Cognitive Task Analysis

Cognitive Task Analysis (CTA) is the most rigorous extraction methodology available. Developed originally for military and aviation training, it is designed specifically for the problem of extracting expertise that practitioners cannot self-report accurately. CTA goes beyond behavioral task analysis — "what do you do?" — to cognitive task analysis — "what do you know, notice, decide, and judge while you are doing it?" The difference in output quality is significant.

1. Task Decomposition Interview Start with a broad task — "conduct a technical discovery call" — and decompose it into subtasks, then sub-subtasks, until you reach individual cognitive operations. The decomposition is not linear. It is hierarchical. Use a flowchart or hierarchical task diagram to capture the branching structure. The branching points are where expert judgment lives.
2. Knowledge Audit Probe For each subtask, probe the cognitive requirements: What do you need to know to do this? What are you looking for? What would cause you to do something different? What would a novice miss here? What is the most common mistake? These probes surface cues, conditions, and decision criteria that task description alone cannot reveal.
3. Scenario-Based Deepening After the initial decomposition, present three to five scenarios — one routine, one difficult, one where the expert made a mistake or encountered an unexpected outcome. For each scenario, walk through the task decomposition and probe what changed. Scenario-based deepening reveals the contextual variation in expertise that abstract description always flattens.
4. Simulation Probing Where possible, replicate the task environment and observe the expert performing the task while narrating. Simulation probing catches the actions the expert cannot recall in interview because they are automatic. "I notice you just skipped to step four. What told you step three was not necessary here?" That question cannot be asked without the simulation.

# Cognitive Task Analysis Session Template

## Setup
- Expert: [Name, role, years of experience]
- Task Domain: [Specific task being analyzed]
- Session #: [1 of N — plan 3+ sessions]
- Analyst: [Name]

## Phase 1: Task Decomposition (30 min)
"Walk me through [task] from beginning to end."
→ Capture: steps, sequence, decision points

## Phase 2: Knowledge Audit Probes (20 min per subtask)
For each decision point, ask:
- "What are you looking for here?"
- "What would cause you to do something different?"
- "What would a novice miss?"
- "What is the most common mistake at this step?"

## Phase 3: Scenario Deepening (30 min)
Scenario 1: Routine [describe]
Scenario 2: Difficult [describe]
Scenario 3: Past failure or unexpected outcome [describe]
→ For each: "Walk me through how this changes what you do."

## Phase 4: Synthesis Review (10 min)
Read back the extracted model. Ask:
- "What am I missing?"
- "Where am I wrong?"
- "What would surprise a novice about this?"