Teach-Back Mental Model Elicitation

Description

Asks users to explain or teach how a system works in their own words, then scores the explanation against a reference model to expose mental model gaps.

In medicine, teach-back confirms patient understanding: the clinician asks the patient to explain the instructions back, corrects misunderstandings and repeats until correct. Pask used 'teachback' to externalise a learner's conceptual structure. As a UX method, participants explain to a confederate novice how a feature works (e.g. how their data is shared, how the AI decides), sometimes with drawing. Explanations are coded against a reference conceptual model for correct, missing and incorrect propositions, giving a quantitative mental model accuracy score. It is far more diagnostic than asking 'did that make sense?', and directly tests whether interface explanations transfer understanding.

Attributes

Loss type
Unmigrated
Data
Mixed
Effort
Days

Steps

  1. Write a reference conceptual model of the system as a list of key propositions (e.g. 10-20).
  2. After users interact with the product or explanation, ask them to teach a novice (or the researcher playing one) how it works.
  3. Allow drawing and probing with neutral prompts ('and then what happens?').
  4. Code transcripts for correct, missing and incorrect propositions; compute accuracy and misconception frequency with two coders.
  5. Identify common misconceptions and the interface elements that produce them.
  6. Revise explanations and repeat teach-back until a target (e.g. 80% of key propositions correct) is met.

Sources

  • Schillinger et al., Closing the loop: physician communication with diabetic patients who have low health literacy (2003), Archives of Internal Medicine
  • Pask, Conversation, Cognition and Learning (1975)

Also known as

  • Teach-back
  • Teach-back method
  • Explain-to-a-novice protocol

Categories

  • User Research
  • Interviewing & Elicitation