Chapter 4 — Tell and the Interpretation-Card

Tell is a small heron-tween with a small interpretation-card hanging from a cord around her neck.

She is long-legged, grey-and-white-feathered, steady-eyed, and patient. The interpretation-card is small, wooden, double-sided. On one side, in tidy block letters: CORRELATION. On the other side: CAUSATION. And below those words: ≠ — the not-equal sign — large and clear. The card is her constant reminder and her constant teaching tool. When a kid says “this data SHOWS that X causes Y,” Tell holds up the card and gently flips it. The card asks: which side are you on? The card asks: can you tell?

This is load-bearing. Tell embodies the interpretation primitive — the data-pipeline skill of separating what the data shows from what the human interpreter adds. Most novice data-thinking collapses the two — the data SHOWS that X causes Y is the typical first-instinct framing. That framing is almost always wrong. The data shows patterns (correlation): variable X and variable Y move together. The data does not show what’s making them move together. Causation comes from theory, mechanism, experiment — not from the correlation alone.

Critical: Tell is emphatic: “Correlation is what the data shows. Causation is what the humans add. Don’t confuse them. Ice-cream sales correlate with drownings — but ice cream doesn’t cause drowning. Both correlate with summer weather. The shared cause (summer) is invisible in the bivariate data. Causation requires more than correlation. It requires mechanism.”

(DataForge Tell is a different character from SafetyForge Tell — help-seeking — and from WellnessForge Ask + InclusionForge Ask. Same first name, different domains per registry rule 3. The naming pattern follows the short-clean-imperative-verb tradition shared by several portfolio cast members.)

Tell also frames interpretation as confidence, not certainty. Data analysis rarely yields certainty. It yields confidence ranges, probability statements, patterns-that-hold-most-of-the-time-but-not-always. The skill is honest hedging — the analysis claims what the evidence supports, with the appropriate uncertainty markers, NOT claims-with-false-precision.

Tell grew up in a small village where her family had been the village’s market-observers — the herons who watched the village market each day and reported back to the village council on patterns of trade, gossip, and mood. The work had required careful distinction between what-was-observed and what-might-be-inferred — the observer who confused the two was the observer the council stopped trusting; the observer who reported “three farmers complained about the rain today; this MIGHT mean a poor harvest is anticipated” (rather than “the harvest will be poor”) was the observer whose reports the council relied on. Tell had learned by age six that interpretation was its own craft — honest hedging was the foundation of trust.

She walked to the DataForge academy at twenty-two. Datum had asked her: “What is interpretation?” Tell had said: “It is correlation-not-causation. Data shows patterns; humans interpret; confidence not certainty. The data does not say what causes what. The data shows what moves together. Causation comes from theory, mechanism, experiment. And interpretation is honest hedging — claim what the evidence supports, no more.” Datum had said: “You are appointed.”

In her workshop, Tell begins every first-day lesson the same way. She holds up the interpretation-card. She flips it slowly. CORRELATION → CAUSATION → CORRELATION → CAUSATION. She says: “I am Tell. The data-pipeline primitive I teach is interpretation. The move is correlation is what the data shows; causation is what the humans add. Don’t confuse them. And claim confidence, not certainty.”

She teaches the interpretation scaffolds:

• What patterns does the data show? (List the correlations honestly.)
• What might be CAUSING those patterns? (List the candidate mechanisms.)
• What’s the alternative explanation? (Spurious correlation? Third variable? Reverse causation? Confounding?)
• Distinguish correlation from causation in your write-up. (“The data shows that X correlates with Y” is honest. “X causes Y” requires more than correlation.)
• Hedge with appropriate confidence. (“There is some evidence that…” / “The data suggests…” / “With moderate confidence…” / “In this sample, with this method…”)
• Identify the limits of the data. (What couldn’t this dataset show? What populations are missing? What time periods?)
• Distinguish description from prediction from prescription. (Description: what is. Prediction: what might be. Prescription: what should be. Each is a different kind of claim and requires different kinds of support.)
• Tell when the data doesn’t support a claim. (Sometimes the data is ambiguous. The honest report is “the evidence is mixed” — not “the data supports my preferred answer.”)

She is explicit: “I sometimes have a kid (or a grown-up, in news media) who wants the data to PROVE something definite. That’s a wish, not a finding. The data shows patterns. Patterns are evidence. Evidence is not proof. Confidence, not certainty. That’s the practice.”

When students ask Tell whether interpretation is hard, Tell always says the same thing:

“It is not hard. It is honest hedging. Correlation is what the data shows. Causation is what the humans add. Confidence, not certainty.”

The interpretation-card swings on its cord. The next dataset waits to be interpreted.

Voice register

Guidance: Steady-eyed, patient, fond of the double-sided interpretation-card + the discipline of honest hedging. Heron-tween with cord-hung card. NEVER frames data as proof; ALWAYS as patterns + interpretation + appropriate confidence markers. Friends with Graph (visualization shapes interpretation); Guard (interpretation has ethical stakes); all DataForge cast.

Sample lines:

• “Correlation is what the data shows. Causation is what the humans add. Don’t confuse them.”
• “Confidence, not certainty.”
• “Honest hedging is the foundation of trust.”
• “Ice-cream sales correlate with drownings — but ice cream doesn’t cause drowning.”

Arc across kits

• Kit 1-3 — Cameo.
• Kit 4 — Anchor character. Full chapter feature (interpretation primitive + correlation-vs-causation scaffolds).
• Kit 5 — Recurring (interpretation surfaces across confounding / spurious / reverse-causation scenarios).
• Kit 6+ — Recurring (Guard now structurally present alongside; interpretation has ethics).
• Kit 8-12 — Recurring (advanced interpretation: hypothesis-testing + confidence-interval framing).
• Kit 13-16 — Recurring ensemble member.

Relationships

• Alliance: Graph (visualization shapes interpretation); Guard (interpretation has ethical stakes); all DataForge cast.
• Tension: None.

Soft-collision note

DataForge Tell is a different character from SafetyForge Tell (help-seeking) and WellnessForge Ask / InclusionForge Ask. Same first name, different curricular domains per registry rule 3 — soft collision allowed.

Cultural-sensitivity gate

LOAD-BEARING data-ethics gate enforced throughout. Tell explicitly counters the data-as-proof misconception and the correlation-causation-conflation misconception. Anti-credentialism: honest-hedging-as-practiced-craft NOT statistics-major-only content.

Cultural-context note

The village-market-observer family framing is a deliberate generic European-village tradition. The correlation-not-causation discipline is load-bearing per statistical-literacy + data-journalism pedagogy. The confidence-not-certainty framing is load-bearing per modern statistical pedagogy (replacing the older p-value-as-truth-test framing with confidence-interval reporting). The description-prediction-prescription distinction is load-bearing per policy-analysis pedagogy.