Four CS Courses, Scored Against the National Academies Competencies

The report's framing is the part that made me stop reading and go check my own work. Everybody wants their subject added to K–12 right now: the CS people, the data science people, the AI people, the quantum people. The curriculum is already full, so all of that good intention piles up on schools until nobody can move. The committee calls it the messy garden problem. Their answer is to stop adding territory for each field and instead name the competencies that sit underneath all of them.

I read the list and assumed I was mostly fine. Eight years teaching CS, four courses rebuilt from scratch, an AI literacy scaffold two years in the making. So I checked. And I want to be specific about how, because that is the part that mattered: I did not check against unit titles or against my standards crosswalk. I checked against the lesson files. Every day plan, every worksheet, every activity. If a competency was covered, I wanted to find the exact place where a student is asked to do the thing.

The seven competencies and their components

Each competency has three named components (Competency 6 has two). Coverage has to be judged per component, not per competency name, which is where most self-assessment goes wrong.

1. Problem Posing and Problem-Solving Processes — define a problem or question · identify the steps needed to address it · reflect and iterate. The report stresses that problems do not arrive pre-defined, and that knowing when a task is done is itself complex.
2. Producing and Working with Data — produce data (measurement and provenance) · organize data (case/attribute structure and data moves) · exploration and visualization.
3. Abstraction, Algorithmic Thinking, and Automation — creating abstractions and applying them in programming · algorithmic thinking at small and large scales · recognizing and building automated solutions. Note the wording: expressed in a step-by-step systematic and deterministic way.
4. Probabilistic and Inferential Reasoning — identify sources and impacts of uncertainty and variability · recognize the roles of probabilistic, statistical, and deterministic reasoning and apply them · make inferences and predictions with appropriate degrees of certainty. The body text specifies inference using hypothesis tests and confidence intervals.
5. Models and Representations — construct and use models and representations · assess their quality · recognize the limitations of all of them.
6. Technology and Society — reasoning about process and product · reasoning about self and society.
7. Data and Computing Systems — recognizing elements of data and computing systems · selecting and using them · ownership, openness, and access.

Course-level coverage

Strong = taught explicitly with dedicated instruction and assessment, most components · Partial = real but confined to one or two units, or only one component · Incidental = the vocabulary appears, the competency does not · None = absent.

¹ ACS I Competency 5 was raised from Partial to Strong on 2026-08-08 by curriculum edit, not by re-reading. Component 3 was the only gap and is now taught in two places. See the close-out.

What the full report changed

1. Competency 4 is partly already taught

The practitioner summary folded the probabilistic-versus-deterministic distinction into Competency 6. The report does not. It is Competency 4, component 2, and Competency 3 is defined against it as the deterministic one. That relocation matters, because ACS I already teaches it directly: Unit 6 has students distinguish deterministic functions (same input, same output) from probabilistic ones, states that simulations and AI models that sample from probabilities are probabilistic, and teaches why probabilistic code is harder to test — a wrong answer from deterministic code is a bug, but a different answer from probabilistic code is not. CET carries an informal version: LLMs generate text by predicting words, not by referencing facts.

So Competency 4 is not the empty column it looked like. ACS I touches all three components. CET moves from None to Partial on component 2 alone.

2. Competency 5 was scored too harshly

The competency is models and representations, and component 1 is construct and use them. Storyboards, wireframes, and level diagrams legitimately satisfy component 1. DGD I assesses whether level geometry matches the approved storyboard, which is using a representation to govern construction. DGD I and WDD I move from Incidental to Partial.

The failure is narrower and cleaner than "design artifacts are not models": components 2 and 3 are absent almost everywhere. Nothing asks students to assess the quality of a representation or name its limitations. That is the precise gap — not the artifacts, the interrogation of them.

3. Competency 2's first component is the real thin spot

The components are produce, organize, explore. ACS I is strong on provenance, on data moves, and on visualization. The weaker half is produce data: measurement — students mostly receive datasets rather than design what and how to measure. Worth noting that the CET robotics block is generating measurement data already and is not being taught as data production. That is the cheapest available upgrade in the portfolio.

Evidence behind each rating

Competency 1 — Problem posing and problem-solving

ACS I Strong. Five Practice Performance Task units, the AP CSP Create Performance Task, and the capstone. Student-chosen problems cover the define component, which most CS curricula skip. Debug, trace, and iterate language is dense through the programming units. DGD I Strong. Playtest-revise cycles run through the back half; the final unit is explicitly finish, present, reflect. CET and WDD I Partial. Build-test-revise exists, but the problems are teacher-specified, so component 1 is largely absent.

Competency 2 — Producing and working with data

ACS I Strong. Two units teach provenance directly: students interrogate data that is wrong because of who collected it, who is in it, and what labels were used. They reason about who is likely to be missing from a dataset — students who moved mid-year, people who mistrust data collection — and recognize that dropping those records is not fixing the dataset. They clean messy real data and build visualizations. CET Partial through its data block. DGD I None — zero data signal across all eight units, the only clean zero in the matrix. WDD I Incidental.

Competency 3 — Abstraction, algorithmic thinking, automation

ACS I Strong across seven units. CET Strong via robotics programming. DGD I Partial — real in the engine and scripting units, absent in the front half. WDD I Incidental — markup is not algorithmic work, and the redesign added no scripting strand. The weakest component portfolio-wide is the third: recognizing when automation is warranted as problems scale. Loops get taught as syntax more than as a scaling decision.

Competency 4 — Probabilistic and inferential reasoning

ACS I Partial across all three components. The data unit handles variability well: correlation versus causation as the stated core conceptual challenge, confounding variables, alternative explanations, distribution as the shape of the data, a correlation extension task, proxy variables inside an algorithmic risk-score activity. Unit 6 handles the probabilistic/deterministic distinction explicitly.

Still missing: probability itself, sampling design, quantified uncertainty, and the formal inference the report names. No unit asks for a claim with an attached degree of certainty. CET Partial on component 2 only. DGD I Incidental and WDD I None — the apparent hits there are false positives on everyday words like "random."

Competency 5 — Models and representations

Component 1 is well covered everywhere: storyboards and level diagrams, wireframes and mockups, flowcharts, binary encoding as representation, network-layer abstraction.

Component 2 was already strong in ACS I and was under-scored in the first two passes. A chart-critique lesson runs four charts with known flaws — truncated axis, truncated axis over time, wrong chart type, dual axes implying correlation — and asks of each: what conclusion is it pushing, does the data support it, what would you change. The dual-axis chart is the strongest beat, because nothing in it is false; the misleading part is a scaling choice rather than a data error, which is exactly the judgment component 2 asks for.

Component 3 was the real gap. It is now closed in ACS I and remains open in the other three, where students produce representations constantly and are never asked what those representations discard.

The C5 close-out

Two edits, no added class time, both hosted in lessons that already existed:

• Compression lesson — a new two-minute beat generalizing from lossy compression to representation as such: lossy throws data away and admits it, every other representation does the same without announcing it, and a representation that kept everything would be the thing itself.
• Chart critique — the worksheet gains a single question, asked once rather than per chart: assume your fix is applied and the chart is now honest. What does it still not show? The exit ticket became the same question, and the assessment note now distinguishes incomplete from dishonest as the thing being graded. Runtime unchanged at 30 minutes.

The design choice worth preserving: component 3 is taught as a working question, not an accusation. Something is always missing; the skill is naming what, and judging whether it matters for the question being asked. Compression is the ideal host because it is the one case where the discarding is deliberate, visible, and not a flaw.

What this means

• ACS I is Strong on six of seven. Portfolio-wide, competencies 1, 3, 6, and 7 are Strong. Competency 6 is strong in all four courses and no other competency is.
• Competency 4 is the gap, but a narrower one than it first appeared. The conceptual half is taught; the statistical half — probability, sampling, uncertainty, formal inference — is absent everywhere.
• ACS I should probably stop at six. The report specifies inference using hypothesis tests and confidence intervals, which is a statistics course's job. The honest ceiling in a CS course carrying an AP exam is Strong on components 1 and 2 with a stated, reasoned Partial on component 3. A declared partial is more credible to a district reader than a full row of checkmarks that a framework-literate reviewer would probe.
• The cheap Competency 4 add, if it is ever wanted: reframe an existing fictional survey as a sample and resample it in the browser so students watch the average move. That is inference with a degree of certainty and no formulas, it is already aligned to the AP CSP framework, and it ties back to the deterministic-versus-probabilistic lesson. One to two class periods. Not scheduled.
• Competency 5 was the most fixable. The same move — one question appended to a critique that is already happening — is available in the other three courses wherever a wireframe, storyboard, or diagram is already being reviewed.
• DGD I is the weak course — the only None on Competency 2, and Partial or worse on five of seven.
• The report names its own gap as instructional guidance and teacher professional development — PD tailored to the disciplines teachers actually teach, expanded preservice pathways, and familiarity with the competencies for all preservice teachers regardless of subject. That is the part of the report with the least existing infrastructure behind it.
• Cheapest real gains, in order: limits-of-model questions attached to critiques already happening (C5, all four courses) · sampling and uncertainty added where the data infrastructure already exists (C4) · teach robotics sensor output as produced data (C2) · a data strand in DGD I, where playtest telemetry is the natural fit (C2).

Method and limits

This section exists because a coverage claim without one is worth very little, and because the first version of this analysis was wrong in three places.

• Regex-and-context scan over 448 lesson files across the four courses, grouped by unit. Excluded: teaching notes, crosswalk sheets, audit summaries, syllabi, and index files. Instruction files only.
• Every rating at Partial or above on the contested competencies (2, 4, 5) was hand-verified by reading the matched lesson text in context. That check moved ratings in both directions across two passes.
• Ratings are analytic judgment against the report's stated components, not a validated instrument. They are defensible in a portfolio or PD context. They are not a formal alignment study.
• Not a full read of all 448 files. Cells rated None or Incidental were not individually verified beyond spot checks, so a competency taught without any of its standard vocabulary would be missed.
• Scoring from the practitioner summary alone produced three wrong ratings. Score against the report.