Meadow
Deliverable Review
Incorrect. Try again.
M2 Build Spec  /  M2-003 + M2-004

SCS Conversation Engine

How Meadow turns single words into back-and-forth conversations — the core of what makes this app a communication tool, not just a soundboard.
M2-003 + M2-004
v2.0 · May 2026
Engine Delivered
Design sources: Sentence Engine Interaction Design Vocabulary & Routines
01 The Big Idea
Beyond single words

Most AAC apps stop at “tap a word, hear it spoken.” Meadow keeps the conversation going.

When a child taps “banana” in most AAC apps, the iPad says “banana” and that’s it. The child has to start over for every new thought. That’s a soundboard, not a conversation tool.

Meadow does something different. After the child taps “banana” and hears it spoken, three follow-up options appear instantly — like “I want banana,” “yummy banana!” and “more banana.” The child taps one, it speaks, and new follow-ups appear. We call this the Speak → Choose → Speak loop, or SCS.

The child isn’t just pressing buttons. They’re having a real back-and-forth exchange — the same way language naturally develops through conversational turns.

🗣 SPEAK Child taps a word 💬 CHOOSE 3 follow-ups appear 🔊 SPEAK Selection plays aloud loop continues
02 How It Adapts to Each Child

Three developmental tiers, one system

Every child develops at their own pace. The SCS engine adjusts its behavior based on the child’s developmental tier — set by the parent or speech therapist in the settings. A younger child who is just learning cause-and-effect gets a simpler experience. An older child who is starting to build sentences gets the full conversation loop.

Tier 1 — First Words

~12–18 months

  • Tap a word, hear it spoken. That’s it.
  • No follow-ups, no conversation loop
  • Single words only — “banana,” “milk,” “more”
  • Focus: learning that tapping makes something happen
Tier 2 — Word Combos

~18–30 months

  • SCS loop is active with 2-word follow-ups
  • “want banana,” “more milk,” “no juice”
  • Three suggestions appear after each tap
  • Focus: combining words to express more
Tier 3 — Early Sentences

~30–48 months

  • SCS loop with 2–3 word follow-ups
  • “I want banana,” “give me milk”
  • Richer variety of follow-up types
  • Focus: expressing complete thoughts
Tracing to M1 decisions

These tier behaviors come directly from M1-003 (Interaction Design). Tier 1 children experience the companion’s guided engagement cycle — Spotlight → Present → Wait → Celebrate → Advance — with expectant pauses of 30–45 seconds at Tier 1, 10–15 seconds at Tier 2, and 5–10 seconds at Tier 3. The SCS conversation loop activates only at Tier 2+, where the child has moved beyond cause-and-effect learning into intentional word combination. See Interaction Design for the full pathway specifications.

Scaffolding levels — parent-controlled

Parents or therapists can also adjust how much help the app provides, independent of the developmental tier. This is set in Settings behind the parent gate.

Full support

3 Suggestions

The default. Three follow-up options appear after every word. Best for children who are still learning to use follow-ups.

Light support

1–2 Suggestions

Fewer choices for children who are gaining confidence. Reduces visual clutter while still offering help.

Off

Words Only

No follow-ups at all. The child taps words directly, like Tier 1. For children who have outgrown scaffolding.

03 Smart Suggestion Selection

How Meadow picks the right follow-ups

When a child taps a word, the app doesn’t just show random phrases. It runs through a careful selection process to show the most useful, varied, and contextually appropriate follow-ups. Here’s how it works, step by step.

1

Find matching follow-ups

Look up all the follow-up phrases that connect to the word the child just said, filtered by the child’s developmental tier. A Tier 2 child only sees 2-word combos; a Tier 3 child sees longer phrases too.

2

Prefer the current scene

If the child is in the Kitchen scene and taps “want,” kitchen-specific follow-ups like “want banana” and “want milk” appear first. Scene context makes suggestions feel natural and relevant.

3

Mix different types of communication

The app ensures variety. Instead of three requests (“want banana,” “want milk,” “want juice”), it mixes in a comment (“yummy banana!”) and a feeling (“banana yay!”). This exposes the child to different ways of using language.

4

Avoid repetition

If the child has tapped “banana” three times in a row, the app shows different follow-ups each time. Previously shown suggestions get shuffled to the back of the line.

5

Fill in the gaps

If there aren’t enough authored follow-ups for a word, the app automatically generates simple ones like “more banana,” “no banana,” or “banana please.” The child always gets three choices.

Communication types

Speech therapists call these “pragmatic functions” — the different purposes a sentence can serve. Meadow ensures follow-ups span multiple types so children experience the full range of communication, not just requesting.

🤲

Requesting

Asking for something. “want banana,” “more milk”

👁

Commenting

Describing or observing. “big dog!” “yummy banana”

🙅

Protesting

Saying no. “no bath,” “don’t want”

👋

Greeting

Social hellos. “hi mommy,” “bye bye”

Asking

Seeking information. “where mommy?” “what that?”

💕

Feeling

Expressing emotions. “happy!” “sad” “I love you”

👉

Directing

Telling someone to act. “come here,” “look!” “help me”

Clinical traceability

These eight communication functions trace to Beukelman & Light’s Participation Model, documented in M1-006 Clinical Traceability. The M1 framework identifies seven core functions: Requesting, Refusing, Expressing Feelings, Commenting, Social Routines, Directing, and Answering. The SCS engine’s pragmatic function tags ensure every follow-up suggestion maps to one of these clinically validated communication purposes.

Why variety matters

Research on early language development shows that children who are exposed to diverse communication functions — not just requesting — develop broader conversational skills. Many AAC apps inadvertently train children to only make requests because that’s the easiest interaction to model. By mixing follow-up types, Meadow ensures the child practices commenting, protesting, greeting, and feeling — the full toolkit of human communication.

04 Scenario Walkthrough

See it in action

Three real examples showing the SCS loop at work across different scenes and tiers. Notice how each turn produces new, contextually relevant follow-ups.

🍽 Kitchen · Tier 2

Snack time conversation

Turn 1: Child taps “banana” → iPad speaks “banana!” → three follow-ups appear.

🍌banana
🤲want banana
😋yummy banana
🙌more banana

Turn 2: Child taps “want banana” → iPad speaks “want banana!” → new follow-ups appear for the next thought.

🤲want banana
🙌want more
🙏banana please
🥣want juice
🛌 Bedroom · Tier 3

Bedtime routine

Turn 1: Child taps “sleep” → iPad speaks “sleep!” → longer, richer follow-ups appear for Tier 3.

💤sleep
🤲I want sleep
🙅no more sleep
😴sleepy time

Turn 2: Child taps “I want sleep” → iPad speaks the full phrase → follow-ups continue the thought.

🤲I want sleep
💦sleep now
🙏sleep please
📚read book
🛵 Park · Tier 2

A full conversation at the playground

This shows the loop running through three consecutive turns — the child driving a real conversation about what they want to do at the park.

Turn 1
🚴go
🛵go slide
🧸go swing
Turn 2
🧸go swing
🙌more swing
😄fun swing
Turn 3
🙌more swing
🙏swing please
🖐push me

Three turns, zero navigation. The child drove the entire conversation from a single starting word.

05 Safety & Clinical Grounding

Communication comes first. Always.

The SCS engine is designed with several non-negotiable safety rules that protect the child’s communication experience.

🔒

Communication is never interrupted

Celebrations, animations, and companion reactions always queue up and wait. If the child is mid-conversation — follow-ups visible, words being spoken — nothing cuts in. The child’s voice always takes priority.

💔

Feelings are immediate

When a child taps “sad” or “angry,” the word speaks instantly. It does not trigger a conversation loop or force follow-ups. Expressing a feeling should never be complicated. One tap, one word, heard.

12-second timeout

If the child stops interacting, the speech bubble and follow-ups clear after 12 seconds. This prevents a stale conversation from confusing the child. The timeout is configurable by the parent in Settings.

🎯

Tier 1 is protected

Children at the earliest developmental stage (Tier 1) never see the SCS loop. Their experience is simple cause-and-effect: tap a picture, hear the word. No follow-ups, no choices, no cognitive overload.

Research foundation

The SCS loop is grounded in how children naturally acquire language. Decades of research in speech-language pathology show that children learn to communicate through conversational turns — back-and-forth exchanges where each person responds to the other. Single-word drills teach vocabulary, but conversational turns teach communication. The SCS loop brings this principle into the AAC app experience, giving every child the opportunity to practice real conversational interaction.

06 What Was Delivered

The engine is built and tested

The SCS Conversation Engine was assembled from five components, each depending on the one before it. All five are complete, integrated, and passing tests on the simulator. iPad 9 hardware validation happens at milestone review.

🗃

Follow-Up Data

Phrases tagged by trigger word, scene, tier, and communication type

🧠

Selection Algorithm

5-step process: match, rank, diversify, avoid repeats, fill gaps

🔁

SCS Loop

The state machine — speak, choose, speak again

💬

Follow-Up Pills

Tappable buttons with Fitzgerald Key color coding

🔊

TTS Integration

Every follow-up speaks aloud on tap

19
Automated Tests Passing
8
Communication Types
3
Scaffolding Levels
12s
Configurable Timeout

Content vs. engine

The engine is complete — it can accept any number of follow-up phrases and run the full SCS loop. Right now it has ~20 test entries to validate the algorithm. In M3 (Kitchen Scene), this engine gets loaded with 200+ scene-specific follow-ups covering every Kitchen vocabulary item. The engine scales to that volume without changes.

07 Acceptance Criteria

How we know it works

Each criterion is a testable checkpoint. Green items pass in automated tests today. The remaining item needs M3 Kitchen content to validate fully.

ID What We Test Must Pass Status
M2-003 Tier-appropriate combinations The engine produces follow-ups matching the child’s tier. Tier 1: no follow-ups. Tier 2: 2-word combos. Tier 3: 2–3 word combos. Full coverage of every Kitchen item comes in M3. 🟡
M2-004 Follow-ups appear instantly After tapping a Tier 2 or Tier 3 word, speech fires and 3 follow-up suggestions appear within the 400ms word-to-speech budget — fast enough that the child perceives them as instant. 🟢
M2-004 Communication type diversity The 3 follow-up suggestions span at least 2 different communication types (for example, one request and one comment — not three requests). 🟢
M2-004 Loop runs without errors The SCS loop runs for at least 5 consecutive exchanges without errors, freezes, or crashes. 🟢
M2-004 Timeout clears the bubble If the child stops interacting, the speech bubble and follow-ups clear after 12 seconds. No stale conversations remain on screen. Timeout is configurable by the parent. 🟢

Testing approach

19 automated tests verify the engine on every build. Performance will be validated on a real iPad 9 (not just the simulator) at milestone review to confirm the experience matches what an actual child would feel.

08 What’s Next

The engine is ready — now it needs a world to live in

The SCS conversation engine is complete and tested. The next milestones fill it with content and connect it to the rest of Meadow.

🍽

M3: Kitchen scene follow-ups

200+ authored follow-ups for every Kitchen vocabulary item — fridge, stove, table, snacks, drinks, utensils. This is where the engine goes from test data to a real conversation experience.

🤖

Companion integration

The companion character will react to SCS conversations — signing along, celebrating milestones, and modeling language. The engine already fires events the companion can listen to.

🎉

Celebration triggers

The celebration system will watch for conversation milestones — first 3-turn conversation, new word combinations, returning to a word after time away — and respond with contextual celebrations.

🌎

All scenes

In M4, every routine scene (Bath, Bedtime, Getting Dressed, Playtime) gets the same follow-up depth as Kitchen. The engine scales without changes — only the content library grows.