Co-Constructing AI Boundaries Framework Component - Integration

Definition in This Study

Integration refers to how students incorporated, revised, resisted, or transformed AI outputs into their final work. This component captures the boundary of authorship and trust—the most visible evidence of Boundary-work in the entire framework.

Mollick & Mollick (2023) Connection

The Integration component resonates with M&M's core mandate:

• Students must "remain the human in the loop"
• Avoid outsourcing analysis to AI
• Actively oversee and complement AI output with human perspectives

Key M&M Principle: "Students should actively oversee the AI's output, check with reliable sources, and complement any AI output with their unique perspectives and insights."

This principle is most directly observable in how students integrate (or fail to integrate) AI outputs into their final work.

What This Component Analyzes

Primary Focus

• Transformation: Evidence of modification, revision, rewriting
• Rejection: Explicit refusal to use AI output
• Incorporation: How AI text appears in final work
• Attribution: How students mark AI vs. human contributions
• Voice: Whose voice is present in the final product?

Secondary Focus

• Degree of human intellectual contribution
• Evidence of synthesis (not just stitching AI outputs together)
• Maintenance of authorial control

Agency in Integration: The Authorship Decision

This component captures Agency over Final Product:

Integration Strategies: A Continuum

Low Agency ←―――――――――――――――――――――――――――――――――→ High Agency

Verbatim    Light      Moderate     Heavy         Complete      Rejection
Copy-Paste  Editing    Revision     Rewriting     Rewriting     (Non-use)
    |          |           |            |              |             |
AI is      AI is      AI is       AI is         AI is       AI not
Author     Author     Co-author   Draft         Thinking    Used
                                  Creator       Partner

Boundary-work in Integration

Integration is where Boundary-work becomes most visible:

Boundary-Setting Actions

Key Analytic Questions

When coding Integration, ask:

1. Transformation Level:

   • How much did the student change the AI output?
   • Is it recognizably the AI's language or the student's?

2. Voice:

   • Whose voice is present in the final text?
   • Can you distinguish human from AI?

3. Authorship:

   • Who is the author of the final product?
   • Did the student maintain intellectual control?

4. Attribution:

   • Does the student mark what came from AI?
   • Is there transparency about AI use?

5. Synthesis:

   • Did the student integrate multiple sources (AI + human)?
   • Is there evidence of original thinking?

Examples from Data

High Agency Integration: Heavy Transformation

AI Output:
"Critical literacy emphasizes the importance of questioning
texts and understanding power dynamics in society."

Student Final Text:
"Rather than accepting texts as neutral conveyors of
information, critical literacy educators—following Freire
(1970) and Janks (2010)—position reading as an inherently
political act. Students learn to interrogate whose voices
are centered, whose are marginalized, and what ideological
work texts perform."

Analysis:

• Complete rewriting
• Human voice dominates
• Added citations and theoretical framing
• Moved from general to specific/scholarly
• AI served as thinking partner, not author

Medium Agency Integration: Moderate Revision

AI Output:
"Teachers should scaffold literacy instruction to support
diverse learners."

Student Final Text:
"Teachers should scaffold literacy instruction to support
diverse learners, particularly through culturally sustaining
pedagogies (Paris & Alim, 2017) that honor students'
linguistic and cultural repertoires."

Analysis:

• Kept AI sentence structure
• Added scholarly citation and specificity
• Mixed voice (AI base + human enrichment)

Low Agency Integration: Verbatim Copy-Paste

AI Output:
"Reading comprehension strategies include predicting,
questioning, clarifying, and summarizing."

Student Final Text:
"Reading comprehension strategies include predicting,
questioning, clarifying, and summarizing."

Analysis:

• No transformation
• AI voice present
• No human contribution visible
• AI positioned as author

Highest Agency: Rejection

AI Output:
"The achievement gap can be closed through more rigorous
standards and accountability measures."

Student Response (in reflection):
"I rejected this output entirely because it frames the
issue through deficit thinking. Instead, I researched
opportunity gap frameworks (Ladson-Billings, 2006) and
wrote the section from scratch."

Analysis:

• Complete rejection
• Explicit rationale (deficit thinking)
• Student researches alternatives independently
• Strong epistemic stance

Connection to Epistemic Stance

Integration reveals epistemic stance about knowledge authority:

• AI-Authoritative: AI text incorporated verbatim; AI positioned as knowledge producer
• Self-Authoritative: Heavy transformation; human positioned as knowledge producer
• Co-Constructed: Synthesis of AI suggestions with human ideas; shared authorship

Coding Categories for Integration

Relationship to Other Framework Components

• ← Co-Constructing AI Boundaries Framework Component - Outputs: Output quality influences integration approach
• → Co-Constructing AI Boundaries Framework Component - Reflection: Students may reflect on integration decisions
• Integration is the culmination of earlier components (Inputs → Prompts → Outputs → Integration)

The Critical Question: Who Is the Author?

Integration analysis ultimately asks: Who is the author of this knowledge?

M&M's framework assumes students must remain authors, even when using AI tools. Integration evidence reveals:

• Whether students maintained authorship
• Whether they outsourced intellectual labor
• Whether they "remained the human in the loop"

Pedagogical Implications

Teaching effective integration:

1. Model transformation techniques (paraphrase, rewrite, synthesize)
2. Discuss authorship and intellectual property
3. Practice voice analysis (AI vs. human)
4. Teach attribution norms
5. Provide revision examples (AI draft → human final)
6. Emphasize transformation as learning (not just efficiency)
7. Make integration visible through reflection

Special Case: Citation and Attribution

How students cite or attribute AI contributions reveals their understanding of authorship and boundaries:

Data Collection Notes

Where to find evidence:

• Final projects: Compare to NotebookLM outputs
• Track changes: If using collaborative docs
• Reflections: Student descriptions of integration process
• Voice analysis: Stylistic comparison (AI vs. student writing)
• Citation practices: How AI is acknowledged

Triangulation:
Compare NotebookLM outputs → Final project text → Student reflections

Related Notes

• Analytic Framework for AI Human Meaning-Making Practices
• How learners should engage Large Language Models framework
• Agency
• Boundary-work
• Epistemic Stance