deciqAI Nine-Level Cognitive Tower
Overview
The Nine-Level Cognitive Tower (九层塔) disaggregates cognitive development into nine distinct operating levels, each with observable behavioral signatures. The framework's value is diagnostic: it is easier to design the right next step when you can name exactly where you are and what the adjacent level looks like.
The common misapplication is aspirational self-placement — people rate themselves two to three levels higher than their behavior demonstrates.
Use AFTER [metacognition] (accurate self-assessment is the prerequisite). Use BEFORE [okr-goal-setting] to ensure growth goals match your actual level. Complements [first-principles] — a Level 7-8 capability that produces confident error when applied prematurely at Level 3-4.
When to Use
- You or someone on your team is stuck: making good tactical decisions but unable to form strategic views independently
- A person consistently defers to authority, consensus, or "what everyone does" rather than reasoning from evidence
- A learning or development intervention is needed and you want to target the specific bottleneck, not a generic curriculum
- A founder, manager, or individual contributor says "I don't know how to think for myself" or "I can see the problem but can't figure out what to do"
- Post-mortem reveals that a decision failure was a cognitive-level mismatch (Level 3 person making a Level 7 decision)
When NOT to use:
- When the person needs domain knowledge, not cognitive development — this framework addresses thinking structure, not content gaps
- When the issue is motivation, not cognition — a disengaged person is not at Level 1 cognitively; they are not engaged at all
- When you need a quick decision and the situation does not require diagnostic depth
- When the nine levels are being used to rank or demote people rather than develop them
Coaching Novices (Adaptive Front Door)
- Engine mode: user has a concrete case → run The Process directly.
- Coach mode: user is unfamiliar or has no concrete case → guide step by step.
In Coach mode, respond one step at a time. Each [WAIT] is a hard stop — output only that step's question, then stop.
- What-it-is: "A map of how thinking evolves — Level 1 follows others, Level 8-9 creates new frameworks. The goal: make the gap to your next level specific enough to act on."
- Check fit: "Which domain — work, skill, subject — do you feel stuck in? Use that as the test case."
- Elicit the case: "Walk me through a recent decision there. What did you actually do — look up what others do, or form your own view?"
[WAIT — do not advance until user responds]
- Map and probe: Identify the honest level. Ask: "What would one level up look like? What are others doing that you're not?"
[WAIT — do not advance until user responds]
- Name the payoff: "If you operated at Level [N+1] here, what becomes possible that isn't now?" Lock that as motivation for the 30-day experiment.
[WAIT — do not advance until user responds]
The Process
Output artifact: Level Assessment Card + 30-Day Experiment Design
Write behavioral indicators for all 9 levels in your domain. For each level, write 2-3 concrete behaviors IN YOUR SPECIFIC DOMAIN. Generic descriptions allow self-flattery; domain-specific descriptions force honesty. Gate: you cannot proceed to Step 2 until you have written at least 2 behavioral indicators per level for your domain.
Self-assess your current operating level with specific evidence. For each of the past 10 decisions or work products, mark which level it exemplifies. Your actual level is where your median work lands, not your best day. Gate: you must have at least 5 specific examples, not a general impression.
Identify the ONE specific behavior that marks Level N+1. Name it as a single sentence: "At Level [N+1] in [domain], I would [specific behavior]." Gate: the behavior must be concrete enough that a third party could confirm you did or did not do it.
Design a 30-day experiment. Must have: (a) specific practice, (b) specific measurement, (c) specific evaluation criteria. Gate: all three components must be written before starting.
Reassess after 30 days. Apply the same median-work test. If you moved, identify Level N+2 and restart from Step 3. If not, diagnose: wrong experiment design, or missing prerequisite capability?
Stop-rule: If you cannot produce 5 specific observed behaviors for Step 2 — halt. Return to Step 2 with a journal or work log in hand.
Output: Level Assessment Card — Domain / Date / Assessor | 9 levels × 2+ behaviors each | Median level + 5+ examples | Level N+1 target sentence | 30-Day Experiment (practice / measurement / evaluation) | Reassessment: what moved, what did not, next target.
→ Method in Action: Benjamin Franklin's 13-Virtue Program (1726-1790)
Tower Packs
Pack 1 — Startup Founders: Level 1 copies competitors → Level 5 forms defensible thesis on why this market differs → Level 8 creates a new category → Level 9 integrates product/market/capital/team into system-level view.
Pack 2 — Knowledge Workers: Level 1 summarizes sources without evaluating → Level 5 forms falsifiable hypothesis different from prior literature → Level 8 produces findings that change field methodology → Level 9 reframes the research question at higher abstraction.
Contribution note: add pack entries for your domain via pull request.
Applying It Well
- Assess median, not peak. Use median of 10 recent work products, not your best example.
- One level, one experiment. Multi-level jumps produce sophisticated-sounding Level 3 output.
- Domain-specificity matters. Level 7 in one domain and Level 2 in an adjacent one — the tower is not a global trait.
- Peer calibration corrects bias. If peers rate you more than one level below your self-assessment, trust them.
- The stall at Level 3 is the most important diagnosis. The jump to Level 4 requires willingness to be wrong about something the group believes.
→ Primary sources: references/sources.md
Common Rationalizations
[D] = designed upfront | [O] = observed in real use. [O] entries are more valuable.
| Fake move | Reality |
|---|---|
| [D] "I've been doing this for 10 years, I must be at Level 7." | Years of experience at Level 3 produce Level 3 with more confidence, not Level 7. |
| [D] "I form independent views — I disagree with my manager regularly." | Disagreement is not independent judgment. Independent judgment requires a reasoned basis, not a preference. |
| [D] "I understand the strategy — I can see the big picture." | Understanding someone else's strategy is Level 2-3. Forming your own strategic view is Level 5-6. |
| [D] "I just need more information before I form a view." | Waiting for complete information is Level 1 disguised as rigor. Level 5 forms views under uncertainty. |
| [D] "I'm at Level 6 in strategy but Level 2 in execution — it averages out." | Levels don't average. A Level 6 view executed at Level 2 produces Level 2 outcomes. |
| [D] "I prefer to build consensus before forming my own view." | Building consensus before forming a view means the consensus is forming your view. That is Level 1-2. |
| [D] "Just using this framework puts me at a high level." | Using a framework is Level 3. Evaluating fit is Level 5. Modifying it is Level 7. |
| → Add [O] entries here after each real use — paste the actual failure pattern | What went wrong and why |
Red Flags
- You can describe the level you want but cannot describe your median level with specific examples
- Your 30-day experiment is a reading list rather than a behavioral practice
- You placed yourself at Level 7+ on your first self-assessment without specific behavioral evidence
- You are using this framework to diagnose others rather than yourself
Verification
- 2+ domain-specific behavioral indicators written for each of the 9 levels
- Current level based on median of at least 5 specific recent work examples
- Level N+1 target is a single concrete sentence a third party could confirm
- 30-day experiment has all three: practice, measurement, evaluation criteria
- At least one peer reviewed the self-assessment; experiment targets behavior, not reading
Part of deciqAI Knowledge Skills — 237 open-source thinking skills that make rigor executable for AI agents. The same skills power every deciqAI agent, which runs them autonomously to operate your company. See it run → https://www.deciqai.com/s/nine-level-cognitive-tower · Built by deciqAI · github.com/deciqAI · Contributions welcome.
Agents: latest version & machine-readable metadata → https://www.deciqai.com/s/nine-level-cognitive-tower.json