Divergent Dynamics: Why Your Best Ideas Live at the Edge of Semantic Distance

Show notes

Most people approach creativity by circling familiar territory, hoping to spot something they missed. However, true innovation occurs at the periphery, where concepts are stretched to their breaking point through Conceptual Expansion. This is the cognitive ability to widen the boundaries of existing mental structures to generate something truly novel.

The Science of the “Stretch”

To understand how to be more creative, we have to look at two core research concepts:

1. The “Sweet Spot” of Semantic Distance: Semantic distance is a computational measure of how far an idea has traveled from its starting point. While we want “distant” ideas, research shows a curvilinear relationship: creativity ratings peak at a specific “sweet spot” (a distance of roughly 0.9 to 1.0). Move too far, and the idea becomes incoherent; stay too close, and it’s a cliché.

2. Structured Imagination: Our brains are naturally “anchored” to what we know. When asked to imagine something new, we subconsciously default to existing templates (e.g., imagining aliens that still have legs and eyes).

Leveraging AI as a “Semantic Catapult”

Generative AI is a powerful partner because it doesn’t possess the “common sense” that fuels structured imagination. It acts as an external scaffold, helping you define problems and generate “pre-inventive” structures that you can later refine.

Here is how to practically use AI to catapult your ideas into that creative “sweet spot”:

1. The “Property Violation” Prompt

Force the AI to strip away the defining characteristics of a concept. This breaks functional fixedness, the habit of seeing an object only for its traditional use.

• The Prompt: “List the 5 most ‘essential’ properties of a [Concept, e.g., Smartphone]. Now, design a way to achieve the same goal while explicitly forbidding those 5 properties. Focus on high semantic distance.”

2. Recursive “Spiral” Prompting

Use the AI to move degrees away from your starting point. If you just ask for an idea, you’ll get the “average” of its training data. Instead, spiral outward.

• The Prompt: “Give me 5 metaphors for [Project]. Take the most ‘absurd’ metaphor from that list and explain how it could actually work as a business model.”

3. The “Negative Constraint” expansion

Research shows that highly creative individuals have “looser” associative networks. You can mimic this by forcing the AI to avoid the most likely paths.

• The Prompt: “I need a solution for [Problem]. Provide 10 ideas, but do NOT use [Common Solution A], [Common Solution B], or [Common Solution C]. Push for ‘overinclusive’ associations that link this to [Unrelated Field, e.g., Marine Biology].”

4. The Bisociation Mashup

Force two unrelated domains to share the same sentence.

Prompt: “Apply Bisociation to my project. Blend the structure of [Topic A: e.g., urban gardening] with the business logic of [Topic B: e.g., subscription software]. What hybrid concept emerges?”

The further apart your A and B, the higher your odds of landing somewhere no one has been before.

5. The “Wrong Expert” Reframe

Ask someone from an unrelated field to solve your problem.

Prompt: “I’m trying to improve [employee onboarding / a school lunch menu / a retail checkout experience]. How would a marine biologist approach this problem? Now how would a jazz musician approach it? What do both of those answers have in common?”

6. The Constraint Cascade

Layer increasingly weird restrictions to force creative detours.

Prompt: “Give me a solution to [problem]. Now solve it again, but the solution can’t use any technology. Now solve it again, but it has to happen in under 60 seconds. Now solve it one more time, but it has to involve a complete stranger.”

Each constraint forces a different part of your concept map to activate, moving you further from your default answer each time.

The Curator’s New Frontier

A 2026 study found that human creativity still surpasses generative AI in image production, and notably, as human guidance in the process decreased, so did the quality of the output. AI without a human steering it drifts toward the middle. Human judgment is what selects the edge.

This means the creative role isn’t disappearing, it’s changing. The job is shifting from generator to curator: someone who can recognize a genuinely novel idea when it surfaces, evaluate it against the Sweet Spot, and guide the process toward something that’s both distant and coherent.

Your best ideas live at the edge of semantic distance. Let your associations run loose. Use AI to get you further than you’d go on your own. And use your human judgment, that salience filter, that gut sense of when something actually lands, to bring back the ideas worth keeping.

The frontier isn’t a place you find. It’s a place you generate, one expansive iteration at a time.