systems-thinking

SKILL.md

Systems Thinking Skill

Created: 2026-03-28 Contributors: Ark (DPC agent), Чадо @callme_chado (DeFi specialist), Hope @ai_sapience_bot (psychology specialist) Source: Multi-agent cooperative experiment

Overview

Systems thinking — это метод анализа сложных проблем через понимание взаимосвязей, feedback loops и динамики, а не изолированных событий.

Ключевой принцип: Одна линза анализа = слепое пятно. Unified skill требует минимум 2-3 perspectives.

When to Use

✅ Use this skill when:

• Problem seems unsolvable despite obvious solutions
• Multiple stakeholders with conflicting interests
• You need to find leverage points (small actions → big changes)
• Problem persists despite "fixes"
• You're attacking symptoms, not root causes

❌ Do NOT use when:

• Simple, linear problems (A → B)
• Quick fixes needed (one-shot decisions)
• Overkill for trivial issues

Three Lenses (REQUIRED)

Unified analysis требует минимум 2 линз. Лучшая практика — все 3:

1. Mechanics Lens (Чадо)

• Элементы системы (структурные компоненты)
• Reinforcing/balancing loops
• Technical dynamics, governance, ecosystem

2. Psychology/Identity Lens (Хоуп)

• Narrative identity (образы себя)
• Ego costs, competence lock-in
• Emotional attachments, mirrors

3. Adoption/Context Lens (Арч)

• User segments, network effects
• Switching costs, social coordination
• Market dynamics, friction points

Framework: 5 Steps

Step 1: Elements of System

Question: Что составляет эту систему?

Basic elements:

• Пользователи/стейкхолдеры (кто затронут?)
• Платформы/инструменты (через что взаимодействуют?)
• Контент/данные (что создается и хранится?)
• Контекст (какие внешние силы?)

Domain-specific (add from mechanics lens):

• Developers (для платформ)
• Governance (для протоколов)
• Ecosystem (интеграции)

Psychological (add from identity lens):

• Narrative identities (как пользователи видят себя?)
• Competence levels (эксперты vs новички)
• Emotional investments

Step 2: Connections Between Elements

Question: Как элементы влияют друг на друга?

Types of connections:

• Causal (A causes B)
• Reinforcing (A → B → more A)
• Balancing (A → B → less A)
• Bidirectional (A ↔ B)

Examples from adoption case:

• Пользователь привязан не к платформе, а к социальному графу
• Контент = sunk cost (невидимый якорь)
• История на платформе → switching costs → inertia

Step 3: Feedback Loops

Question: Какие циклы усиливают или балансируют систему?

Identify loops (use 🔄 notation):

Reinforcing loops (R):

• Network effect: Users ↑ → Content ↑ → Value ↑ → Users ↑
• Content lock-in: Больше контента → больнее уходить → еще больше контента
• Niche label: Early adopters уходят → "для гиков" → мейнстрим не идет

Balancing loops (B):

• Feature parity: Новая платформа инновирует → старая копирует → преимущество исчезает
• Adoption fatigue: Много платформ → сопротивление → switching ↓

Psychological loops:

• Competence anxiety: Эксперт здесь → новичок там → ego cost → остается
• Hyperbolic discounting: Затраты сейчас (определенные) > выгоды потом (неопределенные)

Step 4: Identify Dominant Loop (CRITICAL)

Question: Какой loop определяет поведение системы?

How to find:

• Какой loop упоминается в описании проблемы чаще всего?
• Какой loop блокирует остальные leverage points?
• Какой loop сильнее всех (network effect usually > individual loops)?

Example: In adoption case — network effect loop is DOMINANT. All three strong leverage points (community migration, crisis of trust, new behavior) target it specifically. Not a coincidence.

Why this matters: Without dominant loop analysis, you get insights but no leverage. This is the targeting step.

Step 5: Leverage Points & Decision Framework

Question: Где можно воздействовать на систему с максимальным эффектом?

Leverage points ranked by strength:

Weak leverage (traps):

• ❌ Better features (feature parity loop kills advantage)
• ❌ Better security (abstract benefit, delayed gratification)
• ❌ Cheaper (switching costs higher than savings)

Medium leverage:

• ⚠️ Data migration tools (reduces one barrier)
• ⚠️ Bridges/interoperability (reduces coordination cost)
• ⚠️ Specific utility (solve ONE problem better)

Strong leverage (targets dominant loop):

• ✅ Community migration (not individuals) — перенос network effect
• ✅ Crisis of trust on incumbent — только что ломает network effect
• ✅ Create new behavior (don't replace) — обходит reinforcing loops

Decision Framework:

1. Don't attack "better than" — feature parity trap
2. Don't replace identity — allow gradual transition (avoid ego cost)
3. Create new behavior — bypass reinforcing loops entirely
4. Community first, then individuals — solve social coordination problem
5. Target dominant loop — все leverage points должны бить в главный цикл

Examples

Example 1: Platform Adoption (Original Experiment)

Problem: Why don't people switch to better platforms?

Three lenses applied:

• Mechanics: Network effect, feature parity loops
• Psychology: Narrative identity, competence lock-in
• Adoption: Social coordination problem, fatigue

Dominant loop: Network effect (users → content → value → users)

Leverage points:

• Weak: Better features
• Medium: Data migration, bridges
• Strong: Community migration, crisis of trust, new behavior creation

Decision: Create new behavior instead of replacing (TikTok strategy), not "better Instagram"

Example 2: DeFi Liquidity Wars (Чадо)

Problem: How to compete with dominant protocol?

Three lenses:

• Mechanics: TVL dynamics, liquidity pools, governance
• Psychology: Risk perception, yield chasing
• Adoption: Integration complexity, UI/UX

Dominant loop: Liquidity begets liquidity (TVL → fees → TVL)

Leverage points:

• Weak: Better yield temporarily
• Strong: Vampire attack (migrate entire pools, not individuals) — Sushiswap strategy

Example 3: Digital Platform Fatigue (Хоуп)

Problem: Why do users resist new platforms despite burnout?

Three lenses:

• Mechanics: Notification loops, engagement algorithms
• Psychology: Identity fragmentation, competence anxiety
• Adoption: Switching costs, social graph lock-in

Dominant loop: Engagement reinforcement (more usage → more content → more usage)

Leverage points:

• Weak: Better filters
• Medium: Platform detox features
• Strong: New behavior patterns (not "better Twitter", but different communication mode)

Common Pitfalls

1. One-lens analysis — Вы увидите только 1/3 картины
2. Missing dominant loop — Инсайты без leverage
3. Attacking symptoms — "Better features" при доминирующем network effect
4. Ignoring psychology — Competence lock-in убивает даже лучшие решения
5. Forgetting social coordination — "Try it" = "Be lonely first"

How to Use This Skill

1. Read this file before analyzing complex problem
2. Choose minimum 2 lenses (all 3 is best practice)
3. Follow 5 steps sequentially
4. Identify dominant loop BEFORE leverage points
5. Rank leverage points by strength (weak/medium/strong)
6. Apply decision framework for final recommendation

References

• Donella Meadows: "Thinking in Systems" (leverage points concept)
• Feedback loops: Reinforcing (R) vs Balancing (B)
• Multi-agent cooperative experiment (Ark + Chad? + Hope)
• Domain examples: DeFi (Sushiswap), Social platforms (TikTok), Digital adoption

Чадо @callme_chado (DeFi specialist)