You Don’t Know What You Think You Know
You think you understand a concept because you can recite its definition. You’re wrong.
True understanding is not about labeling; it’s about recognizing the shape of an idea in the wild. What if the best way to grasp a slippery, counterintuitive concept is to hide it from yourself?
Philosopher Amanda Askell — who leads character design for Claude at Anthropic — developed a prompting technique that does exactly that. Instead of asking an AI for a dry explanation, she asks for a fable that embodies the concept without ever naming it.
The reader must assemble the mental structure blind, only to have the name revealed at the end, locking the pieces into place. This method flips conventional learning on its head. It forces you to map the terrain before you see the label.
The result is a deeper, stickier form of conceptual understanding — one that survives beyond the flashcards.
The Prompt That Hides the Answer
Askell’s template has a simple architecture. But its constraints are designed to engineer a specific cognitive journey.
First, you request a fable — an indirect narrative.
Second, you forbid the direct naming of the concept.
Third, you mandate that the realization dawns only near the story’s end.
Finally, a short explanation connects the dots.
This delay isn’t a gimmick. It’s a carefully calibrated cognitive friction that forces you to model the idea yourself before anyone hands you a label.
I want to understand [concept]. Please explain it by writing a fable — an indirect, narrative version of the concept. The story should embody the concept completely without naming it directly. Ideally, the reader should only start to realize what the concept actually is near the end of the story. After the fable, add a short explanation that names the concept clearly and connects it back to the key moments in the story.
Inside the Reader’s Mind While the Name Is Missing
Why does this work? The technique activates three mental processes simultaneously. First, active modeling: you track characters, infer motives, and trace cause-and-effect chains because the story demands it.
Second, the absence of the concept name creates productive difficulty. You’re forced to construct the logical skeleton from the inside out, feeling for its joints.
Third, the reveal at the end acts as a label that suddenly organizes the mental scaffolding you’ve already built. That moment is deeply satisfying and cements the concept. This combination is rare in traditional instruction, where terms are often defined before we’ve even seen the problem they solve.
Where the Fable Flourishes — and Where It Falters
Askell notes the technique is best suited for concepts that involve agents, actions, and consequences.
Reflexive equilibria, adverse selection, and game-theory dilemmas come alive as characters navigate moral and strategic pitfalls. But conceptual understanding in math of the purest kind — say, the Riemann hypothesis — resists this treatment.
Abstract structures with no human-like agency don’t map neatly onto narrative arcs. The technique also demands that you never introduce the concept name prematurely. If you slip and name it in the prompt, you collapse the essential delay and turn the fable into mere illustration.
Maps, Gardens, and the Price of Premature Definitions
The technique’s power becomes undeniable when you see it applied.
When asked to explain Richard Rorty’s concept of truth, the fable “The Mapmaker’s Daughter” pits three mapmakers against one another — a naive realist, a foundationalist, and a pragmatist. The narrative never utters the word “truth” until the very end, yet the reader gradually realizes that maps aren’t mirrors of reality but instruments for navigating it.
Similarly, applying the template to the Israeli-Palestinian conflict produced “The Garden Between Two Storms,” a story of two peoples — Wanderers and Keepers — both with ancient roots, trauma, and fears.
Without naming Israel or Palestine, the fable made the symmetry of grievances palpable before dropping the historical context in the explanation. In both cases, the technique circumvented reflexive ideological reactions by forcing empathy through narrative before labeling.
Why This Mirrors How Claude Was Designed
There’s a deep resonance between Askell’s fable prompt and her day job shaping Claude’s character.
Anthropic didn’t hard-code rigid rules into Claude; they shaped its underlying dispositions and values. Rigid rules break the moment they’re out of context, just as a definition memorized without understanding crumbles under pressure.
The fable prompt is the user-facing reflection of that philosophy: design a path of emergence, not a set of assertions. It’s a small act of alignment between human and machine cognition. You’re not just extracting output; you’re constructing understanding together, story by story.
Beyond Fables: Detective Stories and Self-Critique
The template is flexible. Swap “fable” for “detective story,” “corporate memo from a future civilization,” or “post-mortem report,” and you force the same concept through a fresh metaphorical lens.
Each genre activates different neural clusters, enriching the model you build. For even deeper understanding, Askell and others suggest following up with: “What critical aspect of [concept] did this fable fail to capture?”
This prompt surfaces the story’s blind spots — its simplifications and edge cases — turning a one-shot explanation into a dialogue about the limits of metaphor.
Critics who dismiss this as a “perfect one-shot” addiction miss the point; the technique isn’t a substitute for follow-up questions but a specific cognitive intervention that plain Q&A rarely replicates.
The Name Is the Enemy of Knowing
Most education hands you the name first and hopes you’ll later fill in the structure. The fable technique inverts that order — and in doing so, it exposes a fundamental truth about conceptual understanding.
You cannot truly own an idea until you’ve wrestled with its implications in the dark. This isn’t a prompting gimmick; it’s a philosophical stance on learning, one that Anthropic itself baked into Claude’s design.
The next time you reach for a definition, resist the reflex. Make the AI tell you a story that hides its point. Then feel the click when the name finally lands — because that click is understanding being built, not borrowed.
