By the end of this session, you can…
- LO 2.1Write a learner description that names a specific population, their current capability, and the context in which they practice.
- LO 2.2Articulate a measurable gap between current and desired performance — something a colleague could verify.
- LO 2.3List the three most binding constraints of use (time, device, setting, supervision) that any solution must respect.
- LO 2.4Produce D1 — a one-page Design Problem Statement — and defend it through one round of peer critique.
- LO 2.5Revise your D1 based on one specific piece of feedback; log what you changed and why.
Generic learners produce generic games
Almost every failed learning game can be traced back to a learner description that was too thin to discipline the design. "High school students" is not a learner. "Trainee pharmacists in their third clinical rotation who can recall drug names but freeze when asked to reason about interactions under time pressure" is a learner.
The level of specificity you reach here determines how sharp your mechanics can be in later sessions. You cannot design a game to teach "judgment" — you can only design a game to teach this group how to resolve this kind of ambiguous situation in this setting.
If you cannot say the learner, the gap, and the context in three short sentences, you do not know them yet. Go talk to one more person before the next session.
The Design Problem Statement
One page. Six fields. This document lives at the top of every subsequent deliverable's repo — it is the reference every later design decision is measured against.
| Field | What belongs | What disqualifies it |
|---|---|---|
| Learner | Named population, career stage, prior exposure to the domain. | "Students." "Learners." Anything that would also describe 100M people. |
| Current state | What they can already do; common patterns of error. | Assertions with no source. Write "observed in interview" or drop it. |
| Desired state | The behavior or judgment you want them to show afterward. | "Understand X." Unobservable verbs are not targets. |
| Gap | The measurable distance between current and desired — stated as a scenario one could score. | Vague direction words ("improve," "be better at"). |
| Context of use | Where / when / how long / with whom / on what device the game will be played. | "Anywhere." Everything is a trade-off; own one. |
| Constraints | Three hard limits: time budget per session, setting, supervision level. | Aspirational caveats. These are walls, not preferences. |
| Success signal | One observable that, if true after play, tells you the game worked. | "They enjoyed it." Engagement ≠ learning. See Session 01. |
A filled-in D1, ready to defend
Night-shift resuscitation decisions for first-year residents
- Learner
- First-year internal-medicine residents in an academic teaching hospital, in the first three months of night-float rotations. They have passed step exams and completed a simulation orientation.
- Current state
- They can recite protocols. When a real patient deteriorates, they over-order tests, delay calling the attending, and freeze on low-information decisions (observed in six shadowing sessions).
- Desired state
- When presented with an ambiguous deteriorating patient, they escalate earlier, order fewer but more discriminating tests, and articulate a leading diagnosis within five minutes.
- Gap
- On the departmental sim-lab rubric, current cohort averages 2.1/5 on "prioritizes high-yield data"; target is 3.5/5 after a single 45-minute session with the game.
- Context of use
- Played in the resident lounge, 30–45 minute blocks, alone or in pairs, on hospital-issued laptops during down-time on call.
- Constraints
- (1) No audio — call rooms are shared. (2) Must pause gracefully for pages. (3) No cost to residents; runs in browser.
- Success signal
- On a blinded post-test scenario, residents who played name the leading diagnosis within five minutes at a rate 20 pp above a matched control.
Notice every field resists a category slip. The gap is a rubric score, not a vibe. The context is a room, not a platform. The success signal is an observable, not a feeling. This is what you are aiming at.
Draft, critique, revise
| Time | What happens | Facilitator cue |
|---|---|---|
| 00:00–25:00 | Solo drafting. Fill all seven fields from your interview notes. | "Do not start with the fun parts. Learner, then gap." |
| 25:00–55:00 | Triads. Each person reads their D1 aloud; others score each field pass/revise; 10 min each. | Circulate; enforce the timer. No design talk yet — just the D1. |
| 55:00–80:00 | Revise your D1 in place, tracked changes on. | "Revise one field. Don't start over." |
| 80:00–90:00 | One triad volunteers; present revised version to the full room. | Capture the sharpest revision on the board. |
Reviewers will want to suggest mechanics. Shut that down. At this stage, feedback is only on the seven fields — specificity, observability, defensibility. Mechanic critique comes in Session 04.
An equity lens for the learner analysis
This week we committed to who the learner is and what gap we are closing. The UDL 3.0 Crosswalk keeps variability visible so your design problem does not quietly assume a single “default” learner.
UDL 3.0 Crosswalk
Translates CAST's UDL Guidelines 3.0 into a working crosswalk for this course — engagement, representation, and action/expression mapped across course design, artifact design, and prototype design.
Why this week Before you lock a gap statement, read the crosswalk and add one sentence to your learner analysis about where variability will be expected and where barriers are likely.
Before next week
Drafting a D1 faster without making it worse
Google AI Studio (aistudio.google.com) is useful for three specific D1 tasks: pulling a first-pass learner description from interview notes, pressure-testing the "gap" field for specificity, and generating three failure-mode variants of your problem statement so you can see what a weak version looks like.
Use case · Draft the learner description from interview notes
Model: Gemini 2.5 · temperature 0.4Paste your raw interview notes into a new chat. The first prompt does one job only — extract a named population from the evidence you collected. You will still write the final description; AI Studio is a scaffold, not a ghostwriter.
You are helping an educator write the "Learner" field of a Design Problem
Statement for an educational game. Your job is to extract — not invent — a
named population from the notes I paste.
Rules:
- Every claim you make must be traceable to a specific note I gave you.
- If the notes do not support a claim, say "not in notes."
- Reject generic descriptors ("students," "learners," "young people").
- Output in exactly this structure:
Population: ...
Career stage: ...
Prior exposure: ...
Evidence per line: "[quote or paraphrase] — note #N"
- Do not suggest mechanics, games, or solutions. Stay on the learner.
Here are my interview notes from 4 shadowing sessions. Extract the Learner
field. Anything generic, flag and ask a clarifying question instead of
smoothing it over.
[paste notes]
Use it when
You have ≥3 pages of interview notes and feel stuck turning them into one paragraph. Or when you want to see a stripped-back version of what the evidence actually supports.
Don't use it when
You have no notes yet. AI will happily invent a plausible learner from a vague prompt — that is the failure mode this session is built to prevent. Talk to a real person first.
Use case · Pressure-test the Gap field
Adversarial modeAfter you have drafted a Gap sentence, ask the model to attack it. The value here is not the answer — it is the list of angles a skeptical reviewer (your peers in the workshop) will come at you with.
My Gap field, draft 1:
"[paste your one sentence]"
Critique it against the D1 quality bar: specific population? observable?
scorable? measurable? Name three concrete weaknesses and three
rewrite options that would survive peer critique. Do not soften. I need
to hear the worst version of the feedback now, not in workshop.
The model's instinct is to broaden ("replace 'night-shift residents' with 'new clinicians'"). That is the opposite of the D1 discipline. Keep rewrites that tighten specificity; throw out rewrites that widen the population.
Paste your revised gap statement
Copy just the "Gap" field of your revised D1 here. If you cannot state it as a scorable scenario, say so — the instructor will review before next week.