AI Decoded: Inside Neural Networks

The Challenge: Demystifying "The Black Box"

AI education for youth often stays on the surface—teaching prompt engineering without explaining the underlying mechanics. For iUrban Teen, the goal was different. We needed to get past the prompts and break down how a Neural Network actually processes data.

I needed to explain complex concepts like Input Layers, Hidden Layers, and Back-propagation to middle and high schoolers in a 50-minute sprint without relying on heavy mathematics or coding.

The Solution: The "Human AI" Simulation

I designed an "unplugged" curriculum where students stop being passive users and become the algorithm.

Neural Node Operations

The session breaks the class into specialized "Node Teams" (Subject, Setting, Vibe, Style). These teams simulate the processing layers of a real machine learning model:

• Structured Processing: Teams use physical worksheets to decode visual data into specific parameters.
• The Generation Phase: Student-led inputs are fed into a live AI generator to produce immediate results.
• Iterative Learning: Much like real "Back-propagation," students analyze errors in the output and "re-train" their logic to improve the next generation.

Program Enablement Kit

To make sure this program could scale and actually hit the mark, I built out a full resource suite for educators and students:

• Interactive Presentation: A visual anchor designed for high engagement and energy not just lecture mode.
• Operational Worksheets: Structured docs that walk the "node" teams through their specific logic layers step by step.
• AI Literacy Survival Kit: A takeaway guide containing ethical frameworks, prompt structures, and verified AI resources.

The Results

The workshop turns a black box into something students can see, touch, and argue about. Advanced ML concepts landed through physical systems and collaboration — no math, no code. And the biggest shift: students stopped seeing AI as magic and started seeing it as a structured system they can understand and control.

Learning Design Framework

• Bloom's Cognitive Levels: Remember → Understand → Apply → Analyze
• Gagne's 9 Events: Achieved 8 of 9 instructional events with built-in feedback loops
• Learning Objectives: 5 mapped objectives from basic recall (explaining neural networks) through analysis (recognizing their role in shaping AI systems)
• Knowledge Transfer: AI Literacy Survival Kit ensures application beyond the session
• Measurement: Hands-on simulation provides immediate feedback; designed exit ticket validates learning outcomes