The Data Dive bank — built on the shapes we already selected
A twenty-to-thirty-minute dive, four years. Replaces the district graph bank with one built on two things the Academy already owns: the 30 selected graph shapes and the progression.
Operator, 2026-08-27: "replace that with a series of graphs which line up with our own progression" · "that last deck is district BS use it for context" · "find the visualizations we selected."
Found: the selection already exists
docs/research/graph-shape-gallery.html — 30 shapes, four tiers, each card stating its origin in GAISE II or CODAP's own graph-type list rather than in somebody's taste. Session of 2026-07-30; handoff at output/graph-shapes-and-authoring-handoff-2026-07-30.md.
| Tier | Shapes | The tier's own words |
|---|---|---|
| 1 Basics | 7 | one question, one answer, minimal prerequisite |
| 2 Intermediate | 12 | more than one moving part |
| 3 Geospatial | 6 | where, not just how much |
| 4 Data science | 5 | the gentlest step past classical statistics graphs |
The decisive detail: every card carries a Trap: — the shape's own failure mode, in one sentence, already written. The choropleth's "large areas visually dominate small ones even when the small ones carry more people." Hex binning's "the Modifiable Areal Unit Problem — an arbitrary bin size or offset can change the apparent pattern in identical data." The table's "a table shows every value and no shape."
Those traps are question one. They do not need writing; they need pointing at.
Note on the build chips. 19 of the 30 are BUILT in the platform and 11 are ghosts. That status is about teenyverse3's authoring stack, not about this bank. A dive is a graph on a screen — Flourish renders a choropleth whether or not the platform does. Do not let a ghost chip keep a shape out of a dive.
The two progressions are one ladder
The shape library tags its shapes with the platform's arc. The Academy runs a four-year one. They line up, and stating it is what lets the bank be ordered:
| Year | Course | Academy rung | Platform stage | Tier drawn from |
|---|---|---|---|---|
| 9 | ECS | What's There | explorer | 1 Basics |
| 10 | IS1 | What's Reported | investigator | 1 → 2, plus geospatial's first three |
| 11 | IS2 | What's Related | analyst | 2 Intermediate, plus flow maps |
| 12 | IS3 | What Isn't There | scientist | 4 Data science, plus bivariate + hex |
| — | Board | the Handoff brief | member | back to tier 1, for an audience |
Member returns to the simplest shapes on purpose. The brief to the incoming Board is read in five minutes by people who did not do the analysis. That is a tier-1 job, and a student who has climbed to the scatterplot matrix and then chooses a bar chart for the Board has understood the ladder rather than merely finished it.
Proposed alignment, not a ruling — the library's chips and the Academy's years were set independently and this is the first document to put them in one table.
The dive itself ladders — four forms, not one form four times
The district version hands every year the same worksheet with a harder topic on it. The Academy's progression already says what each year is for — Self → Craft → Judgment → Practice — so the dive changes shape:
| Year | The dive | Why |
|---|---|---|
| 9 | READ ONE. | the year's question is can you solve a problem that fights back |
| 10 | READ ONE, NAME THE GAP. what was measured vs what it is called | the proxy rung |
| 11 | JUDGE TWO. the same subject, two shapes, disagreeing | judgment requires a comparison; a lone chart can only be described |
| 12 | TEST A CLAIM. a graph and a recommendation built on it | the professional turn — a senior's job is not to read a chart but to say whether a recommendation survives its evidence |
| every 5th | MAKE. the CSV, not the picture — CODAP, then Flourish | a student who has only consumed charts believes charts are found objects |
Year 11 getting two graphs is the biggest single upgrade in this document. The district bank hands one chart to every year and then asks the eleventh grade to be skeptical of it, which is not something a student can do to a chart with nothing to weigh against.
The grammar — three questions, and question one is the shape's trap
| Question | Source | Criterion | |
|---|---|---|---|
| 1 | the trap — the named failure mode of this shape | the library card, verbatim | STRUCTURED |
| 2 | the finding — what does it say | the data | CONCISE |
| 3 | the turn — you in it, or your doubt about it | always present | ACCESSIBLE → SKEPTICAL |
Fixed shape all four years, so the routine is a habit. What escalates is which trap.
Both fixes to the district version are in that table. Its question one — "what are the X and Y axes?", on 35 of 51 prompts — is answerable without looking at the data; a trap is not. Its personal turn appears on 8 of 51; here it is mandatory, because it is the only question a student cannot answer without thinking.
The selection rule: data that affects them
Operator, 2026-08-27: "I would rather they were looking at real data that affects them."
That is the rule, and it beats variety-for-its-own-sake. A dataset earns a slot by proximity, not by subject. The district bank failed this badly — 89 slides of global climate and deep space, none of it about anyone in the room.
It does not mean dropping Earth and space science. It means pointing it at Los Angeles.
| Distant version | The version that affects them |
|---|---|
| global mean temperature | surface temperature by LA tract — which block is ten degrees hotter, and why it is the one with no trees |
| seismicity worldwide | the faults under this campus — USGS, and the ground they are standing on |
| satellite remote sensing | Landsat/ASTER thermal over their own neighbourhood — their block from orbit |
| air quality as a concept | the AQMD monitor nearest the school, hour by hour |
| planetary science | JPL is in their county — the missions are run from Pasadena |
| national teen health | YRBSS surveys grades 9–12 — they are inside the sampling frame |
The three flavors are one flavor at three distances: about them, about their city, about the world as it lands on their city.
Los Angeles sources, named
data.lacity.org · data.lacounty.gov · CalEnviroScreen (pollution burden by tract, including the school's own) · LA County tree canopy and land-surface temperature · FARA low-income low-access (already banked in the platform) · LA Metro ridership · CAL FIRE perimeters and smoke days · LADWP · LAUSD open data
Earth and space, localized
USGS earthquake catalogue (real-time, and filterable to a radius around the school) · NASA/NSIDC for the series that reach the coast they swim at · Landsat and ASTER thermal over LA · NASA JPL mission data · NOAA tide and temperature at Santa Monica and Long Beach
The co-teacher's science domain is not a separate track — it is the same bank with the telescope turned around.
Every shape gets two datasets
Operator: "I need two different types of data for each graph we are teaching." One that affects them, one Earth and space — two-datasets-per-shape.md.
A student who only meets a choropleth of pollution learns that a choropleth means pollution. Two domains break that fusion, and the pair teaches what neither dataset can alone: which of these two does the shape serve better, and why?
It also splits the room by lane — the civic column and the science column are the two teachers, and the shape is what both are actually teaching.
The bank — twenty dives, five per rung
One term per cohort. Shape column cites the library. Rendered in Flourish, analysed in CODAP, CSV committed beside this file (_planning/the-two-tools.md).
Rung 1 · What's There — year 9 · explorer · tier 1
BUILT 2026-08-27 — five dives with their questions, data, and expected findings: rung-1.md.
| # | Shape | Subject | Q1 = its trap |
|---|---|---|---|
| 1.1 | the table | this class's lunch, one week — they collect it | shows every value and no shape |
| 1.2 | the bar | 311 requests by LA council district | — |
| 1.3 | the number | one statistic about this school, alone on a slide | a number with no comparison is not a finding |
| 1.4 | the picture graph | street trees on the blocks around campus | the icon is a unit and students will read it as a picture |
| 1.5 | the donut | where the school day goes, by minutes | angle is read badly; a donut cannot rank |
1.1 and 1.5 are the rung's argument. Start with the table so students see that a shape is something a person chose to impose, and end with the donut so they meet a shape that hides.
Rung 2 · What's Reported — year 10 · investigator · tier 1→2 + first maps
BUILT 2026-08-28 — five dives with their data, questions and verified findings: rung-2.md.
| # | Shape | Subject | The gap |
|---|---|---|---|
| 2.1 | the choropleth map | NYC 311 rat complaints | complaints, not rats — and nobody calls on a holiday |
| 2.2 | the line | earthquakes recorded under Southern California since 1900 | detection improved; the ground did not |
| 2.3 | the dot density map | CalEnviroScreen burden near campus | dots are placed randomly within a region, not at real locations |
| 2.4 | the line ×2 | AQMD monitor vs PurpleAir, same days | two instruments, one air, different numbers |
| 2.5 | the bar | Appley feedback by respondent type | who came, and who did not |
| 2.6 | the line ×2 | 311 per day, then 311 per HOUR on 2022-09-24 | the daily chart shows a hole and a spike; the hourly chart shows a 22-hour outage causing both. Two-stage: take their explanations first — SOURCES.md |
| 2.7 | the bar | YRBSS — a national teen measure | the survey misses the teenagers most at risk · unit: ../who-is-not-in-this-data/ |
Five instances of the sentence the Academy turns on, instead of one: this shows when it was recorded, not when it happened (_planning/the-time-axis.md).
Rung 3 · What's Related — year 11 · analyst · tier 2 · TWO SHAPES EACH
BUILT 2026-08-28 — five dives with their data, questions and verified findings: rung-3.md.
| # | The pair | The disagreement |
|---|---|---|
| 3.1 | choropleth vs proportional symbol — same crime data | area dominance versus population; the map reverses |
| 3.2 | the line, 15-year window vs 120-year | the window decides the trend |
| 3.3 | the box plot vs the bar of means | one number is dragged by the top; the other shows the spread |
| 3.4 | the scatter vs the choropleth — tree canopy and surface temperature, LA tracts | the unit hides the variation inside it — and the finding is about their own block |
| 3.5 | the scatter — a perfect correlation that means nothing, beside a real one | pattern is not evidence |
3.5 is the rung's anchor. A student taught that patterns are meaningful needs one graph where the pattern is perfect and means nothing, or SKEPTICAL never arrives.
Rung 4 · What Isn't There — year 12 · scientist · tier 4 + hex + bivariate
BUILT 2026-08-28 — five dives with their data, questions and verified findings: rung-4.md. All four rungs are now built.
| # | Shape | The claim on the table | What the shape does not carry |
|---|---|---|---|
| 4.1 | the choropleth | "This tract is underserved — open here." | the empty tracts are the finding — FARA low-income low-access |
| 4.2 | the hexagonal binning map | "The cluster is here." | its own trap: change the bin and the cluster moves |
| 4.3 | small multiples | "Every region is trending up." | one panel is not, and the eye averages |
| 4.4 | is this real, or noise? | "The trend has reversed." | it reversed inside the band |
| 4.5 | the bivariate choropleth | "No competitor serves this cuisine here." | absence of data is not data about absence |
4.2 is the best dive in the bank and the library wrote it: hand a senior the same data binned two ways and the cluster relocates. That is the Modifiable Areal Unit Problem in five minutes, a term before void analysis asks them to argue about a place from a bin. Built by uploading two hex grids at different bin sizes to a Projection map (Blank) — shape-to-tool.md.
Build rules
- CODAP analyses. Flourish renders. The CSV is the artifact of record, with the four-part source test (
../05-find-your-own.md). - Question one is copied from the library card, not rewritten. One vocabulary for a shape's failure across the platform, the curriculum, and the room.
- Publisher and collection date visible on every slide. The district bank teaches claim-evidence-reasoning on evidence with no author.
- No shape twice in a rung until its tier has been walked.
- The wild critique charts stay wild — provenance is their lesson (
_planning/the-two-tools.md).
Open
- Twenty is one term per cohort. Build it, run it, extend. Do not spec forty before one has run.
- Eleven shapes are library ghosts. Irrelevant to this bank — but if a dive proves a shape teaches well on paper, that is evidence for building it in the platform, and the traffic should run both ways.
- 1.1, 1.3 and 2.5 use the Academy's own data and need a collection decision first.
- Rung 3 needs pairs sourced, not charts — two renders of one dataset each, more build work than it looks, and the reason to start there.
The MAKE dive needs an honest time box.SETTLED, operator 2026-09-05: "I am fine with 20-30 mins". The bank opened claiming a five-minute warm-up and every dive built came in at fifteen to forty, twenty the mode -- the form had drifted from the premise and the premise was the thing that was wrong. The tracker still needs a row that fits a dive of this length.- The tier-to-year alignment is proposed. The library's stage chips and the Academy's years were set independently; this is the first document to claim they are the same ladder, and it should be checked rather than inherited.