Divergent Ideation

Divergent Ideation

Generate a large quantity of ideas with judgment explicitly deferred, on the principle that quantity produces quality's raw material.

Why

Evaluating each idea as it's generated kills fluency — the mind self-censors toward the safe and obvious. Deferring judgment to a separate pass produces both more ideas and, reliably, better ones deeper in the list.

Use when / Don't use when

• Use when: starting any generative task before a single approach has been chosen; a group is anchoring on the first idea proposed.
• Don't use when: evaluation criteria are the actual bottleneck, not idea supply — use idea-convergence or weighted-scoring instead.

Inputs → Outputs

• Inputs: a prompt or problem open to multiple solutions.
• Outputs: a large raw list of ideas, unevaluated, ready for a separate convergence pass.

Principles

• Defer all judgment during generation — no "that won't work" until the separate convergence phase.
• Quantity correlates with quality — the 50th idea is, on average, more novel than the 5th.
• Build on others' ideas rather than only originating your own.
• Write every idea down, including the obvious and the silly — they often trigger better ones downstream.

Procedure

1. State the prompt precisely; set a target quantity or time-box, not a quality bar.
2. Generate rapidly; capture everything without commentary or filtering.
3. When generation slows, apply a push technique — a constraint, a random stimulus, a reversal — to break the plateau.
4. Continue past the first plateau; the second wave after a push is often where the novel ideas live.
5. Stop at the time-box or quantity target; hand the full raw list to a separate convergence step without evaluating yet.

Common mistakes

• Silently evaluating while "generating" — self-censorship disguised as brainstorming.
• Stopping at the first plateau instead of pushing through it.
• Mixing generation and evaluation in a group setting, which suppresses quieter contributors' ideas.

Examples

• Naming a product.
• Generating candidate root causes before a diagnosis.
• Listing possible marketing channels, or alternative interpretations of an ambiguous result.

Related

• [[idea-convergence]] — the required next phase.
• [[scamper]] — a systematic push technique for plateaus.
• [[morphological-analysis]] — the combinatorial alternative to unstructured generation.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: skyf0xx
• Source: skyf0xx/better-thinking
• License: MIT
• Homepage: https://zenbin.org/p/better-thinking

Install and usage instructions live in the source repository linked above.