Bloom's AI Collaboration Framework
A cognitive model for human-AI partnership — who thinks what, at each level of Bloom's Taxonomy.
Jasem Neaimi
AI Collaboration Researcher
In 1956, Bloom's Taxonomy defined six levels of cognitive thinking — from basic recall to creative synthesis. In 2001, Anderson and Krathwohl revised it into the version used today: Remember, Understand, Apply, Analyze, Evaluate, Create.
I built on this foundation to create a collaboration model for working with AI: who should lead at each cognitive level, and where the trust boundary sits. The result is a practical framework tested across legal, research, and coding domains.
The Split Cognitive Stack
The framework maps every task to six cognitive levels, split by a trust boundary:
The Trust Boundary
There's a critical line between Level 3 and Level 4:
- Below it (Levels 1–3): You can mostly trust AI output with light verification. AI excels here — infinite recall, reliable summarization, pattern application.
- Above it (Levels 4–6): You must verify and direct. AI assists but doesn't lead. The human brings context, values, and judgment.
This maps directly to the "production gap" in vibe coding — people trust AI at Levels 1–3 (build the prototype), then hit a wall at Level 4+ (is this secure? does this scale? should we even build this?).
How to Use It: The 6 Universal Questions
Before AI does any work, answer these evaluative questions. They're universal — they work for any project, domain, or decision:
| # | Question | Purpose |
|---|---|---|
| 1 | What is the purpose? | Why does this matter? |
| 2 | What does each side bring? | How does it fit together? |
| 3 | What does success look like? | What's the outcome I care about? |
| 4 | What am I afraid of? | What's the risk priority? |
| 5 | What's the scope? | What are the boundaries? |
| 6 | What's the commitment level? | How deep am I going? |
These questions force you to evaluate before anything is built, researched, or drafted. AI cannot answer them — they require your personal context, values, and goals.
The Iterative Spiral
The framework isn't linear. It works as a collaborative spiral:
- Level 5 (Evaluate) — You answer the 6 questions
- Level 4 (Analyze) — AI researches, you steer
- Level 5 again — AI's analysis surfaces new decisions; you evaluate
- Levels 3→2→1 — AI executes, explains, and saves
The back-and-forth between Level 4 and 5 may repeat multiple times. This iterative loop is where the real value emerges.
Worked Example: UAE Partnership MOU
I tested this framework on a real scenario — drafting a binding MOU for a training partnership in the UAE, with zero legal expertise.
| Step | Level | Who Led | What Happened |
|---|---|---|---|
| 1 | Evaluate | Human | Answered the 6 questions: purpose, contributions, success criteria, fears, scope, commitment |
| 2 | Analyze | AI ↔ Human | AI researched UAE MOU law. Key finding: MOUs are binding by default in UAE (opposite of Western jurisdictions) |
| 3 | Evaluate | Human | Made strategic decisions: subcontract model, penalty structure, exclusivity terms |
| 4 | Analyze | AI | Further research on penalty enforceability, arbitration best practices |
| 5 | Evaluate | Human | Final decisions on revenue split, delivery model |
| 6 | Apply | AI | Drafted full 14-clause MOU with UAE-compliant structure |
| 7 | Understand | AI | Explained every clause in plain language |
| 8 | Remember | AI | Saved everything to Second Brain |
The framework naturally created multiple Level 4↔5 loops. I walked away not just with a document, but with a deeper understanding of partnership law — proving the framework is also a learning accelerator.
Why This Matters Now
The market is drowning in "AI tool" content but starving for "AI thinking" content. Based on analysis of 419 posts across 6 platforms, every competitor occupies 1–2 Bloom's levels. This framework covers all six.
The framework sits in the only quadrant that is both deep (grounded in cognitive science) and actionable (comes with questions, levels, and a worked example).
Connection to Top-Down Learning
Dr. Justin Sung's research shows that starting at Level 5 (Evaluate) and working downwards leads to deeper learning than the traditional bottom-up approach. With AI, this becomes even more powerful:
- AI eliminates the drudgery of Levels 1–3
- Humans spend all cognitive energy on Evaluate and Create
- The collaboration amplifies both strengths
AI should make us sharper thinkers, not lazier ones.
Get new insights
Subscribe for the latest research and frameworks, delivered to your inbox.