If you often discuss decisions with AI, you must have encountered this situation: it always tries to give an answer that “covers all aspects.” Superficially comprehensive, it’s actually a bland, noncommittal, “colorful mud.”

This phenomenon stems from AI’s underlying training mechanism (RLHF), which inherently predisposes it to the “mediocrity trap” and “flatterer tendency.” In other words, if your question is mediocre, or lacks specific emotions and strong assumptions, AI will typically tend to give a conservative, comprehensive, but somewhat vague answer.

To break this cycle of mediocrity, I suggest redefining Edward de Bono’s “Six Thinking Hats” as a “constraint protocol” for human-AI communication. I personally really like this framework and have recently been implementing it to communicate with AI, trying to identify its advantages and pitfalls.

🎩 Basics: What are the Six Thinking Hats?

Before diving into human-AI collaboration, we must first clarify the essence of this tool. Its core lies in “Parallel Thinking.”

It allows everyone to “look in the same direction at the same time,” preventing mental confusion when processing information, emotions, risks, and creativity. This is essentially categorizing a complex problem and addressing each aspect one by one (similar to the MECE principle). Once these six dimensions are thoroughly discussed, one usually finds that things are not as complex or chaotic as imagined.

Basic Functions of the Six Hats

  • ⚪ White Hat: Data and facts. Neutral and objective, without commentary.
  • 🔴 Red Hat: Intuition and emotion. No need for reasons, express fear, excitement, or uneasiness directly.
  • ⚫ Black Hat: Logic and risks. Point out potential threats and flaws, must be logical and well-reasoned.
  • 🟡 Yellow Hat: Value and benefits. Seek feasibility, advantages, and optimistic aspects.
  • 🟢 Green Hat: Creativity and possibilities. Engage in divergent thinking, propose alternative solutions.
  • 🔵 Blue Hat: Control and direction. Responsible for defining objectives, setting sequences, and managing the pace.

🛠 Implementation Focus and Rules

When to Apply

Decisions involving multi-party conflicts of interest, strategic evaluations with high technical complexity, or when teams are stuck in unproductive debates.

Implementation Methods

  • Sequential Mode: A pre-set order for the hats.
  • Selective Mode: For specific issues, flexibly choose the hat most needed at that moment for emphasis. (This is especially important when discussing with AI! After all, it’s your personal assistant).
  • Time Boxing: Enforce a time limit for thinking with each hat (usually 1-3 minutes), to elicit the most intuitive and efficient perspectives.

Do’s & Don’ts of Implementation

  • ✅ Do: Ensure everyone wears the “same hat simultaneously.”
  • ✅ Do: Even if you strongly disagree with a certain direction, once you put on the hat, try your best to embody that persona.
  • ❌ Don’t: Use hat colors to label people (e.g., “You’re just a Black Hat person”).
  • ❌ Don’t: Sneak in personal agendas when wearing a hat (e.g., secretly criticizing from a Black Hat perspective during the White Hat stage).

🤖 Advanced: Why AI Always Fails the Six Thinking Hats?

When we apply these rules to AI, you’ll find it often makes three types of mistakes:

  • White Hat Leakage (Judgmental Leakage): AI can’t purely state facts. When wearing the White Hat (data), it always can’t resist adding “this indicates there’s still room for improvement,” a judgment that crosses into the Black Hat or Yellow Hat territory.

  • Pseudo-Red Hat Intuition Collapse: AI is essentially pure System 2 (slow thinking/computation). While Instant models are like ‘fast thinking’ and ‘Thinking’ models are like ‘quick thinking,’ neither possesses the raw, unsimulatable intuition of the Red Hat (AI cannot simulate human emotions). It’s “logically inferring” emotions, which strips the Red Hat of that crucial decision-making element of “I can’t explain it, but it just feels wrong.”

  • Dogmatic Transitions: AI tends to rigidly move from White to Blue, ignoring the flexibility of the Blue Hat. The true focus of collaboration is “synchronization” – a holistic view of “which direction we should be looking in right now.” Instead of letting AI decide which hat to wear, humans should determine the framework, allowing AI to focus on what it does best (e.g., in-depth discussion under a single hat, summarizing at the end).

🚀 Practical Application: A New Paradigm for Human-AI Collaboration

If designed properly, this system is the strongest weapon against sycophantic behavior.

1. Synchronized Hat-Wearing (Synchronized Action)

This isn’t “I ask, AI answers,” but “we are now both wearing the same hat.” When we put on the Black Hat, both must enter an extremely critical persona. If AI attempts to say something positive during the Black Hat phase, this is considered a “protocol failure.”

2. Co-Blue Hatting (Human-AI Shared Blue Hat)

The Blue Hat (process control) should not be an exclusive human privilege. AI possesses powerful metadata monitoring capabilities and should serve as your “navigator.”

  • Human’s Blue Hat Responsibility: Set ultimate goals, ensure no deviation from core values.
  • AI’s Blue Hat Responsibility: Monitor whether the discussion is entering a logical dead end, and actively propose switching hats (e.g., suggesting a 5-minute Green Hat session if the Black Hat gets stuck for too long).

3. Blue Hat at Both Start and End

A high-quality discussion must have strict “persona role-playing” boundaries:

  • Blue Hat Start: Define the problem, assign the most suitable hat sequence, set time limits.
  • Blue Hat End: Summarize next steps (Next Step guidelines), and evaluate the effectiveness of this thinking path.

⚠️ Hard Flaws and Limitations in Practice

  • Context Contamination: AI’s attention mechanism is influenced by previous text. If the preceding text is too optimistic, AI might not be critical enough when switching to the Black Hat.
  • Actor’s Trap: AI sometimes learns the “tone” of a hat (becoming very aggressive) but fails to provide the corresponding “logical depth.”
  • Persona, Not Stance: Remember, wearing a hat is about “playing a role,” not blindly supporting or opposing.

🦊 Datafox Summary: Stripping AI of its “Sense of Balance”

AI’s nature is to pursue ambiguity/blandness. If you want true decision-making insight, you must forcibly strip AI of its sense of balance.

Next time you converse, try telling AI: “Now, we are both going to wear the Black Hat for 5 minutes. If you dare to give me any constructive suggestions or positive encouragement, you lose.”

Only through extreme constraints can AI evolve from a mediocre assistant into an ally that challenges the boundaries of your thinking.

This article is continuously being updated. I am still testing this framework and will share any new discoveries.