Case study · Conscious Discipline at Home
Parents don’t need another lesson — they need somewhere safe to try it.
A self-paced module I researched, designed, and built: it teaches parents their school’s discipline framework, then drops them into a real moment and responds to what they actually say.
Walk four screens
Not screenshots. The same HTML the module ships, embedded here.
Unscored, skippable, and deliberately before any teaching. Whatever the parent writes here is the only thing their final answer gets compared to.
Mid-scenario. The child answers what the parent actually wrote, the coach note sits inline under their line, and the phrase help stays collapsed until asked for.
A written answer scored against named criteria. Their own words are quoted back first; the model answer appears only after they have committed to theirs.
The first answer and the last, side by side, with no verdict attached. The ground is dark for the same reason it was on the baseline screen: the module is talking about the parent, not teaching them.
My role
Research, architecture, scenario writing, assessment design, UX, visual design, and the build. No client, no brief — I’m the parent this was written for.
Scope
3 sections · 35–40 minutes, resumable at any point.
Built with
HTML, CSS, JavaScript — the screens above are the real thing. Rise 360 for the reference section, Rive for concept animation.
The problem
My child’s school trains its staff in a discipline framework, and sent the parents a one-page PDF. The parents I know read it and agreed with it. But the first time we need any of it, it’s 7:40 in the morning, there’s a shoe across the hallway, and the bus is coming.
So the module is built around one bet: trying the words is not the same as reading them. A parent who has typed their own sentence and had it answered starts from a different place than a parent who has only read the handout.
How it’s built
Section 1
Learn
Four principles, each opened by the old frame it replaces.
Section 2
Test your knowledge
Four written scenarios, each needing more than one principle.
Section 3
Try it out
Three multi-turn role-plays that escalate: two turns, three, then four.
Before any screen got styled, a wireframe audit triaged 41 items into settled, blocking, and open; seven blockers closed.
An unscored baseline sits before section 1 and its reflection after section 3, so the comparison is self-referenced: the parent measures against their own first answer, not a model answer and not a number. Attempting the moment before being taught it is also what makes the teaching land.
Six decisions, and what I rejected
Every one of these had a defensible alternative.
Assessment
No item maps to a single principle
The knowledge check is four written scenarios — supermarket, forced apology, the thing you said, the skeptical grandparent — each needing more than one principle at once.
Instead of
One item per principle
Mapping one item to one principle teaches the module’s own shape instead of the material: parents learn that the fourth principle was about limits, and answer from the table of contents rather than from the situation in front of them. Real moments never arrive labeled with which principle applies.
Assessment
No percentages anywhere
Every result is a band — Strong, Nearly there, Not yet — with the criteria named in plain sentences. There is no overall score at module end.
Instead of
Per-scenario percentages with no total
The incoherent middle: precise enough to invite arithmetic, incomplete enough that the arithmetic is meaningless. And a parent who scores 62% on staying calm with their own child has been handed a number about their parenting. The bands survived because they can be said out loud without that.
Assessment
The composure cap is never explained as a rule
If the parent loses their composure, the band is Not yet and the other three criteria go gray rather than ticked, under one line: they only work from a calm voice.
Instead of
Showing three ticks and a failing result
Accurate, and it feels unfair — I did three of the four things and you failed me. Naming the rule up front (“composure gates everything”) would have been worse: parents would then read every calm reply as gaming the gate.
Help
Hot help inline, cold help in one panel
Phrase suggestions live in a collapsed strip directly above the answer field. Everything read once — glossary, save state, privacy — lives in a single panel behind a ? key.
Instead of
One unified Help drawer; or a permanent reference rail
Merging them puts a calm reference panel in a hot path every time the parent needs three words fast. The permanent rail had the best case on paper — always visible, nothing to learn — but it costs 268px, and at 375px that leaves no room for the answer it exists to support. A third option — dragging phrase cards into the field — died on the detail: it fights touch scroll and the raised keyboard, is unusable by keyboard and screen reader, and needs a tap fallback anyway, so you ship two interactions.
Help
The strip never opens itself
Collapsed by default. When the coach sees a miss it offers the phrases in words — “there are three phrasings under Words to try, if you want them” — and remembers the parent’s choice after they open it once.
Instead of
Auto-expanding on a wrong answer
A panel that springs open the moment you get something wrong is a correction, and it arrives exactly when a struggling parent is least able to hear one. Said as an offer, the help is still right there, and the parent decides whether to take it.
Visual
Dark ground means the module is talking about you
Three colors carry meaning and nothing else: blue for emphasis while teaching, green only to confirm something the parent got right, plum for developmental feedback. Dark screens are reserved for the reflective moments — baseline, reflection, close.
Instead of
Dark as a section marker
Using dark to mean “you are in section 3” made it decoration. It told the parent where they were, which the header already does. Tying it to a mode of address means the shift in ground is doing work no label could. Tinted fields also lost their strokes: once tint carries meaning, a border around it is a second, competing signal.
The decisions above about tone and stakes are the oldest part of this practice. Asking someone to try a thing they expect to be bad at, and keeping them willing to try it again, is what I was doing in a classroom long before I built any of this.
“I’ve come to understand that learning is never just about mastering what’s written in books. What meant even more to me was the sense of warmth I felt in Dr. LeMay’s class — it truly made FSU feel like a home away from home.”
What this transfers to
Parenting is the subject matter, not the design problem. The design problem was scoring what someone wrote, in a domain where being told you got it wrong stings.
- Free-text scenario design. Writing prompts that can’t be answered by pattern-matching the last screen, with matcher criteria and per-branch feedback for each.
- Branching feedback that names criteria, not scores. Including the case where one criterion cancels the other three.
- Behavior change over recall. A baseline captured before instruction and shown back after.
- Help design under load. Separating reference a learner reads once from support they need mid-task, in one tap.
- Low-stakes assessment in a high-stakes emotional context. Same problem as compliance, feedback, and difficult-conversation training.
How it gets tested
No learner has used this yet. The module goes live in mid-October, and what follows is the plan — written before the results exist, so it can’t be written to fit them.
Oct 14–24
Eight parents, unmoderated, think-aloud on three. The question is completion, not satisfaction: does anyone finish all three role-plays in one sitting when told it’s three?
Ongoing
Read the free-text answers against the matcher. Every answer the matcher called wrong that a human would call right is a content bug. The “I think my answer covered this” link exists to collect them.
+30 days
One question, by email: did you use any of it? This is the only row that tests the module’s actual bet, and it’s the weakest instrument on the page: self-report, thirty days out. Short of putting a researcher in someone’s hallway, it’s what I have. With a second cohort I’d replace it with an unannounced fourth role-play.
Before that
Subject-matter review. A practitioner reads the four principles for substance before any parent sees them. Plain-language paraphrase is a copyright position, not a guarantee of accuracy.
The working documents
The alternatives, the reasoning, and the two I lost.
Matcher criteria, feedback branches, failure behavior.
Every screen in parent-journey order, gaps marked as gaps.
Type scale, the three-color system, the light/dark rule.
What I’m looking for
A team where a design that isn’t landing gets fixed in weeks. If you’re hiring for instructional or learning experience design, I’d like to hear about it — and I’m glad to walk through any decision on this page in more detail.