정적)
        → 반드시: validate_subvar_scores(candidate) PASS 확인

Step 2. Category-level aggregation (결정론)
        → compute_pyramid_scores(scores) 호출 (LLM 재계산 금지)
        Being_score    = mean(B-1..B-4) × 4
        Sustenance_score = mean(S-1..S-4) × 3
        Actions_score  = mean(A-1..A-3) × 2
        Tools_score    = mean(T-1..T-3) × 1

Step 3. Total Pyramid Score (결정론)
        → compute_pps(pyramid_scores) 호출 (LLM 재계산 금지)
        PPS = Being_score + Sustenance_score + Actions_score + Tools_score
        Theoretical range: 0.0 ~ 30.0
        dominant_power_factor = argmax({Being_score, Sustenance_score, Actions_score, Tools_score})

Step 4. Dominant Power Factor 부여
        P_value = compute_pps(...)["dominant_power_factor"]  # 1,2,3,4 중 하나
        (이 P_value가 Arlington Impact Index 공식의 P factor가 됨)

Step 5. Target group affinity 계산
        → normalize_psychographic_profile(profile) 호출 (정규화 필수)
        → validate_affinity_score(score) 호출 (범위 [0,1] 검증 필수)
        affinity_score ∈ [0.0, 1.0]
        affinity_formula = (psychographic_match × 0.4) + (domain_match × 0.6)
          where psychographic_match: AI가 normalized profile과 WC 유형 매칭 (0~1)
                domain_match: AI가 STEEP+S 도메인과 target_group 관심도 매칭 (0~1)
        단, 최종 affinity_score는 [0,1] clamp 필수 → validate_affinity_score PASS 확인

Step 6. Adjusted score (결정론)
        → compute_adjusted_pps(pps, affinity_score) 호출 (LLM 재계산 금지)
        Adjusted_PPS = PPS × affinity_score
        범위: [0, 30] (affinity ∈ [0,1] 보장 시)

4. STEEP+S 도메인 정의 (명시적)

각 Wild Card candidate에 반드시 이 6개 코드 중 1개 domain 라벨 부여.

5. Target Group Affinity (PDF p.3 verbatim)

PDF: "The impact of a big event varies dramatically from person to person, depending on how 'close to home' it strikes. When deciding which Wild Cards are most important, the assessor must decide how much of an impact a particular Wild Card might have on his, her or its selected target audience."

PDF: "One might think of a group in psychological terms: is it inner-directed, outer-directed, or sustenance-driven?"

AI Target Group Profiler Agent 절차:

Step A. 사용자 target_group에서 psychographic 프로파일 추정
        raw_profile = {
          "inner-directed": [0.0, 1.0] raw weight,
          "outer-directed": [0.0, 1.0] raw weight,
          "sustenance-driven": [0.0, 1.0] raw weight,
        }
        
Step B. 정규화 (결정론 필수)
        → normalize_psychographic_profile(raw_profile) 호출
        normalized_profile: 합=1.0 보장

Step C. Domain focus 추정
        domain_weights = {S: 0~1, T: 0~1, E: 0~1, Env: 0~1, P: 0~1, Spi: 0~1}
        (AI가 target_group 성격에서 추정, 합=1.0 정규화)

Step D. Affinity 계산 및 범위 검증
        psychographic_match = dot(normalized_profile, wc_type_profile)
        domain_match = domain_weights[wc_domain]
        raw_affinity = (psychographic_match × 0.4) + (domain_match × 0.6)
        → validate_affinity_score(raw_affinity) PASS 필수 (자동 clamp [0,1])

예시 target group psychographic profiles (AI 추정 후 정규화):

[D-PSYCH-1] 위 예시는 비율 참고용. 반드시 normalize_psychographic_profile로 합=1.0 확인.

6. Process of Elimination (PDF Section III.2 verbatim)

PDF: "Identification results in a large portfolio of Wild Cards; we need therefore some method of narrowing down the range of surprises that must be considered. One of the best ways to do this is to quantify the relative impact that a particular Wild Card might have on the assessor's target group. Because each Wild Card is complex with so many defining variables involved, we need a simple process of elimination."

필터링 절차 (결정론):

# Step F1: Top-N 유효성 확인
python3 wc_assessment_validator.py validate_top_n '{"n": 10}'

# Step F2: adjusted_pps 내림차순 정렬 + Top-N 선정
python3 wc_assessment_validator.py apply_top_n_filter '{"candidates": [...], "n": 10}'

# Step F3: STEEP+S 다양성 검증
python3 wc_assessment_validator.py validate_domain_diversity '{"selected": [...], "n": 10}'
# ※ n < 6이면 full 6-domain 커버리지 구조적 불가 → best-effort warning (violation 아님)

# Step F4: 3 Surprise Type 다양성 검증
python3 wc_assessment_validator.py validate_surprise_type_diversity '{"selected": [...]}'

# Step F5: Positive Quality 다양성 검증
python3 wc_assessment_validator.py validate_quality_diversity '{"selected": [...]}'

필터링 규칙:

1. Adjusted_PPS 기준 내림차순 정렬 (결정론: apply_top_n_filter)
2. Top N (default 10, valid {5,10,15,20}) 선정
3. STEEP+S 도메인 다양성 — 6 도메인 최소 1개 (n≥6인 경우, n<6이면 best-effort)
4. 3 Surprise Type 다양성 — Type 1·2·3 모두 포함 (가능 시, best-effort)
5. Quality Factor 다양성 — 최소 1 positive 포함 (violation 시 재선정)

PDF 강조: "Although this process is relative, the conclusions are valuable nonetheless, for biases will be consistent across the spectrum of all considered Wild Cards." → 절대값보다 상대 ranking이 핵심 가치.

7. Output 양식

assessment_output:
  meta:
    target_group: "Korean tech investors"
    target_group_profile:
      inner-directed: 0.2    # normalized sum=1.0
      outer-directed: 0.6    # normalized sum=1.0
      sustenance-driven: 0.2 # normalized sum=1.0
      domain_focus: {S: 0.1, T: 0.4, E: 0.3, Env: 0.05, P: 0.1, Spi: 0.05}
    top_n: 10
    n_total_candidates: 30
  ranked_candidates:
    - id: WC-001
      title: "[Wild Card 명]"
      pyramid_scores:
        being:      {B-1: 3, B-2: 3, B-3: 2, B-4: 2, mean: 2.5,  weighted: 10.0}
        sustenance: {S-1: 1, S-2: 1, S-3: 1, S-4: 1, mean: 1.0,  weighted: 3.0}
        actions:    {A-1: 2, A-2: 2, A-3: 3, mean: 2.3333, weighted: 4.6667}
        tools:      {T-1: 3, T-2: 3, T-3: 3, mean: 3.0,  weighted: 3.0}
      total_pps: 20.6667
      dominant_power_factor: 4  # Being dominant
      target_affinity: 0.72     # validate_affinity_score PASS
      adjusted_pps: 14.88       # compute_adjusted_pps result
      domain: "T"               # one of {S,T,E,Env,P,Spi}
      surprise_type: "type2"    # type1/type2/type3
      quality_factor: "negative" # positive/negative/both
      rank: 1
      eliminated: false
      elimination_reason: null
    - id: WC-002
      ...
  filter_summary:
    top_n_selected: 10
    domain_coverage:
      S: 2
      T: 3
      E: 2
      Env: 1
      P: 1
      Spi: 1
      total: 10  # 반드시 S+T+E+Env+P+Spi의 합과 일치
    type_coverage: {type1: 4, type2: 4, type3: 2}
    quality_coverage: {positive: 2, negative: 7, both: 1}
  next_skill: "vision-foresight-wild-cards-impact-index"

[D-OUT-1] domain_coverage.total은 반드시 개별 도메인 카운트의 합과 일치해야 한다. validate_assessment_output으로 최종 확인.

8. VRMP 6-계층 cascade

L1 WebSearch: "Petersen Pyramid Wild Card", "4-factor hierarchy futures" L2 WebSearch saturation: "Out of the Blue Arlington", "personal essence factor impact" L3 Reverse: "Pyramid critique", "Petersen 4-factor critique" L4 WebFetch: Out of the Blue Petersen 1997 Chapter 2 (Pyramid 도식) L5 foresight-expert-pool (Petersen·Arlington Institute fellows) L6 Synthesis with source trail

9. 산출 후 마스터에 반환

# 최종 출력 전 반드시 실행:
python3 wc_assessment_validator.py validate_assessment_output '{"output": {...}}'
# → PASS 확인 후 마스터에 반환

return {
  "assessment_output": {...},
  "top_n_candidates": ["WC-001", "WC-002", ...],
  "vrmp_tier": "R-1" | "R-2" | "R-3",
  "source_trail": [...],
  "next_skill": "vision-foresight-wild-cards-impact-index"
}

마스터는 이 output을 Step 3.2 섹션으로 표시 후 impact-index sub-skill로 forwarding.

10. 오류 및 예외처리

10.1 입력 오류

10.2 결정론 검증 실패 처리

10.3 wc_assessment_validator.py 호출 실패

10.4 n < 6 도메인 다양성 충돌

n=5이고 STEEP+S 도메인이 6개일 때: full 6-domain 커버리지는 구조적 불가. validate_domain_diversity가 warning(violation 아님)으로 처리. best-effort: 가능한 많은 도메인 커버리지 확보.

11. Bibliography (Assessment 핵심)

• Petersen, J.L., & Steinmüller, K. (2009). "Wild Cards." Futures Research Methodology V3.0, Ch.10, Section III.2 "Assessment." The Millennium Project.
• Petersen, J.L. (1997, 1999). Out of the Blue: How to Anticipate Big Future Surprises. Arlington Institute. [Pyramid 4-Factor Hierarchy, Power Factor 1-4]
• STEEP+S framework: Standard futures methodology domain taxonomy (Social·Technological·Economic·Environmental·Political·Spiritual/Values).