Scaffolding Success: AI and the 5-Step Routine in Introductory Physics
Deepak Kapoor, The City College of New York
Bio
My name is Deepak Kapoor, a National Board Certified physics educator and Math for America Master Teacher inspiring New York City students since 2006. After migrating to the US in 2001, I dedicated my career to hands-on, project-based science education. Currently teaching at Gregorio Luperon High School, City College’s STEM Institute, and College Now, I am fiercely committed to empowering diverse learners and motivating the next generation of STEM innovators.
Course Setting
I facilitate a College Now Physics course with City College, hosted at Gregorio Luperon High School. This on-site setting is vital for my students—all English Language Learners who qualify for free lunch. Because these resilient students work to support their families, commuting to a college campus isn't feasible. Offering this foundational course directly at our school removes logistical barriers, granting them equitable access to rigorous, college-level STEM education.
Problem of Practice
While College Now students often grasp fundamental physics concepts within highly structured settings, many lack the proficiency to independently construct and apply mathematical models to represent real-world physical situations. This cognitive gap becomes particularly evident when students must transition from step-by-step, guided examples to navigating complex, open-ended problems. Furthermore, they frequently struggle with synthesizing raw experimental lab data into meaningful algebraic or graphical models, tending to rely on rote formula memorization rather than deep conceptual reasoning.
How might AI-supported practice and feedback help students actively demonstrate deep physics reasoning, so they can independently explain, justify, and apply concepts on Regents-style assessment questions rather than relying on in-class recognition alone?
Strategy
To address this cognitive gap, I embedded a pervasive 5-Step Modeling-First Routine (Diagram, Knowns/Unknowns, Model, Solve, Interpret) into warm-ups, classwork, and labs. Grounded in Cognitive Load Theory and AMTA’s Modeling Instruction, this framework actively trains students to translate physical reality into coherent mathematical narratives. I support this crucial transition through:
Lab-Driven Discovery: Students construct core equations (e.g., V=IR) directly from their own lab data, shifting from verifying textbook formulas to genuine scientific inquiry.
Faded Scaffolding: I gradually remove guided instructional setups to cultivate greater student autonomy and problem-solving confidence.
Just-in-Time Algebra: I provide targeted, five-minute math refreshers exactly when needed to prevent cognitive overload and maintain conceptual focus.
AI Integration
To scale this rigorous framework and individualize instruction, I leveraged AI to generate tiered modeling problems tailored to diverse readiness levels. Additionally, AI will help formulate comprehensive rubrics, produce reasoning-focused example solutions, and build realistic lab datasets across all fundamental physics topics
Documentation
Five-Step Modeling First Routine
Figure #1: Five-Step Modeling First Routine
5-Step Electricity Problems Designed by AI
The document below excerpts an assessment, designed with the help of an AI tool, builds conceptual clarity and problem-solving skills in Current Electricity using a structured modeling approach. Instead of rushing to formulas, students must systematically Diagram the situation, identify Knowns/Unknowns, select the physics Model (e.g., Ohm's or Kirchhoff's laws), Solve calculations, and Interpret the physical reasonableness of their results.
Figure #2: 5 Steps as Applied to Electricty Problems
Kinematics Comprehensive Assignment
This assignment reinforces the "Modeling-First Routine" for College Now and AP Physics students, focusing on 1D/2D motion. By prioritizing diagrams and symbolic modeling over quick calculations, students master free-fall, projectile trajectories, and circular motion. This workflow requires explicit coordinate definition, minimizing algebraic errors and ensuring physical plausibility.
Measuring Impact
Initial Survey of Student Confidence
The initial survey revealed that students lacked confidence and structured routines to tackle high-level questions. They struggled to break complex scenarios into manageable steps—like diagramming or identifying knowns—before attempting to solve. This deficit in a systematic approach was especially prominent in the abstract Electricity unit.
Figure #3: Initial Survey of Student Confidence
Final Survey of Student Confidence
The final survey revealed that the new routine significantly boosted confidence and math skills, making complex word problems easier to approach. Students reported improved comprehension, particularly through diagramming. This positive shift empowered many to help peers, though one respondent noted ongoing struggles with accurate problem-solving.
Figure #4: Final Survey of Student Confidence
Analysis
The implementation of the 5-Step Modeling-First Routine successfully addressed the core Problem of Practice by transforming passive students into active, structured problem-solvers. Pre- and post-survey data from the 20 College Now students (12 boys, 8 girls) demonstrated a clear increase in their confidence and capability when approaching open-ended, lab-based challenges. Grounding the routine in Cognitive Load Theory proved highly accurate; by standardizing the approach (Diagram, Knowns/Unknowns, Model, Solve, Interpret), students were no longer overwhelmed by the blank page. The strategy's impact was significantly amplified by the integration of an AI-powered generator, which rapidly produced targeted cluster questions to establish and reinforce this daily routine.
Analyzing the initial survey data and formative classwork revealed exactly where students were experiencing cognitive friction—most notably in the leap from defining variables to constructing the mathematical model. This insight allowed for dynamic adjustments to the implementation. By leveraging the AI to generate scaffolded problems that heavily targeted this specific transition, students showed measurable improvement in their confidence in translating physical realities into mathematical narratives.
While the routine built crucial foundational skills, the challenge of fading the scaffolding remains. The predictability of the 5-step prompts may inadvertently create a reliance on the structure. Next time, the implementation will feature a more deliberate, gradual release of responsibility. Slowly removing the explicit step-by-step guidance will better test their independent modeling capabilities and ensure they can initiate the framework entirely on their own when faced with novel phenomena.
Recommendation
Key Fading Strategies:
Implement a Gradual "Fading" Protocol: Consolidate the 5 steps into two broader phases ("Physical Representation" and "Mathematical Execution") or use a reference checklist to build student independence.
Recalibrate Your AI Generator: Have AI generate partially scaffolded problem sets that provide the diagram and knowns, forcing students to immediately construct and justify the mathematical model.
Elevate Peer Defense: Require students to present and defend their mathematical models to peers. Verbalizing reasoning solidifies the translation of physical reality into mathematics.
Scale Through Mentoring: Share this data-backed routine and AI strategies with new and student teachers to help them manage cognitive overload in their classrooms.
Consider: which of these fading strategies would be easiest to introduce to your students next week?
How to use it for this strategy: These conversational AI models are the best starting point for creating the core problem sets. You can write a specific prompt instructing the AI to generate physics word problems formatted exactly to your routine (e.g., "Generate three tiered kinematics problems. Do not solve them, but provide a 5-step student template for each: Diagram, Knowns/Unknowns, Model, Solve, Interpret."). You can also ask it to instantly rewrite those same problems with "faded scaffolding" for your gradual release.
How to use it for this strategy: MagicSchool is an AI platform built specifically for educators. Your colleagues can use its Science Lab Generator and Rubric Generator to quickly draft the assessment criteria for the open-ended modeling labs. It also has excellent tools for generating the 5-minute "Just-in-Time Algebra" refreshers, tailoring the math review to the exact physics concept being taught.
How to use it for this strategy: Julius is a powerful AI data analyst. Since a major part of your strategy involves "Lab-Driven Discovery" where students build equations from data, colleagues can use Julius to instantly generate realistic, slightly "messy" lab datasets (like current vs. voltage readings) for any physics topic. Students can then practice extracting mathematical models from this generated data.
How to use it for this strategy: Eduaide is fantastic for instructional design and differentiation. If your colleagues have a complex, college-level physics text or problem, they can plug it into Eduaide to automatically generate tiered, scaffolded versions of the assignment. This is highly effective for supporting ELL students in a College Now environment, ensuring the language doesn't obscure the physics concepts.
How to use it for this strategy: Because the very first step of your routine is "Diagram," having clear visuals is crucial. Edraw is an AI-powered diagram creator that can quickly generate physical situation images (like circuits, force vectors, or block-and-tackle systems). Colleagues can generate partially completed diagrams to include in their problem sets, giving students a visual starting point for their Knowns/Unknowns.