Free educational resource for instructors and institutions
Developed by Dr. Justin Rose
G3 Scaffolding Framework
A classroom-tested andragogical model for building AI fluency and metacognitive depth in 16-week disciplinary courses through structured, concurrent learning across three cognitive layers.
G3 is grounded in Delta Iota Learning, a framework that treats competence as the product of accumulated, deliberate activations of mental models across different contexts. Delta (small change) and Iota (small amount) together represent how genuine skill develops: not through single exposures or events, but through incremental, sequenced steps that build on what came before. Learning happens simultaneously across multiple cognitive layers, not one at a time. This is why G3 structures the entire 16-week semester to activate the same mental models repeatedly in new contexts, with decreasing instructor support, building transferable competence that lasts. In practice, this principle sets the sequence: each week builds on the last, and support is removed step by step.
How G3 Works: The Five Stages and Three Layers
G3 has three structures that nest. Across the semester, five stages move students from baseline to independent work. Within each week, three cognitive layers run together, and three cognitive methods support them collectively. Every week follows the same four assignments. ANCHOR (prompt design) and AUDIT (output evaluation) are defined under Key Frameworks below. This page serves program leaders (For Program Leaders), instructors (The Weekly Reflection Report), and students (the weekly sequence below).
The Semester: Five Stages
Week 1StartupBaseline without AI. Initial AI check discussed in the lecture.
→
Weeks 2-5GuidedInstructor models. Students use complete copy-and-paste prompts.
→
Weeks 6-10GroundedStudents write prompts from the ANCHOR acronym.
→
Weeks 11-15GraduatedANCHOR and AUDIT as a reminder. Students decide whether to use AI.
→
Week 16ReviewStudents compare final work with their Week 1 baseline.
Each Week: Four Assignments
Part 1Core instructionThe weekly topic, taught through video and lecture.
→
Part 2Discussion boardPeer assessment without AI. Peers support or add a view, not oppose it.
→
Part 3AI reflectionIndependent review, submitted as the Weekly Reflection Report.
→
Part 4CapstoneThe Takeaway feeds the collective capstone.
The Capstone
The capstone is a collective project that students build during the Graduated stage, integrating all three layers. Students may use AI to coordinate and manage the project, and they present the finished work as a video. Each week's Takeaway is a building block for it, so the weekly work is designed to accumulate rather than stand alone.
Across the semester, five instructional stages move students from baseline to independent work. Within each week, the three cognitive layers run together, and the three cognitive methods support them collectively. In Week 1, students establish baselines across all three layers. In Weeks 2-5 (Guided), the instructor models everything and provides templates; students practice following structured examples. In Weeks 6-10 (Grounded), templates are removed and students design their own approaches using the tool frameworks, with instructor feedback and calibration. In Weeks 11-15 (Graduated), scaffolds fade and students work autonomously, deciding whether and how to use AI in complex, real-world tasks. In Week 16, students review their growth by comparing final work with their Week 1 baseline. Each layer reinforces the others: peer discussion (Layer 1) creates the foundation for AI critique (Layer 3), and metacognitive reflection (Layer 2) connects them both. The result is competence that transfers beyond the course.
The G3 Architecture: Three Layers, Five Stages
This geometric architecture shows how subject knowledge (domain), metacognition, and AI fluency meta-schemas interact and evolve together across the five instructional stages.
The Five Instructional Stages
Each stage represents a distinct learning context, with the same mental models applied in new ways across all three layers.
Week 1Startup
Find your baseline understanding of the domain
Check your current learning habits and study style
Complete all Week 1 work without AI to show your writing style and establish your experience with the topic
Attend the initial AI check in the Week 1 lecture: say whether you have used AI and discuss how it will affect your work. No AI use is required.
Weeks 2-5Guided
Observe step-by-step modeling with think-alouds
Use complete copy-and-paste prompts from your instructor, then evaluate each output with AUDIT
Practice with instructor-provided templates and feedback
Weeks 6-10Grounded
Write your own prompts from the ANCHOR acronym alone, with no provided examples
Evaluate each AI output with AUDIT
Receive feedback and calibration on your approaches
Weeks 11-15Graduated
Complete complex tasks autonomously
ANCHOR and AUDIT are provided as a reminder; decide whether to use AI, then complete the report
Demonstrate integrated competence across all three layers
Week 16Review
Final assessment showing competence across all layers
Reflection on growth and learning progress
Plan how you'll apply frameworks after the course
Key Frameworks to Teach
Two core tools your students will master: ANCHOR for writing prompts to AI, and AUDIT for evaluating AI output.
ANCHOR: Prompt Design Framework
A
Aim
What is the specific goal of this task, and what would success look like?
N
Narrative
What context, background, or story does the AI need?
C
Constraints
What are the boundaries? (format, length, tone, audience)
H
Method
How should the AI reason through the task, and what steps should it follow?
O
Own thinking
What do you think before consulting the AI? Where do you currently stand?
R
Review criteria
How will you judge whether the response is good enough?
AUDIT: AI Output Evaluation Framework
A
Accuracy
Are the facts and claims correct and verifiable?
U
Underlying reasoning
Does the logic hold, and does it follow from the evidence?
D
Departures
Where does the output depart from the case, your thinking, or the disciplinary sources?
I
Inclusion
Whose perspectives, cases, or constraints are missing from the output?
T
Takeaway
What will you keep, change, or apply elsewhere?
Understanding the Three Cognitive Layers
Layer 1: Subject Knowledge (Domain)
What students think about (disciplinary concepts, frameworks, ways of reasoning). Without domain grounding, AI use is mechanical. G3 keeps domain thinking central throughout: students use AI as a thinking partner, not a replacement for thinking.
Layer 2: Metacognition
What students think about their thinking (awareness of cognitive processes, ability to monitor and adjust reasoning). Metacognition bridges domain knowledge and AI fluency. Students who struggle with articulating how they think may struggle with evaluating where AI output ends and their own thinking begins.
Layer 3: AI Fluency Meta-Schemas
Mental models for evaluating AI: recognizing capabilities, limitations, bias, and output quality. Without explicit schemas, students either over-trust or dismiss AI. G3 builds schemas that transfer to novel tools and domains.
The Weekly Reflection Report
Start by mapping your existing course activities to the G3 framework. Identify where you model, where students practice, and where they work independently. Then layer in explicit scaffolds for all three cognitive dimensions simultaneously.
Each week, students submit one page as their weekly assignment. It is a fixed format, so students know what to produce and instructors know where to look.
Weekly Reflection Report: Week [#], [Topic]
Cognition. What I learned this week about the topic. Name one concept I can now explain in my own words, and one I still cannot.
Metacognition. My own thinking before I consulted AI, and where my reasoning changed afterward. What did I notice about how I think?
AI Fluency.
Tool and date used:
ANCHOR prompt (paste the exact prompt):
AUDIT, one quoted claim or omission per element, with my judgment in my own words:
Accuracy:
Underlying reasoning:
Departures:
Inclusion:
Takeaway:
Graduated only: if you did not use AI, skip the ANCHOR and AUDIT fields, justify the choice in one paragraph, and complete Metacognition in full.
Looking ahead. What I am aware of as I move into next week, what I expect to find difficult, and what support or change would help me. Name one thing I will do differently. This section is a suggestion for the student and is not part of the grade; it belongs to the metacognitive reflection.
Example: The Prompt a Student Copies
This is a Guided-stage example from the conflict management course, using the ladder of inference. The student fills in their own thinking first, pastes the text in the box into the AI tool, answers three questions, and checks the one-page report against their own answers before submitting it.
Aim: Help me see where a workplace conflict started in each person's reasoning, using the ladder of inference, well enough to act on it.
Narrative: I am a manager. Two people on my team, [Name A] and [Name B], have disagreed for three weeks about who owns a client handoff. [Name A] believes [Name B] ignored the deadline. [Name B] believes [Name A] held back information. Details: [paste the facts you have].
Constraints: Keep each reply under 150 words until the report. Use plain language. Do not assume facts I have not given you. List any assumptions separately.
Method: First, walk through the ladder of inference for each person, in order: observable data, data they selected, meaning they added, assumptions, conclusions, beliefs, and actions. For each rung, say which person is on it. Then name the first rung where their accounts diverge.
Questioning: Then ask me exactly three questions, one at a time, to deepen my understanding of the conflict. Wait for my answer before asking the next question. Each question should test an assumption I have made or ask me to explain my own reasoning. Do not write the report until I have answered all three.
Report: After my third answer, write a one-page report for me to submit, with these headings: Cognition (one idea I can explain and one I still find hard), Metacognition (my Own thinking and where it changed), AI Fluency (the tool used, this prompt, and for each of Accuracy, Underlying reasoning, Departures, Inclusion, and Takeaway, one quoted claim or omission from your analysis, with a blank line for my judgment in my own words), and Looking ahead (what I should watch for next week). Use only my answers and my Own thinking. Do not invent claims I did not make.
Own thinking: [Before pasting, write your own view here: where do you think the conflict first diverged, and why?]
Review criteria: A good answer names the specific observable data each person selected, separates their assumptions from the facts, and gives me one question I could ask each person to test an assumption.
Grading the Weekly Report
Grade the student's thinking. Do not grade how polished the writing is, and do not grade how much AI they used.
Rate three parts as Meets or Developing: Cognition, Metacognition, and AI Fluency.
Cognition: Meets if the student names one idea they can explain and one idea they still find hard.
Metacognition: Meets if the student states a clear position in Own thinking before reading the AI. Developing if it only repeats the topic.
AI Fluency: Use the AUDIT rubric below. Meets if each part points to something specific in the AI output.
Looking ahead is not graded. You can comment on it, but it does not change the rating.
The rules change by stage. In Guided, the prompt is given, so grade the AUDIT entries and Own thinking. In Grounded, grade the prompt the student wrote, plus Own thinking. In Graduated, the prompt is a reminder. If the student used AI, grade it as in earlier weeks. If not, grade the reason they gave and their Metacognition.
Before the term, have two instructors score the same three sample reports. Talk through any part where they disagree. Then give one sentence of feedback for each part, and one question for next week.
Sample Rubric for AUDIT Reflections
Score each element as Meets or Developing. This is a sample; adjust the criteria to your discipline.
AUDIT element
Meets expectation
Developing
Accuracy
Flags at least one claim as unverified or incorrect and states how it was checked against a source.
Accepts claims without checking, or flags errors without a source.
Underlying reasoning
Explains whether the AI's logic follows from the evidence and names the step that holds or fails.
Restates the AI's conclusion without evaluating how it was reached.
Departures
Identifies where the output departs from the case, the student's own thinking, or disciplinary sources, and judges whether it matters.
Lists differences without judging whether they matter.
Inclusion
Names at least one perspective, stakeholder, or constraint the output leaves out, and explains the effect on the answer.
Calls the output balanced or fine without identifying gaps.
Takeaway
States one specific thing the student will keep, change, or apply elsewhere, with a reason.
Summarizes the output without deciding how to use it.
Depending on the demands of your course, you can encourage students to complete supplemental AI fluency training, such as Anthropic's AI Fluency Fundamentals course, which awards a certificate. Completion can be offered as extra credit or as professional development, and it does not replace any part of G3.
For Program Leaders and Curriculum Coordinators
G3 integrates AI instruction into existing disciplinary courses without creating a separate AI track. It scales across departments by giving every instructor a common andragogical architecture and shared frameworks (ANCHOR, AUDIT) that students encounter consistently.
→
Reduces redundancy by embedding AI fluency into existing 16-week courses rather than adding standalone modules
→
Provides consistent andragogical language across all disciplines so students transfer learning more easily
→
Scales with minimal professional development: one training on G3 architecture, and instructors adapt it to their discipline
The Research Behind the Practices
Three cognitive methods support the three layers together in every week of G3. Each week applies all three, and the stage determines how much support students receive. G3 has been tested in classrooms, and curriculum impact feedback has shown the model effective. The research behind each method comes from controlled studies, and those effects are smaller and less certain in full semester courses, so G3 treats the methods as design principles rather than as results it has demonstrated itself.
Spaced Retrieval
Recalling material from memory across spaced sessions improves retention more than rereading it (Roediger and Karpicke, 2006). G3 pairs this with knowledge chunking, which organizes material into units that build on one another.
In G3: Each week builds on the one before. Students recall the previous week's material at the start of the next, and every week contributes to the collective capstone, so earlier material returns in later work.
Interleaving
Mixing different cases within a practice session builds the ability to recognize which approach a problem calls for (Rohrer and Taylor, 2007).
In G3: Each weekly topic is practiced through varied scenarios, industries, and situations. This builds schemas of how the topic is experienced and applied across contexts, not one case repeated.
Self-Explanation
Stating the reasoning behind each step, not just the answer, improves understanding (Chi and colleagues, 1989).
In G3: Think-alouds, in which an instructor narrates their decisions while working, model this in Weeks 2-5. ANCHOR has students reflect on the topic and their own thinking before they take an AI-recommended solution or analysis. AUDIT then requires them to analyze the output after it is produced, apply disciplinary knowledge, and judge whether the AI's reasoning is valid.
What This Looks Like in Weeks 5 and 6
This example is adapted from Dr. Rose's course on conflict management and resolution for managers. Each week has one topic and one conflict principle. Each week follows the same sequence: core instruction, a discussion board without AI, and an independent AI reflection.
Week 5 (Guided): Negotiation Techniques in Management. Conflict principle: the Collaborating mode of the Thomas-Kilmann conflict model is high assertiveness and high cooperation, working together to find solutions that fully satisfy both parties. It takes time and trust, but it produces durable outcomes. Every prompt is provided for students to copy and paste, so they do not write prompts yet.
Retrieval. Students start by stating the conflict modes covered in earlier weeks from memory, before the provided prompt is shown.
Interleaving. The core lecture presents negotiations from different settings, such as a vendor contract renegotiation, a staffing budget dispute between two managers, and a schedule change negotiated with a team. Students run the same provided prompt on each and compare whether the AI's proposed solutions reached the Collaborating mode or settled for a compromise.
Self-explanation. On the discussion board, which is AI-free, each student assesses a peer's proposed option, explaining why it does or does not fully satisfy both parties and what time and trust would be needed to reach it. Responses support the peer's stance with another reason or offer a different point of view, rather than opposing it. In the AI reflection, AUDIT has students check the AI output, apply disciplinary knowledge, and judge its validity, writing one sentence on each claim they accept or reject.
Week 6 (Grounded): Mediation and Facilitation. Conflict principle: reconciliation moves beyond conflict resolution to a restored relationship. It often requires acknowledgment of harm, demonstrated change, and the choice to release resentment. The copy-and-paste prompts are removed. Students receive only the ANCHOR acronym and write their own prompt for this topic.
Retrieval. Students rebuild a prompt from the acronym and what they recall of the Week 5 negotiations, without a model in front of them. They then state what separates the Collaborating mode from a settlement.
Interleaving. The Week 6 core lecture alternates between two mediation cases from different settings, such as a hiring disagreement between two team members and a shift conflict between two crews. Students write a prompt for each and compare whether the case reached reconciliation or stopped at a settlement.
Self-explanation. On the discussion board, which is AI-free, each student explains what acknowledgment of harm, demonstrated change, and release of resentment would look like in a mediated case, and why a settlement alone would not meet the standard. Responses build on a peer's assessment by supporting it or adding a different perspective. In the AI reflection, AUDIT checks whether the AI output treats a settlement as reconciliation.
The shift from Week 5 to Week 6 is deliberate. Week 5 shows students what a good prompt contains, and Week 6 asks them to write one. The work from both weeks feeds the collective capstone, which students present as a video.
About the Developer
Dr. Justin Rose
With nearly two decades of experience across higher education, private industry, and workforce development, I have designed and implemented training at every level from individual classrooms to statewide and national programs. My work spans instructional design, program management, organizational consulting, and AI integration in education. G3 emerges from this breadth of experience, bridging andragogical rigor with practical implementation needs.
G3 Framework Adaptation Prompt
Copy this prompt and paste it into Claude (or your preferred AI assistant) to get personalized guidance on adapting G3 to your course:
The G3 Scaffolding Framework builds AI fluency alongside disciplinary learning across 16 weeks. Five stages move the semester from baseline to independent work: Startup (Week 1, no AI), Guided (Weeks 2-5, the instructor models and provides copy-and-paste prompts), Grounded (Weeks 6-10, students write prompts from the ANCHOR acronym), Graduated (Weeks 11-15, scaffolds fade, ANCHOR and AUDIT are provided as a reminder, and students decide whether to use AI), and Review (Week 16, students compare final work with their Week 1 baseline). Three cognitive layers run together in every week: Layer 1 (Subject Knowledge, Domain), Layer 2 (Metacognition), and Layer 3 (AI Fluency). Three cognitive methods support the layers collectively: spaced retrieval with knowledge chunking, interleaving across varied scenarios, and self-explanation. Each week has a weekly topic, core instruction (video and lecture), an AI-free discussion board where peers support or add to each other's assessments, an independent AI reflection submitted as a one-page Weekly Reflection Report, and a link to the capstone, which students present as a video. Two tools drive the work: ANCHOR (Aim, Narrative, Constraints, Method, Own thinking, Review criteria) for prompt design, and AUDIT (Accuracy, Underlying reasoning, Departures, Inclusion, Takeaway) for evaluating AI output. Full framework details: g3.lpsnc.com
I want to adapt the G3 Scaffolding Framework to my course. Ask me 3 to 5 questions, one at a time, about my course structure, student background, disciplinary context, course demands, and my AI policy. Then guide me through implementation week by week, starting with the Week 1 no-AI baseline and showing how support changes from Guided to Review. Where useful, suggest supplemental AI fluency training, such as Anthropic's AI Fluency Fundamentals course, as optional extra credit or professional development.
Copy and paste into any AI assistant (Claude, ChatGPT, Gemini, or similar). The AI will ask you the questions needed to understand your course and guide you through implementation.
Want to explore more about AI integration in instruction?
