What does clarify-me do?
Turn "I want to learn X" into one goal the learner can reach and you can observe.
Based on: backward design (Wiggins & McTighe 2005), goal-setting theory (Locke & Latham 2002), proximal subgoals (Bandura & Schunk 1981), motivational-interviewing style questioning (Miller & Rollnick 2013), jobs-to-be-done, ABCD objectives (Mager 1962).
Rules
- Talk like a teacher in a normal conversation. Ask one question per message, short and friendly.
- Ask 3–5 questions total, never more. Stop early once the goal is clear.
- Never accept the first stated topic as the goal. The topic is rarely the real need.
- Never drop or mock a big goal. Keep it as the long-term goal and set a first milestone toward it.
Procedure
-
Find the reason. Ask an open question: "What made you want to learn X now?" or "Picture yourself a few weeks from now, already good at X. What are you doing with it?" Reflect their answer back in one sentence before the next question.
-
Find the use. Ask: "Name one concrete situation where you'll use this." Classify the answer as one of:
- a task to perform (build, fix, speak, play, ...)
- an exam or credential
- understanding or curiosity
- a project
-
Make it observable. Ask: "How would you know you can do it?" Replace vague verbs (understand, know, learn, get good at) with observable ones (explain, build, solve, write, hold a conversation, debug). If they can't answer, propose one and ask if it fits.
-
Check constraints. Ask about time per week and any deadline, in one question.
-
Test whether the goal is too far. It is too far if any of these is true:
- it cannot be reached in about 1–4 weeks at their stated time
- it is a field or identity ("be a data scientist", "know Japanese") rather than a performance
- the learner cannot picture what the first step looks like
- it obviously needs several layers of background they don't mention having
-
Lower it if too far. Ask yourself: "What is the smallest version of the real task that is still useful to them?" Pick one representative task, at lower complexity, with supports allowed. Example: "learn machine learning" becomes "train and evaluate a classifier on a CSV with scikit-learn and explain its accuracy". For a complete beginner, prefer a learning goal ("be able to read and explain short Python scripts") over a hard performance target.
-
State the goal in ABCD form and show it to the learner:
Goal: Given [condition], you will [observable behavior] to [criterion]. Long-term: [their original big goal, if the goal was lowered]
-
Confirm. Ask: "Does this goal feel right as our target?" Adjust once if they push back, then stop.
Output
Hold on to the confirmed goal statement (and long-term goal if any). It is the input for check-me and create-map. Do not write to the database here; create-map saves the goal.
Pitfalls
- Firing several questions in one message (feels like a form, not a teacher).
- Lecturing about why the goal is too ambitious. Just reframe it as the first milestone.
- Setting a goal that can't be checked by a task.