grad-dual-process
用 System 1 和 System 2 診斷判斷為什麼出錯，並設計去偏誤的介入
中文翻譯
--- name: "grad-dual-process" description: "用雙歷程理論診斷判斷是來自快速直覺（System 1）還是緩慢分析（System 2）的處理，並找出由此產生的認知偏誤。當使用者需要解釋為什麼快速決策會出錯、設計考慮到認知預設值的選擇架構、稽核決策流程中的捷思錯誤，或問到「為什麼人會誤判機率」「怎麼減少直覺判斷的錯誤」「直覺什麼時候會失靈」時，就用這個 skill。" metadata: category: "WP-29 心理學理論" tags: ["dual-process", "system-1", "system-2", "heuristics", "cognitive-bias", "Kahneman", "thinking-fast-and-slow"] --- # Dual-Process Theory ## Overview 雙歷程理論（Kahneman，2011；Stanovich & West，2000）區分兩種認知處理模式：System 1（快速、自動、由捷思驅動）和 System 2（緩慢、審慎、基於規則）。大多數判斷預設走 System 1，它很有效率，但捷思失靈時容易產生系統性偏誤。 ## When to Use - 解釋為什麼利害關係人在時間壓力或複雜情境下，會出現可預測的判斷錯誤 - 設計能補償 System 1 預設值的決策環境（推力、檢查清單） - 稽核現有流程，找出捷思捷徑在哪裡帶來風險 - 評估直覺專業什麼時候可靠、什麼時候會誤導 ## When NOT to Use - 決策本來就有結構化的演算法程序（偏誤已被工程化排除） - 拿來當藉口，全盤否定直覺判斷：在高效度的環境中，專家直覺可以很準 - 問題出在動機，不是認知（人知道正確答案，卻選了別的） ## Assumptions ``` IRON LAW: System 1 operates by DEFAULT — System 2 engagement requires cognitive effort and is easily depleted. Under time pressure, cognitive load, or ego depletion, System 1 dominates and heuristic biases amplify. ``` 關鍵假設： 1. System 1 和 System 2 是處理模式的隱喻，不是腦中真正分離的系統 2. 捷思通常是適應性的，偏誤出現在邊界條件 3. System 2 可以推翻 System 1，但前提是有被提示，而且有足夠的認知資源 ## Methodology ### Step 1 — Identify the Judgment or Decision Context 描述決策的特徵：時間壓力、複雜度、熟悉度、賭注大小、情緒投入程度。 ### Step 2 — Classify Processing Mode | Feature | System 1 | System 2 | |---------|----------|----------| | Speed | 快、自動 | 慢、費力 | | Awareness | 無意識 | 有意識 | | Capacity | 高（平行） | 低（序列） | | Basis | 捷思、聯想 | 規則、邏輯 | | Error type | 系統性偏誤 | 計算錯誤 | | Triggered by | 預設、熟悉感 | 新奇、偵測到衝突 | ### Step 3 — Map Relevant Heuristics and Biases 常見的 System 1 捷思與其失靈模式： - **Availability**：用回想的容易程度判斷頻率，會被顯著性和近期性帶偏 - **Representativeness**：用相似度判斷機率，忽略基本比率 - **Anchoring**：從初始值開始調整來估計，調整不足 - **Affect**：用情緒反應判斷風險與利益，忽略統計證據 ### Step 4 — Design Intervention - **De-bias**：放慢決策、要求明確說明理由、做 pre-mortem - **Nudge**：重組選擇架構，讓 System 1 的預設值和期望結果一致 - **Leverage**：在高效度、回饋快速的領域，善用 System 1 的優勢（模式辨識） ## Output Format ```markdown ## Dual-Process Analysis: [Context] ### Decision Environment - Time pressure: [High/Medium/Low] - Complexity: [High/Medium/Low] - Emotional involvement: [High/Medium/Low] - Dominant processing: [System 1 / System 2 / Mixed] ### Heuristic-Bias Map | Heuristic | Bias Triggered | Evidence | Risk Level | |-----------|---------------|----------|------------| | [heuristic] | [bias] | [observation] | [High/Med/Low] | ### Intervention Design 1. [De-biasing or nudge strategy] 2. [Process change] 3. [Environmental redesign] ``` ## Gotchas - System 1/System 2 是有用的隱喻，不是字面上的大腦架構，避免把這個區分實體化 - 在回饋明確、模式規律的領域（例如西洋棋、消防），專家直覺（System 1）非常準 - 去偏誤訓練的遷移效果很差，改變環境比訓練個人更有效 - 認知耗竭效應有爭議，不要假設有簡單的「意志力電池」模型 - System 2 本質上不是比較「好」，它比較慢、成本比較高，而且一樣會受動機性推理影響 - 人常把信心和準確度搞混，System 1 的高信心不代表正確 ## References - Kahneman, D. (2011). *Thinking, fast and slow*. Farrar, Straus and Giroux. - Stanovich, K. E. & West, R. F. (2000). Individual differences in reasoning: implications for the rationality debate. *Behavioral and Brain Sciences*, 23(5), 645-665. - Tversky, A. & Kahneman, D. (1974). Judgment under uncertainty: heuristics and biases. *Science*, 185(4157), 1124-1131.
英文原文
--- name: "grad-dual-process" description: "Apply dual-process theory to diagnose whether judgments arise from fast intuitive (System 1) or slow analytical (System 2) processing and identify resulting cognitive biases. Use this skill when the user needs to explain why quick decisions go wrong, design choice architectures that account for cognitive defaults, audit decision processes for heuristic errors, or when they ask 'why do people misjudge probability', 'how to reduce snap-judgment errors', or 'when does intuition fail'." metadata: category: "WP-29 心理學理論" tags: ["dual-process", "system-1", "system-2", "heuristics", "cognitive-bias", "Kahneman", "thinking-fast-and-slow"] --- # Dual-Process Theory ## Overview Dual-process theory (Kahneman, 2011; Stanovich & West, 2000) distinguishes two modes of cognitive processing: System 1 (fast, automatic, heuristic-driven) and System 2 (slow, deliberate, rule-based). Most judgments default to System 1, which is efficient but prone to systematic biases when heuristics misfire. ## When to Use - Explaining why stakeholders make predictable judgment errors under time pressure or complexity - Designing decision environments (nudges, checklists) that compensate for System 1 defaults - Auditing existing processes to identify where heuristic shortcuts introduce risk - Evaluating when intuitive expertise is reliable vs. when it is misleading ## When NOT to Use - When decisions are already well-structured with algorithmic procedures (bias is engineered out) - As an excuse to dismiss all intuitive judgment — expert intuition can be accurate in high-validity environments - When the problem is motivational rather than cognitive (people know the right answer but choose otherwise) ## Assumptions ``` IRON LAW: System 1 operates by DEFAULT — System 2 engagement requires cognitive effort and is easily depleted. Under time pressure, cognitive load, or ego depletion, System 1 dominates and heuristic biases amplify. ``` Key assumptions: 1. System 1 and System 2 are metaphors for processing modes, not discrete brain systems 2. Heuristics are generally adaptive — biases emerge at the boundary conditions 3. System 2 can override System 1, but only when cued and when cognitive resources are available ## Methodology ### Step 1 — Identify the Judgment or Decision Context Characterize the decision: time pressure, complexity, familiarity, stakes, emotional involvement. ### Step 2 — Classify Processing Mode | Feature | System 1 | System 2 | |---------|----------|----------| | Speed | Fast, automatic | Slow, effortful | | Awareness | Unconscious | Conscious | | Capacity | High (parallel) | Low (serial) | | Basis | Heuristics, associations | Rules, logic | | Error type | Systematic biases | Computational mistakes | | Triggered by | Default, familiarity | Novelty, conflict detection | ### Step 3 — Map Relevant Heuristics and Biases Common System 1 heuristics and their failure modes: - **Availability**: judge frequency by ease of recall — biased by salience and recency - **Representativeness**: judge probability by similarity — ignores base rates - **Anchoring**: estimate by adjusting from initial value — insufficient adjustment - **Affect**: judge risk/benefit by emotional reaction — neglects statistical evidence ### Step 4 — Design Intervention - **De-bias**: slow down decisions, require explicit justification, use pre-mortems - **Nudge**: restructure choice architecture to align System 1 defaults with desired outcomes - **Leverage**: use System 1 strengths (pattern recognition) in high-validity, rapid-feedback domains ## Output Format ```markdown ## Dual-Process Analysis: [Context] ### Decision Environment - Time pressure: [High/Medium/Low] - Complexity: [High/Medium/Low] - Emotional involvement: [High/Medium/Low] - Dominant processing: [System 1 / System 2 / Mixed] ### Heuristic-Bias Map | Heuristic | Bias Triggered | Evidence | Risk Level | |-----------|---------------|----------|------------| | [heuristic] | [bias] | [observation] | [High/Med/Low] | ### Intervention Design 1. [De-biasing or nudge strategy] 2. [Process change] 3. [Environmental redesign] ``` ## Gotchas - System 1/System 2 is a useful metaphor, not a literal brain architecture — avoid reifying the distinction - Expert intuition (System 1) is highly accurate in domains with clear feedback and regular patterns (e.g., chess, firefighting) - De-biasing training has poor transfer — changing the environment is more effective than training individuals - Cognitive depletion effects are debated; do not assume a simple "willpower battery" model - System 2 is not inherently "better" — it is slower, more costly, and still subject to motivated reasoning - People often confuse confidence with accuracy; high System 1 confidence does not indicate correctness ## References - Kahneman, D. (2011). *Thinking, fast and slow*. Farrar, Straus and Giroux. - Stanovich, K. E. & West, R. F. (2000). Individual differences in reasoning: implications for the rationality debate. *Behavioral and Brain Sciences*, 23(5), 645-665. - Tversky, A. & Kahneman, D. (1974). Judgment under uncertainty: heuristics and biases. *Science*, 185(4157), 1124-1131.