Skip to main content

Module 7: AI Literacy and Assisted Engineering

Use AI as a tutor, debugging partner and reviewer while retaining the ability to verify its output. Build a small evaluator that exposes plausible-but-wrong suggestions. This is practical model literacy, not a machine-learning theory course or a repetition of the ethics strand.

Planning allowance: 12–18 focused hours, included in semester workload. Authored candidate; executable checks do not replace independent review or learner calibration. Full-book reading is not required.

Before you start​

Complete the immediate Programming Basics gate. Use Learning and Calibration for independent attempts and Reading and Evidence for claims. The AI/privacy strand remains required; link its declaration instead of duplicating it. No model account or paid API access is required.

Diagnostic: Write the expected count for [24,25,25,26] at an inclusive threshold of 25. Explain why a confident suggestion is not a test result. You may be new to AI tools; the offline route teaches the same verification workflow without an account.

Outcomes​

  • Distinguish model output, context, retrieval and tool actions.
  • Specify bounded assistance and preserve an independent first attempt.
  • Evaluate suggestions against explicit schema and behavioral criteria.
  • Recognize untrusted instructions and choose controlled tool permissions.

Concept map​

Concept clusters and learning order​

Cluster 1: Model Literacy​

Apply both lessons in Practice 1.

Cluster 2: Assisted Learning​

Apply both lessons in Practice 2.

Cluster 3: Verification and Control​

Apply both lessons in Practice 3.

Build, review and assess​

Use worked examples, guided lab, exercises, case studies and mistake clinic. Finish the independent assessment under the rubric, then consult calibration.

Workload and outputs​

Lessons/selected reading 4–6; prompt and review practice 2–3; evaluator lab 4–6; assessment and one bounded repair 2–3: 12–18 hours. Existing AI disclosure/privacy work remains in the strand allocation.

Preserve your implementation, original predictions, actual results, a repaired defect and transfer evidence. Use existing Git/assistance records; keep generated code and copied examples distinguishable from independent work. The common lab archive is downloaded once, but each module implements and verifies a distinct boundary.

Depth and future connection​

Reading guide → resources → reference. Capstone throughline explains what to retain. Technical integration specifies prerequisites, evidence reuse and later-semester boundaries.

Bridge from Launchpad​

Bring the no-code AI learning record. Use the first cluster to apply its model/task concepts to functions, source context and code-review contracts, not to repeat the earlier schedule exercise. The 12–18 hours here cover technical lessons, tutoring/debug/review practice and a Python evaluator with independently specified cases. The Launchpad 6–9 hours are separate preparation, counted only there.