Science Training (eduKateSG) — Fence Learning Systems for PSLE + Lower Secondary Science

TITLE: Science Training (eduKateSG) — Fence Learning Systems for PSLE + Lower Secondary Science
SUBTITLE: Reach P3 reliability in Science under load: concepts + keywords + data + experiments
VERSION: CivOS Unified Spec v1.x
MODE: Almost-Code / WordPress paste-ready
SCOPE: Z0 learner + Z1 parent + Z2 tutor/school; PSLE (Primary) + Sec foundation
GOAL: Science becomes precise output under time pressure (not “memorise notes”)
========================================================
0) DEFINITION LOCK BOX
========================================================
Science mastery = reliable explanation and decision-making under constraints.
Science.P3 :=
A learner can:
- understand cause→effect (not just words),
- use correct scientific keywords precisely,
- interpret data (tables/graphs),
- answer exactly what is asked,
- and do so under time pressure with low volatility.
Fence principle:
Science collapses when:
- concept holes are hidden by memorisation,
- keywords are missing under load,
- questions are misread,
- data handling fails,
and TTC shrinks.
So Science training must be instrumented and repaired, not “study more”.
========================================================
1) THE SCIENCE STACK (WHAT SCIENCE IS MADE OF)
========================================================
Science = 4 sub-systems:
S1 Concept Model:
- cause → effect, mechanism, conditions
S2 Keyword Precision:
- correct terms, correct phrasing, correct specificity
S3 Data & Evidence Handling:
- tables, graphs, patterns, comparisons
S4 Scientific Process / Experiments:
- variables, fair test, conclusion from evidence
If any sub-system fails, marks leak fast.
========================================================
2) UNIVERSAL DASHBOARD (SCIENCE EDITION)
========================================================
Metrics:
- TTC: days to next test/exam
- Buffer: sleep/routine/stress slack (science needs calm precision)
- R: are weak nodes shrinking weekly?
- Θ: reaction time after a drop (24–72h)
- Λ: routine enforceability same day
Science-specific sensors:
- Misread count (wrong focus)
- Keyword miss count (missing “evaporation / condensation / conserve / dissolve” etc.)
- Explanation clarity (1–5)
- Data question accuracy
- Variable identification accuracy (IV/DV/controls)
========================================================
3) Z0 ROUTER: FAST DIAGNOSIS (10 MINUTES)
========================================================
Step 1: Identify the failing lane:
- Concept model (S1)?
- Keywords (S2)?
- Data (S3)?
- Experiments/process (S4)?
Step 2: Classify dominant failure mode (pick ONE):
F1 Concept hole (memorised but not understood)
F2 Keyword weakness (vague answers)
F3 Misread / command-word drift (answers wrong focus)
F4 Data interpretation failure (graph/table confusion)
F5 Experiment template failure (IV/DV/controls)
F6 Transfer failure (new context breaks understanding)
F7 Time collapse (rushing → careless)
Step 3: Fence (truncate):
- shrink to the smallest failing node
- stop-loss when stuck
- protect buffer if low
Step 4: Stitch (repair loop):
- apply matched loop 4–7 days
- verify with a timed mini-set
========================================================
4) FAILURE MODES → SENSORS → REPAIR LOOPS
========================================================
--------------------------------------------------------
F1 Concept Hole (no mechanism)
--------------------------------------------------------
Sensors:
- can recite definition but cannot explain “why”
- fails on “predict / explain” questions
Repair Loop SR1 (Micro-Model):
- build a 3-line model:
  Condition → Process → Result
- draw a tiny cause-effect map (5 nodes max)
- do 6 questions that require explanation
Fence:
- truncate: stop new chapters; repair this micro-model first
--------------------------------------------------------
F2 Keyword Weakness (vague answers)
--------------------------------------------------------
Sensors:
- teacher comments: “too general”, “not specific”
- loses marks in open-ended
Repair Loop SR2 (Keyword Bank + Answer Frames):
- keyword bank per topic (10–20)
- answer frames:
  “The reason is…”
  “This causes…”
  “Therefore…”
- daily 5 keywords used in 5 sentences
Fence:
- reduce writing length but increase precision (quality fence)
--------------------------------------------------------
F3 Misread / Command-word Drift
--------------------------------------------------------
Sensors:
- answers correct concept but wrong question
- misses “compare / explain / state / suggest”
Repair Loop SR3 (Question Parsing Template):
- underline command words
- restate question in own words
- identify required format:
  - 1 word? 1 sentence? 2 differences?
Fence:
- stop-loss: if unsure > 60s → parse again before writing
--------------------------------------------------------
F4 Data Interpretation Failure (tables/graphs)
--------------------------------------------------------
Sensors:
- wrong trend, wrong axis, wrong comparison
Repair Loop SR4 (Daily Data Micro-Sets):
- 1 table + 1 graph daily (5–8 minutes)
- answer 3 standard prompts:
  “highest/lowest”, “trend”, “difference”
Fence:
- short daily reps > long occasional practice (time fence)
--------------------------------------------------------
F5 Experiment / Process Template Failure
--------------------------------------------------------
Sensors:
- cannot identify IV/DV/controls
- cannot design fair test
Repair Loop SR5 (Experiment Template):
- always fill:
  IV / DV / constants / method / conclusion
- 1 experiment scenario daily
Fence:
- truncate: do template only; no extra writing until stable
--------------------------------------------------------
F6 Transfer Failure (new context breaks understanding)
--------------------------------------------------------
Sensors:
- can answer textbook-style, fails on novel scenario
Repair Loop SR6 (Context Swaps):
- same concept, 3 different contexts
- “predict + explain” in each context
Fence:
- no new topics until transfer improves (R<1)
--------------------------------------------------------
F7 Time Collapse (rushing errors)
--------------------------------------------------------
Sensors:
- leaves blanks
- messy answers, missing keywords under time
Repair Loop SR7 (Timed Micro-Sets):
- 6 MCQ in 6 minutes
- 2 OEQ in 8–10 minutes
- then mixed mini-paper
Fence:
- “finish fence”: last 5 minutes reserved for check keywords + command words
========================================================
5) SCIENCE TOPIC CLUSTERING (FOR FAST REPAIR)
========================================================
Cluster C1: Matter & Materials
- states, changes, dissolving, separation
Cluster C2: Water & Heat
- evaporation/condensation, temperature, heat transfer (basic)
Cluster C3: Plants & Animals
- life cycles, adaptations, systems (primary level)
Cluster C4: Forces & Energy (primary)
- simple machines, friction, light/sound basics
Cluster C5: Earth & Environment
- water cycle, weather, human impact basics
Rule:
When TTC is short, do not “cover everything”.
Pick the cluster where marks leak most (R>1) and repair it first.
========================================================
6) Z1 PARENT TOOLKIT (SCIENCE)
========================================================
Parent role = enforce fences and protect buffers, not “lecture science”.
Home fences:
- fixed short daily slot (15–25 min)
- require child to:
  - underline command words (F3)
  - state 3 keywords to include (F2)
  - give one “because” sentence (F1)
- sleep fence before tests
Low-effort prompts:
- “What is the condition?”
- “What process happens?”
- “What is the result?”
(These train SR1.)
========================================================
7) Z2 TUTOR/SCHOOL ENGINE (TRANSFER UNDER LOAD)
========================================================
Weekly cycle:
- diagnose the dominant failure mode
- 4 days targeted repair loop (SR1–SR7)
- 1 timed mini-set
- redo protocol: rewrite the weakest answer using keywords + command words
- every 2–4 weeks: mini-simulation (increase frequency as TTC shrinks)
Non-negotiables:
- daily data reps (SR4)
- keyword precision training (SR2)
- parsing drills (SR3)
- transfer swaps (SR6)
========================================================
8) MINIMUM VIABLE 14-DAY PLAN (TTC-RED)
========================================================
Day 1–3:
- SR3 parsing + SR2 keyword bank (stabilize mark leakage fast)
Day 4–7:
- SR1 micro-models for top 2 weak concepts + SR4 daily data
Day 8–10:
- SR5 experiment template + SR6 context swaps
Day 11–14:
- timed micro-sets (SR7) + 2 mini-papers + redo weakest answers
Rule:
Shrink scope until R<1 (weak nodes shrink weekly).
========================================================
9) FAILURE MODE TRACE (REQUIRED)
========================================================
Trace:
concept holes hidden by memorisation + keyword weakness + TTC short
→ misread + vague answers under time
→ R>1 (same leaks repeat)
→ late reaction (Θ>1)
→ P2→P0 collapse near exam
Fence Repair:
detect TTC-RED → truncate to SR2+SR3+SR4 first
→ enforce routine same day (Λ<1)
→ fence within 24h (Θ<1)
→ stitch SR1/SR5/SR6 next
→ timed verification (SR7) → R<1 → restore P3 reliability
========================================================
10) WORDPRESS PASTE-READY STRUCTURE
========================================================
H1: Science Training: How to Reach P3 Reliability Under Time Pressure (Fence Learning Systems)
H2: What Science Mastery Really Is (Mechanism + Precision)
H2: The Science Stack (Concepts, Keywords, Data, Experiments)
H2: Dashboard (TTC, Buffer, R, Θ, Λ) + Science sensors
H2: Fast Diagnosis (pick the failing lane)
H2: Failure Modes → Repair Loops (SR1–SR7)
H2: Topic Clusters (how to repair fast)
H2: Parent Toolkit (Z1)
H2: Tutor/School Engine (Z2)
H2: 14-Day TTC-Red Plan
H2: Failure Mode Trace (Short)
H2: Copy/Paste Weekly Plan
========================================================
11) COPY/PASTE WEEKLY PLAN (SCIENCE)
========================================================
Daily (20–35 min):
- 5 min parsing drill (SR3)
- 8 min data micro-set (SR4)
- 10–15 min micro-model or keyword drill (SR1/SR2)
- log 1 recurring leak
Weekly:
- 1 timed mini-paper
- redo 2 weakest answers using keywords + command-word check
END SPEC
========================================================