Assessment and Pedagogy in CT and AI

 

Assessment and Pedagogy in CT and AI: Rubrics, Competencies, and the Shift to Inquiry and Project-Based Learning

A Domain-II perspective for CT & AI classrooms, Grades 3–8

If you've ever tried to grade a student's trained image classifier the same way you'd grade a spelling test, you already know the problem. Computational Thinking (CT) and Artificial Intelligence (AI) don't live neatly inside right-or-wrong answer columns. A student might write code that "runs" but completely misses the point of the task — or write messy, half-working code that shows exactly the kind of reasoning we want to nurture. Domain-II of the CT & AI framework exists precisely because of this mismatch: pedagogy and assessment have to evolve together, or neither one works.

This post walks through four ideas that, together, make that shift possible: rubrics, competency-based assessment, inquiry-based learning, and project-based learning.

Why Pedagogy and Assessment Can't Be Separated in CT & AI

In most subjects, we can teach first and assess later — cover the content, then test it. CT & AI resist that split. A maze-coding task, a Teachable Machine project, or an ethical chatbot prototype is simultaneously the lesson and the evidence of learning. The moment of "doing" is the moment of "showing." That means the way we plan a CT & AI lesson has to already contain the way we'll judge it — which is exactly what a good rubric is for.

Rubrics: Giving Thinking a Common Language

The biggest practical challenge in CT & AI classrooms is that the artefacts look completely different from grade to grade — a physical maze in Grade 3, a trained classifier in Grade 5, a chatbot prototype in Grade 7 — yet the underlying skills are the same. A rubric solves this by scoring the thinking pattern, not the surface artefact. A four-pillar rubric built around:

• Problem Decomposition — can the student break a big problem into clean, manageable parts?
• Data Literacy & Bias Awareness — does the student notice when data is unbalanced or unfair?
• Algorithmic Design & Logic — is the solution efficient, or just functional by accident?
• Iterative Refinement (Debugging) — does the student treat errors as information, or give up?

...can score a Grade 3 maze and a Grade 8 capstone project on the same four-point scale, because both are really asking the same underlying question: how is this student thinking? That consistency is what lets a school track real growth across years, not just across units.

Competency-Based Assessment: Assessing What Students Can Do

Competency-based assessment asks a different question than a traditional test does. Instead of "What does the student remember?" it asks "What can the student actually do, unaided, with a new problem?" In CT & AI, that shows up as a deliberate mix of:

• Formative checkpoints that happen during learning — live debugging games, peer-prediction sessions where students test a classmate's trained model, structured "bias audit" discussions over a flawed dataset.
• Summative checkpoints that happen at the end of a unit — a completed maze path, a validated model log, a fully built capstone prototype with a public presentation.

The formative layer matters more than it usually gets credit for. A single summative score tells you whether a student succeeded; the formative trail tells you how — whether they got there through genuine reasoning or through guesswork and teacher hints. Competency-based assessment treats that "how" as data, not noise.

Inquiry-Based Learning: Starting With Questions, Not Instructions

CT & AI pedagogy works best when students start from a question rather than a procedure. Instead of "Here are the five steps to train a classifier," inquiry-based teaching starts with: "Why did this model just misclassify three photos in a row?" Students investigate, propose explanations, test them against the data, and revise. The teacher's job shifts from instructor to facilitator — curating a dataset, posing a provocation, and then stepping back to let students argue it out.

This is also where bias and fairness enter naturally. When students are handed a skewed dataset and asked "what's wrong here?" rather than told "this dataset is biased, remember that for the test," the ethical reasoning becomes something they construct, not something they memorise.

Project-Based Learning: From Blocks to Real-World Prototypes

If inquiry is the engine, project-based learning is where it gets steered toward something real. A capstone like a Community AI Prototype — say, an ethical chatbot layout addressing a genuine local problem — forces students to hold decomposition, data literacy, algorithmic logic, and debugging together at once, inside a single sustained piece of work. It's also naturally inclusive: a group project has room for a researcher, a writer, a designer, and a presenter, so a student who isn't yet confident writing logic can still own a meaningful part of the outcome.

Crucially, a well-designed project is the assessment. The rubric doesn't get bolted on afterward — it's visible to students from day one, so they know exactly what "good thinking" looks like before they start building.

Bringing It Together in the Classroom

A few practical anchors for teachers trying to put this into practice:

1. Share the rubric before the project starts, not after — competency-based assessment only works if students know the target.
2. Build in at least one formative checkpoint per week of a project — a peer-prediction session, a bug log, a short bias-audit discussion.
3. Let the first version of a project be rough. Iterative refinement can't be assessed if there's no visible iteration.
4. Reuse one rubric across grade bands wherever possible, scaling the artefact's complexity rather than the criteria.
5. Make the "why" public. Ask students to explain a decision (a dataset choice, a debugging fix) out loud — reasoning that stays in a student's head can't be assessed.

Closing Thought

Rubrics, competency-based assessment, inquiry, and project-based learning aren't four separate initiatives — they're one coherent design. The rubric defines what "good CT & AI thinking" looks like; competency-based assessment insists we measure that thinking directly; inquiry gets students constructing it themselves; and project-based learning gives that construction somewhere real to land. Get the four working together, and assessment stops being something that happens to CT & AI learning, and becomes part of how the learning happens at all.

This post draws on practice developed for Domain-II of the CT & AI framework (Grades 3–8) at Delhi Public Sr Sec School, Balotra. Feedback and classroom adaptations from peer schools are welcome.