Behavioural experiments for social anxiety

The difference between exposure and an experiment

Exposure is about approaching what you avoid and staying with the discomfort. A behavioural experiment adds a specific question: what does my anxiety predict, and what actually happens if I test it?

The two often overlap. Asking a cashier a question can be exposure. If your prediction is 'they will be annoyed', and you observe the outcome afterward, it becomes an experiment too.

Use the prediction-outcome format

Keep the structure simple. You need a prediction, an action, and an outcome.

• Prediction: If I ask my coworker to clarify the task, they will think I am incompetent.
• Action: Ask one specific clarification question.
• Outcome: They answered normally and added one useful detail.
• Learning: Asking did not create the judgement I predicted.

Make predictions measurable

A vague prediction like 'it will be terrible' is hard to test. A measurable prediction is easier: 'They will sigh', 'They will say no angrily', 'Nobody will respond', 'I will be unable to finish the sentence.'

Measurable does not mean cold or clinical. It just means you will know what happened afterward.

Start with low-stakes experiments

The best first experiment is boring enough to repeat. Ask for the time. Send one text without rereading it five times. Make a neutral comment in a queue. Ask a simple follow-up question. State a real preference about where to eat.

You are training approach, not auditioning for a new personality.

Review the result fairly

Afterward, do not grade the experiment by whether you felt anxious. Grade it by whether you did the action and what the evidence showed.

If the feared outcome partly happened, that is still data. Maybe someone was short with you, and you handled it. Maybe your voice shook, and the conversation continued. The learning may be: 'I can tolerate this', not 'nothing awkward ever happens.'