# Behavioral Experiment Design

> Design a real-world test to gather evidence about a belief that persists despite verbal challenging. Based on Aaron Beck's behavioral experiment methodology in cognitive therapy.

- Skill: `sethmblack/behavioral-experiment-design` (Agent Skill)
- Install (CLI): `npx skillmds add sethmblack/behavioral-experiment-design`
- Raw SKILL.md: https://api.skillmd.com/api/skills/sethmblack/behavioral-experiment-design/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- License: MIT
- Author: sethmblack (https://skillmd.com/u/sethmblack)
- Updated: 2026-09-08
- Page: https://skillmd.com/skills/sethmblack/behavioral-experiment-design

---


# Behavioral Experiment Design

Design a real-world test to gather evidence about a belief that persists despite verbal challenging. Based on Aaron Beck's behavioral experiment methodology in cognitive therapy.

**Token Budget:** ~900 tokens

---

## Constraints
- Do NOT design experiments that could cause harm (physical, social, financial, psychological)
- Do NOT push people into situations they're not ready for
- Experiments should be collaborative—person agrees to the design
- Start with lower-risk experiments before higher-stakes ones
- Respect the person's pace and boundaries

---

## When to Use

- Person knows a thought is probably distorted but still believes it
- Verbal discussion hasn't shifted the belief
- Person asks "But how do I know for sure?"
- Belief is testable through observable action
- Person is ready to gather real-world evidence

---

## Inputs

| Input | Required | Description |
|-------|----------|-------------|
| belief | Yes | The thought/belief to test |
| confidence | No | How strongly they believe it (0-100%) |
| context | No | Situations where this belief is active |

---

## Types of Behavioral Experiments

| Type | Description | Best For |
|------|-------------|----------|
| **Hypothesis testing** | Test a specific prediction | "If I do X, Y will happen" |
| **Discovery** | Gather information about what happens | "I don't know what will happen if..." |
| **Survey** | Ask others about their experience | "Everyone thinks..." beliefs |
| **Observation** | Watch without acting | "People always..." beliefs |
| **Behavioral test** | Try a new behavior | Avoidance-based beliefs |

---

## Workflow

### Step 1: Clarify the Belief
"What exactly is the thought we want to test? Let's make it as specific as possible."

*The belief should be:*
- Specific enough to test
- Stated as a prediction if possible
- Something the person genuinely holds

### Step 2: Rate Current Conviction
"On a scale of 0-100%, how much do you believe this thought right now?"

### Step 3: Identify a Testable Prediction
"If this thought is true, what would we expect to see in the real world?"

*Help translate vague beliefs into testable predictions:*
- "People will judge me" → "At least 3 people will look at me negatively"
- "I'll fail" → "I won't complete the task" or "I'll make more than 2 errors"
- "No one cares" → "If I reach out to 3 friends, none will respond positively"

### Step 4: Design the Experiment
"What's a situation where we could test this prediction?"

*Consider:*
- What will you do specifically?
- When and where?
- What will you observe/measure?
- How long will the experiment last?

### Step 5: Anticipate Obstacles
"What might get in the way? How will you handle it?"

*Common obstacles:*
- Safety behaviors that contaminate the experiment
- "Yes, but" rationalizations
- Selective attention to confirming evidence

### Step 6: Plan the Evaluation
"After the experiment, how will we know what happened? What will you look for?"

### Step 7: Predict Both Outcomes
"What would it mean if your prediction comes true? What would it mean if it doesn't?"

---

## Output Format

```
## Behavioral Experiment Design

**Belief to test:** "{The belief}"
**Current conviction:** {X}%

**Testable prediction:** "If I {action}, then {predicted outcome}."

**Experiment design:**
- **What:** {Specific action to take}
- **When:** {Timeframe}
- **Where:** {Context/setting}
- **Measure:** {What to observe/count}

**Potential obstacles:**
- {Obstacle 1} → Plan: {How to handle}
- {Obstacle 2} → Plan: {How to handle}

**Evaluation criteria:**
- Prediction confirmed if: {Criteria}
- Prediction disconfirmed if: {Criteria}

**After the experiment, reflect on:**
1. What actually happened?
2. How does this compare to your prediction?
3. What does this mean for the original belief?
4. New conviction level: ____%
```

---

## Outputs

**Primary Output:** A structured analysis document that identifies and articulates patterns, insights, and actionable recommendations based on the input data.

**Format:**
```markdown
## Analysis: [Topic]

### Key Findings
- [Finding 1]
- [Finding 2]
- [Finding 3]

### Recommendations
1. [Action 1]
2. [Action 2]
3. [Action 3]
```

**Example output:** See the Example section below for a complete demonstration.

## Example

**Input:** "I believe that if I speak up in meetings, people will think I'm stupid. Conviction: 85%"

**Output:**



**Why this works:**

This example demonstrates the key principles of the skill in action. The approach is effective because:
- It follows the systematic workflow outlined above
- It shows concrete application of the framework
- It produces actionable, specific outputs rather than vague generalizations
- The analysis is grounded in observable details
- The recommendations are prioritized and implementable

**Alternative applications:**

This same approach can be applied to:
- Different contexts within the same domain
- Related but distinct problem types
- Scaled up or down depending on scope
- Combined with complementary analytical frameworks


## Behavioral Experiment Design

**Belief to test:** "If I speak up in meetings, people will think I'm stupid."
**Current conviction:** 85%

**Testable prediction:** "If I make one comment or ask one question in the next team meeting, at least 2 people will react negatively (eye-roll, dismissive response, visible annoyance)."

**Experiment design:**
- **What:** Make one comment or ask one question during the meeting
- **When:** Next team meeting (Thursday 10am)
- **Where:** Regular weekly team meeting
- **Measure:** Count visible negative reactions; note any positive or neutral responses

**Potential obstacles:**
- Avoiding by staying silent → Plan: Prepare one question in advance; commit to saying it in first 15 minutes
- Discounting positive responses → Plan: Write down ALL responses, not just negative ones
- Safety behavior (speaking very quietly) → Plan: Speak at normal volume, even if uncomfortable

**Evaluation criteria:**
- Prediction confirmed if: 2+ people show visible negative reactions
- Prediction disconfirmed if: Fewer than 2 negative reactions; any positive reactions occur

**After the experiment, reflect on:**
1. What actually happened?
2. How does this compare to your prediction?
3. What does this mean for the original belief?
4. New conviction level: ____%

---

## Integration

This skill integrates with the Aaron Beck expert voice:
- Present experiments as collaborative—"What if we tried..." not "You should..."
- Start with experiments the person rates as moderately challenging (40-60 on anxiety scale), not extreme
- Frame disconfirming evidence as valuable data, not "proving them wrong"
- One experiment rarely eliminates a belief; plan for repeated testing
- Connect results back to the cognitive distortion or thought record work