Teaching · Practical guide
Set Learning Goals That an AI-Assisted Lesson Can Actually Assess
A prompt such as “teach networking” does not specify what a learner should be able to do. In this lesson you will build a goal, an activity and an assessment that fit together. AI can suggest wording, but you decide which evidence counts as learning.
Distinguish an intention from evidence
An intention names a broad direction: improve communication, understand AI or become better at mathematics. A learning goal states a performance that can be observed. For example, “explain why a generated claim needs independent evidence” is more assessable than “appreciate verification.” The observable performance gives both learner and teacher something to inspect.
Carnegie Mellon’s learning-objectives guidance connects objectives, assessment and instructional activities. For this DYAP exercise, use a practical sentence pattern: given a condition, the learner will perform an action, producing evidence that meets stated criteria. This is a planning aid, not a mandatory institutional formula. Preserve the wording and level of any approved course learning outcomes.
Conditions clarify the task: with a reference sheet, without notes, using a provided data table, or in a disposable lab. Criteria describe successful evidence: correct calculation, explicit assumptions, justified choice or reproducible procedure. Avoid unnecessary numerical thresholds just to make a goal look precise. “80% accuracy” means little until you define the tasks, their difficulty and the consequences of errors.
Build a manageable goal hierarchy
Start with the course outcome, then select a smaller lesson goal that contributes to it. A course may require learners to evaluate AI-supported work. One lesson could focus only on distinguishing unsupported claims from calculations. Another could address sources. A single short session cannot credibly promise mastery of every part of a broad outcome.
Specify prerequisites separately. If learners cannot yet calculate a percentage, evaluating percentage claims may require a short arithmetic review. That does not mean lowering the intended outcome; it means identifying the necessary preparation. Ask AI to propose prerequisite questions, then check whether each actually tests a required skill.
Describe the finished evidence before listing activities. A completed worksheet, corrected explanation or annotated calculation is easier to review than a statement that a learner “watched a video.” Participation can support learning, but completing an activity is not by itself evidence that the target skill was acquired.
Worked example: align a percentage lesson
Initial goal: “Students will understand percentages using AI.” This combines an unclear performance with a tool requirement. Revised goal: “Given two fictional attendance counts, calculate the attendance percentage and explain why the registered total is the denominator.” AI is optional; the goal concerns the reasoning, not use of a particular product.
Activity: learners compare two generated explanations for 27 attendees out of 36 registered participants. One says 27%; the other calculates 27 ÷ 36 × 100 = 75%. Learners choose the supported answer, annotate the numerator and denominator, then write their own explanation. The teacher checks that the example counts represent the same session and population.
Assessment: use a new case with 21 attendees out of 28 registered participants. Require the ratio, calculation and one sentence explaining the denominator. The answer is 75%, because attendance is compared with all registered participants. The new counts prevent simple copying while preserving the skill.
A three-part success checklist is appropriate here: correct ratio 21/28; correct percentage 75%; clear statement that 28 is the registered total. A multiple-choice question asking only the definition of “denominator” would not fully assess the goal. Nor would a beautifully formatted slide deck demonstrate that the learner can calculate and explain.
Use AI to challenge alignment
Give the model the goal, activity and proposed assessment together. Ask it to identify mismatches rather than merely improve the wording. It might notice that the activity supplies every calculation while the assessment requires independent calculation. Decide whether supported practice followed by independent practice is needed before assessment.
Check verbs carefully. “Compare,” “justify” and “design” imply different evidence from “list” or “identify.” A generated goal can sound sophisticated while the task still demands recall only. Ask what a successful learner would physically write, say, create or demonstrate. If you cannot answer, the goal is not ready.
Keep scope realistic. If learners have ten minutes, requiring three research sources, a calculation and an oral defence is probably too much for an introductory activity. You can split the work into a sequence while retaining a clear end goal. Do not treat the model’s suggested timing as a measured estimate.
Practice: repair a mismatched lesson in 20 minutes
Minutes 0–4: inspect this plan: goal—“justify whether an AI summary is supported”; activity—copy a summary; assessment—name two chatbot brands. Write why the three parts do not align. Minutes 4–9: choose one short fictional source paragraph and formulate a goal about checking claims against it.
Minutes 9–14: design an activity requiring learners to label supported and unsupported statements. Minutes 14–20: create a fresh two-claim assessment and write the evidence you will accept. Example source: “The workshop had 12 places and 9 bookings.” Example claims: “75% of places were booked” and “people avoided it because it was difficult.”
Example solution: the goal is to justify two claim classifications using the source. The first claim is supported by 9 ÷ 12 × 100 = 75%; the second invents a cause. Successful evidence includes the calculation and an explanation that booking totals do not reveal motivation. An activity comparing claims prepares learners for that assessment more directly than copying a summary.
Self-review: does the assessment require the action named in the goal? Has the activity prepared learners for that action? Are conditions and expected evidence clear without inventing a new institutional grading policy? Save the aligned version and one note about the mismatch you removed.
Reusable prompt
Select and copy this template, or customise your own using the free builder.
Review this learning design: goal [text], learner prerequisites [text], activity [text], assessment [text], available time [minutes]. Identify specific alignment gaps. Rewrite one observable lesson goal with conditions and evidence, propose a matching practice task and an assessment using a new example. Keep the approved course outcome unchanged. Explain what a learner would produce; do not claim that the plan has been tested.Open the prompt builder →
Pause · Think · Check
Test your understanding
Choose one answer per question, then check your reasoning. This is practice; scores are not saved.