Match the learning approach to the problem

01 / Begin with the outcome

Your learning targets

Unit goal: Make a defensible decision about a small AI pilot before collecting data or building a model.

What must the system produce, and what learning signal is actually available?

O1
I can match a task to an appropriate learning approach.
Success looks like: All four matches are correct; each justification identifies the output and whether the task uses labeled examples, unlabeled patterns, or reward feedback.
Pearson AI 1.2
O2
I can tell which values are inputs and which value is the answer to learn.
Success looks like: Both arrival-time inputs and the target are correctly identified; the future-only field is excluded with a time-based explanation.
Pearson AI 1.2
O3
I can explain what a grouping does not prove.
Success looks like: The critique distinguishes similarity from a reviewed label and proposes one appropriate validation step without asserting security ground truth.
Pearson AI 1.2

Before you begin · materials and readiness

AI-01 or an equivalent project-framing discussion. No programming or model training is required. An input is information available to a system; an output is the result it produces.

The task cards and fictional rows on this page
A four-column organizer: task, output, learning signal, justification
Paper or a text document

Scope and safety

Use the fictional records only. Do not train on school tickets or use live network equipment as a reinforcement-learning environment. This activity selects an approach; it does not train or validate a model.

Keep your work on paper or in your own document. This site does not save answers, submit work, or award credit. Follow your teacher’s submission instructions.

02 / The essentials

Build your understanding

Predict a category or a quantity

In supervised learning, examples include a target answer, often called a label. Classification predicts a category; regression predicts a numeric quantity. A numeric label such as queue 1, 2, or 3 can still represent categories. Ask what the number means, not just whether it looks numeric.

Find patterns without target answers

Unsupervised learning looks for structure without a supplied target answer for each example. Clustering groups similar records. K-means is a clustering method, not ordinary supervised classification. A person still has to investigate what a discovered group means; cluster membership alone does not prove a cause or threat.

Learn from actions and rewards

Reinforcement learning concerns an agent choosing actions in an environment using reward feedback. In our classroom example, the environment would be a sandbox simulator. Reading labeled tickets repeatedly is not reinforcement learning merely because the system gets better.

Recognize generation as a different output

Drafting a paragraph is a content-generation task, not simply predicting a queue or a repair duration. Generative systems can involve more than one training approach. Prompting an existing system is also different from choosing how to train your own model. This lesson focuses on the four task types needed for the cards.

Respect the prediction clock

For a ticket routed when it arrives, the summary and selected device type may be available. A closing note written after repair is not. Using that future-only field during evaluation could create misleading results. A target label may be established later for training; it is not an input secretly available at prediction time.

Check your thinking

Answer before opening the explanation. These are practice checks, not a graded test.

Does a target stored as 1, 2, or 3 automatically mean regression?

No. If those values encode categories, such as queues, the task is classification.

Does an unlabeled dataset mean no useful work is possible?

No. You may explore patterns, or arrange an authorized labeling process. But you cannot pretend reviewed targets already exist.

Can a model use the final repair note when routing a newly arrived ticket?

Not if the note is only written after repair. The experiment must respect what is actually known at prediction time.

03 / Worked example

A queue name is a category—even when stored as a number

The codes and data below are fictional. The prediction is made when a ticket arrives.

Fictional row W-AI2
Input: summary = “Display stays blank”
Input: device_type = desktop
Reviewed target: queue_code = 2
Codebook: 1 = Accounts, 2 = Hardware, 3 = Network
Later record: resolution_note = “Replaced failed display cable”

How a technician reasons through it

The requested output is a queue category. Predicting its code is a classification task, not a regression task.
The reviewed queue supplies a supervised target. A new arrival has a summary and device type but does not yet have that reviewed answer.
The resolution note happens later. Exclude it from an arrival-time prediction input, even if it makes an experiment look accurate.
If the task instead asks how many minutes repair will take, the target changes to a numeric duration. The same organization can have several distinct AI tasks.

04 / Guided practice

Try it with support

For each task, explain the output and signal first; name the approach second.

Fictional tasks G-AI2
A: Predict repair minutes from records with actual past durations.
B: Group unlabeled device-event summaries by similarity for staff review.
C: Learn action choices in a resettable simulator using a reward for completed work and a penalty for unnecessary changes.

Match A, B, and C to an approach and give a reason for each.
Explain why the groups in B cannot automatically be called hardware faults or attacks.
For C, describe why live school infrastructure is not an acceptable practice environment.

Compare your reasoning

A is supervised regression: the target is a measured quantity. B is unsupervised clustering: there are no reviewed target categories. C is reinforcement learning in the described simulator: actions receive feedback through rewards. The groups in B need domain review; a cluster label is not proof. Unreviewed exploratory actions on school systems could disrupt services, so C stays in a separately approved simulator.

05 / Independent application

Show what you can do

Use these four cards and the schema to produce your own task analysis.

Fictional tasks I-AI2
A: Predict a reviewed queue from ticket summaries and known past queue labels.
B: Predict total repair minutes from examples with recorded durations.
C: Discover similar patterns among unlabeled event summaries.
D: Choose actions in a simulated environment with rewards for good outcomes.

Arrival-time queue schema:
summary — available at arrival
device_type — available at arrival
reviewed_queue — target assigned by a reviewer
final_resolution — written after repair

Claim about C: “Cluster 3 proves these devices were attacked.”

Complete task, output, learning signal, and justification for A–D (O1).
Identify the two candidate arrival-time inputs, the target label, and the field to exclude. Explain why the excluded field would mislead the experiment (O2).
Critique the claim about Cluster 3 and propose one safe next validation step (O3).

Evidence checklist

Use the targets above to decide what is demonstrated and what needs revision. The criteria are local lesson expectations, not certification passing scores.

O1: Submit the four-row task organizer.
O2: Submit the annotated schema and a leakage explanation.
O3: Submit a two- or three-sentence critique and a next check.

Review after your first attempt

A: supervised classification; B: supervised regression; C: unsupervised clustering; D: reinforcement learning. The candidate inputs are summary and device_type, the target is reviewed_queue, and final_resolution must be excluded from an arrival-time prediction. A discovered cluster is evidence of similarity under a chosen representation, not a confirmed attack. A defensible next step is authorized domain review of representative fictional records and comparison with independent evidence.

Self-review supports learning. Your teacher determines whether the evidence demonstrates each objective.

06 / Feedback into progress

Reflect, revise, try again

Exit prompt

Change one card’s required output so that a different approach fits. Explain exactly what changed.

Need another route?

Sort cards into “known answer examples,” “no target answers,” and “actions with rewards.” For known answers, distinguish category from quantity. Retry with teacher-provided fresh cards and the same criteria.

Ready to go further?

Design a fictional assistant with both queue prediction and a draft explanation. Explain why those are separate outputs and why each needs its own evaluation.

Identify one piece of evidence for each “I can” target. If the evidence is missing, choose a practice step and try a fresh case with feedback.

Teacher plan · objectives, assessment & teaching sequence

Teacher plan

Prepare a four-column organizer and the fictional schema. Correct the common confusion between clustering and classification; do not present a cluster as a verified security label.

Lesson Planning Guide

Objectives, practice, and assessment alignment

Audience: Introductory IT Essentials CTE learners. Each objective specifies audience, behavior, condition, and degree (ABCD).

O1 · Cognitive · Analyze

Pearson AI 1.2 — Initial practice: distinguish problem types and the available learning signal. Algorithm selection and training are not assessed.

Condition: Given the four independent task cards and the lesson reference
Behavior: The learner will classify four IT task cards and justify each choice using the output and available signal.
Degree: All four matches are correct; each justification identifies the output and whether the task uses labeled examples, unlabeled patterns, or reward feedback.
Aligned practice: Compare classification, regression, clustering, and reinforcement cases.
Assessment evidence: Submit the four-row task organizer.

O2 · Cognitive · Analyze

A2 A7

Pearson AI 1.2 — Initial practice: recognize labeled data for the stated problem. Detailed feature engineering belongs to Unit 02.

Condition: Given the independent ticket schema and the time of prediction
Behavior: The learner will identify suitable inputs, a target label, and an unavailable-at-prediction field.
Degree: Both arrival-time inputs and the target are correctly identified; the future-only field is excluded with a time-based explanation.
Aligned practice: Inspect the worked row and explain why a resolution fact cannot be used at arrival.
Assessment evidence: Submit the annotated schema and a leakage explanation.

O3 · Cognitive · Evaluate

A2 A6

Pearson AI 1.2 — Initial practice: interpret an unsupervised framing without treating it as labeled classification.

Condition: Given the independent clustering task and the claim below
Behavior: The learner will critique a claim that discovered clusters are verified incident categories.
Degree: The critique distinguishes similarity from a reviewed label and proposes one appropriate validation step without asserting security ground truth.
Aligned practice: Discuss why cluster numbers do not automatically name a cause.
Assessment evidence: Submit a two- or three-sentence critique and a next check.

Teaching sequence

Hook and retrieval
Teacher: Ask whether queue code 2 is a quantity or a category.
Learner: Explain what the value means.
Check and respond: Separate the representation from the task.
Model the distinction
Teacher: Teach the output/signal distinctions and the prediction clock.
Learner: Annotate the worked record and exclude future-only input.
Check and respond: Use the numeric-code and future-note checks.
Guided card sort
Teacher: Assign explainer and challenger roles, then swap.
Learner: Match G-AI2 and challenge unsupported cluster claims.
Check and respond: Ask for output and signal before accepting a method name.
Independent evidence
Teacher: Assess the four matches, annotated schema, and critique individually.
Learner: Complete I-AI2 using the reference.
Check and respond: Require a justification, not just four memorized labels.
Feedback and closure
Teacher: Return a misconception-specific revision prompt.
Learner: Revise and change a card’s output in the exit task.
Check and respond: Use a fresh card if the distinction remains unclear.

Use the classroom toolkit for a timer or groups. Participation tools do not replace individual evidence of learning.

Access, differentiation, and professional practice

Supply plain-text cards or printed copies; never rely on color alone.
Offer a vocabulary organizer with examples and allow students to explain choices aloud before writing.
Allow typed, handwritten, or dictated analysis under the same criteria; do not require a device or home AI account.

Academic connection

Categorization, technical reading, and reasoning about time-dependent information. No formal academic-standard code is claimed.

Professional and social-emotional practice

Practice questioning a strong claim without dismissing a colleague. Reward the evidence and explanation, not speed at naming terms.

Purposeful technology use

Use simulations on paper to reason safely before later model-training labs. This lesson demonstrates analysis, not operation of a training platform.

Plan for actual learners privately. Do not publish demographics, disability information, IEP/504 details, grades, or identifying student work in these open lesson pages.

After teaching · reflection and next steps

Note whether errors involve output type, signal, or future information. Reteach the specific distinction and reassess with a new case; keep individual records private.

Before classroom use, verify local procedures, materials, accommodations, pacing, and assessment expectations. Draft lessons have not yet been classroom-validated.

Competencies, certifications & sources

Standards, credentials, and work

A2 A7 A6

Links open the full Michigan competency statements. The objective alignment above shows the portion practiced here; it is not an official MDE or exam-provider crosswalk.

Certification preparation

Pearson IT Specialist Artificial Intelligence objective 1.2: introductory, partial practice. Feature engineering, algorithm implementation, and model evaluation will require later lessons.

Career connection

IT teams need to translate a service request into the right technical task before selecting a tool or working with data specialists.

NIST and compliance context

Local AI RMF Map connection: describe what the system is meant to do and what its output means. No risk-framework outcome is signed off here.

An instructional connection is not evidence that an organization meets a regulation or security control. NIST frameworks, state competencies, and certification objectives serve different purposes.

References and review status

Pearson · Artificial Intelligence objectives
Objective 1.2 anchors the topic. The task cards and criteria are original local drafts.
Google · What is machine learning?
Background on learning signals and output types. No Google exercises or assessment items are reproduced.
Google · What is clustering?
Clarifies clustering as an unsupervised approach, including k-means. Task labels should not be confused with algorithm families.

Original IT Essentials lesson · Draft v1 · September 10, 2026. Original lesson explanations and activities; linked technical sources inform the content. No institutional endorsement is implied.