Pages

▼

Friday, March 20, 2026

Using AI Agent like Google's Antigravity to build - Moulding Bread Experiment: Building a Better Science Class: From Prompt to Pedagogy with AI

 

✨ Refined Version (Sharper, More Impactful)

Building a Better Science Class: From Prompt to Pedagogy with AI

In today’s EdTech landscape, one persistent challenge is finding interactive tools that do more than demonstrate concepts—they must actively teach thinking processes.

One such concept is Fair Testing: changing only one variable at a time to ensure valid conclusions. Rather than adapting an existing tool, we designed a custom interactive using generative AI—guided not by code-first thinking, but by pedagogy-first prompting.

The result was a Mould Growth Investigation simulation—a web-based interactive that does not just simulate science, but enforces scientific thinking.

🔹 Phase 1: From Idea to Simulation

We began with a structured prompt to define the core system:

https://vle.learning.moe.edu.sg/community-gallery/module/edit/fe0f1a1b-dc09-4c57-97dc-96031d1c18f7/section/102355123/activity/102355127

  • Two bread samples (A & B)

  • Variables: Temperature, Moisture, Bread Type

  • 7-day mould growth simulation

The AI generated a working sandbox.
But at this stage, it was interactive—but not instructional.

Students could click randomly and still “complete” the activity without understanding anything meaningful.

🔹 Phase 2: Embedding Scientific Thinking (Fair Test Logic)

We shifted the prompt from features to learning outcomes:

  • Add a Run Simulation gate

  • Detect number of variables changed

  • Provide immediate feedback on validity

Now, the system actively evaluates thinking:

  • ❌ Multiple variables changed → “Not a fair test”

  • ✅ One variable changed → identifies:

    • Independent variable

    • Controlled variables

    • Dependent variable

👉 This transformed the simulation into a formative assessment engine.

🔹 Phase 3: Scaffolding Inquiry (Hypothesis + Journal)

Next, we embedded scientific habits:

This introduced:

  • Metacognition (“What do I predict?”)

  • Progression (“What have I already tested?”)

👉 The simulation now supports inquiry-based learning, not just experimentation.

🔹 Phase 4: Teacher Visibility & Evidence of Learning

Finally, we added teacher-facing affordances:

  • Real-time analytics log

    • Tracks student actions

    • Reveals misconceptions (e.g., repeated invalid tests)

  • Screen recording

    • Students submit evidence of thinking process

    • Supports assessment beyond final answer

👉 This bridges learning → evidence → assessment


🎯 The Final Product

The simulation now:

  1. Forces hypothesis formation

  2. Enforces fair testing principles

  3. Provides immediate feedback on scientific validity





  5. Tracks conceptual progress via journal

  6. Captures learning data and student evidence

This is no longer a simulation—it is a guided scientific thinking environment.

💡 Key Insight for Educators

The breakthrough was not the AI.

The breakthrough was how the AI was prompted.

Instead of asking: “Build me a simulation”
We asked: “Build me a system that teaches thinking.”

👉 AI becomes powerful only when the teacher acts as:

  • Pedagogical designer

  • Assessment architect

  • Learning scientist


Tool used:
https://antigravity.google/


Improving Teaching Value of Mould Growth Investigation

This plan outlines the changes required to enhance the pedagogical value of the "Mould Growth Investigation - Fair Test" interactive.

Proposed Changes
index.html
Hypothesis Section: Add a drop-down or short text area before the run buttons for the student to state their hypothesis. (e.g., "I predict that [Bread A] will have more mould").
Time Controls: Replace instantaneous run with a "Days" progress or animated sequence over 7 days.
Science Journal: Add a new panel to track which variables the student has successfully tested (Temperature, Moisture, Bread Type).
script.js
Hypothesis Logic: Validate that they made a hypothesis before running.
Time Animation: Update the 
runSimulation
 to simulate days passing (e.g., using setInterval or standard async delays) to progressively show mould.
Enhanced Feedback: Update 
checkFairTest
 to identify exactly which variables were changed, and explicitly state "Independent Variable", "Controlled Variables", and "Dependent Variable" if the test is fair.
Journal Tracking: Write logic to check off items in the Science Journal once a specific fair test is completed.
styles.css
Add styles for the Science Journal checklist.
Add styles for the Hypothesis section.
Add styles for the Time progress indicator.
Verification Plan
Browser Testing
Open 
index.html
 in the browser subagent.
Verify hypothesis can be set.
Run a fair test (e.g., only change Temp) and verify the 7-day animation.
Verify vocabulary explicitly shown in the results area.
Verify that the Science Journal updates correctly to check off "Temperature".
Run an UNFAIR test (change 2 variables) and verify the scaffolding error message correctly identifies the two changed variables.


Testing Plan for Interactive_20260318115345
 Open index.html
 Click "Run Simulation" without a hypothesis and observe warning
 Select a hypothesis
 Setup a Fair Test (Change ONLY temperature: Bread A: Room, Bread B: Warm)
 Run Simulation and watch days 1-7
 Observe output vocabulary ("Independent Variable," "Controlled Variables," "Dependent Variable")
 Scroll to Science Journal and check if Temperature test is checked off
 Verify visual appearance via screenshot


at

No comments:

Post a Comment