English teaching pack · हिन्दी में गतिविधियाँ
Print or save as PDF using your browser’s Print menu (Ctrl+P / ⌘P). Swipe wide tables sideways to see every column.
Class 8 · Artificial Intelligence
Chapter 1 Teaching Pack
AI Project Lifecycle — a complete, print-ready plan for teaching the six stages and the judgement that comes before them.
Where this chapter sits
Class 8 begins by moving students from recognising examples of AI to planning a complete AI project. Chapter 1 gives them the process they will need whenever they later examine an application or propose a solution. Teach the order securely, but spend more time on the decisions made inside each stage.
| # | Chapter | Learning focus | Place in the sequence |
|---|---|---|---|
| 1 | AI Project Lifecycle | Six iterative stages; deciding whether AI is necessary; planning data, testing, release and upkeep | Process foundation |
| 2 | Artificial Intelligence and Its Applications | Uses of AI in environmental work, automation, healthcare and education | Applies the foundation to domains |
The suggested seven-period sequence in this pack is a classroom plan, not a prescribed allocation.
Lesson plan · 7 periods
Period 1 — Start with the honest question: is AI needed?
Put this brief on the board: “At 2:00 p.m., a pump switches off.” Ask teams to design the simplest dependable solution. A timer and a fixed rule will do; learning from examples adds cost without solving a new problem.
Now change the brief: “Before lunch, estimate whether demand will be high or regular using past records.” Fixed rules may provide a baseline, but patterns in data can make an AI approach worth investigating. The correct first move is not “choose a model”; it is define the need and compare it with simple automation.
Period 2 — Build the six-stage map
Give six pairs one stage name each. Ask them to stand in the order in which a responsible team would begin work. Then reveal the map below. Do not draw it as a one-way finishing line: feedback from real use is part of the lifecycle.
Period 3 — Stages 1–2: define before you collect
Use a canteen question: “Will tomorrow’s lunch demand be high or regular?” At Define the problem, identify the decision, user and useful outcome. At Data collection and preparation, choose relevant past records, then deal with missing, repeated, incorrect or unrelated entries. Labelling each past day “high” or “regular” makes the intended answer explicit.
Where students go wrong: they list every field available. Ask what decision each field helps. Collecting more columns is not the same as collecting better evidence.
Period 4 — Stages 3–4: learning is not testing
At Model development and training, a model uses training examples to learn useful relationships. Keep some examples separate. At Model evaluation and refinement, use those unseen examples to check performance, investigate errors, clean or add data where justified, adjust the model, and train again.
Period 5 — Stages 5–6: real use changes the evidence
Deployment places the tested model in the situation where a user can act on its output. Monitoring and maintenance checks predictions and failures after release, notices when conditions change, fixes errors, adds suitable recent data and retrains when needed.
Ask: “A canteen predictor worked in July but fails during examination weeks. Which stage detects this, and where might the team return?” Monitoring detects the drop. The cause decides the return path: perhaps the problem omitted examination days, the data lacked them, or the model needs retraining and evaluation.
Period 6 — Lifecycle clinic
Read short team statements aloud: “We tested on the same rows used to teach the model”; “We released it, so the project is complete”; “We collected names because the register had them”; “We chose AI before writing the problem.” Pairs identify the faulty stage and repair the plan in one sentence. Insist on a reason, not only a stage name.
Period 7 — Run the unplugged simulation
Use Section 05. It makes all six stages visible on one sheet and forces monitoring evidence to send the plan backwards.
Worksheet
Name: Class & Section: Date:
A · Recall and check
B · Name the stage
Write the exact lifecycle stage that best matches each action.
C · Apply and explain
D · Diagnose a project
E · Think beyond the labels
Answer key & teaching notes
| Q | Answer | What to watch for |
|---|---|---|
| 1 | (b) Define the problem | Students often start with data because it feels practical. The goal and need come first. |
| 2 | (a) Data collection and preparation | Removing duplicates is preparation; merely gathering records is collection. Both belong to the same stage. |
| 3 | (c) Putting a tested model into use | Deployment does not mean training or testing. It connects the evaluated model to a real user or process. |
| 4 | (d) Monitoring and maintenance | The timing clue is “after release”. Checking is monitoring; updating is maintenance. |
| 5 | 75% | 18 ÷ 24 = 0.75, then × 100. Watch for 6/24 = 25%, which is the error rate, not the accuracy. |
| 6 | To check whether learning works on new cases, rather than measuring recall of training examples. | “For fairness” is too vague. Look for the distinction between training examples and unseen test cases. |
| 7 | Define the problem | Deciding whether AI is warranted is part of this stage, before collecting data. |
| 8 | Data collection and preparation | All three clues belong here: gathering, correcting and labelling. |
| 9 | Model development and training | The word “learns” points to training, but the model must still match the problem defined earlier. |
| 10 | Model evaluation and refinement | Accept the full paired name only; the action includes both checking and improving. |
| 11 | Deployment | Students may choose monitoring because a display is visible. The key action is first putting the model into use. |
| 12 | Monitoring and maintenance | Comparing later predictions is monitoring; updating the system is maintenance. |
| 13 | No. Use a clock or timer with the fixed timetable rules. | Fast or automatic is not the same as AI. Full credit requires the stable-rule reason. |
| 14 | Any two relevant items, such as period number, weekday, timetable, past visitor count or scheduled activity; plus one issue such as a missing, duplicate, incorrect or unlabelled record. | Do not reward personal details merely because they are available. Each field should help estimate busy versus quiet. |
| 15 | Conditions or usage patterns can change, errors can appear in real cases, or performance can fall over time. | “To make it better” needs a cause or a check. Accuracy before release is not a permanent guarantee. |
| 16 | Any two: the data may not answer the real need; unnecessary data may be collected; the output may be unclear; simple automation may have been sufficient; success cannot be judged clearly. | Look for consequences of the skipped decision, not generic claims that “the model will fail”. |
| 17 | Example: predictions weaken after a timetable change, so return to Data collection and preparation, add representative recent records, then retrain and evaluate. | Other sound routes are valid. The evidence, earlier stage and corrective action must form one causal chain. |
| 18 | Three errors: five special-day records are too narrow for ordinary weeks; the test cases were reused from training; 100% on those rows was treated as proof for a different setting. Correct by gathering representative ordinary-week records, holding out unseen test cases, evaluating there, and refining before deployment. | Students must provide three mistake–correction pairs. “Use more data” alone misses representativeness and honest evaluation. |
| 19 | For the stated threshold job, AI is not justified: a level sensor or float switch plus a fixed rule can stop the pump. A new predictive need—such as estimating tomorrow’s water demand from past use and scheduled events—could make learning from data relevant. Revisit Define the problem first, then plan suitable data. | This discriminates reasoning from memorisation. A different new requirement is valid only if examples and learned patterns are genuinely useful. More equipment or more complexity is not enough. |
Unplugged activity · “Lifecycle Control Room”
Period 7 · 40 minutes · No devices, internet or lab required · Designed for 40–48 students.
What you need
Eight sheets of used-on-one-side paper, six small paper slips per group, pens or pencils, the blackboard and chalk. Make eight groups of five or six. Each group needs a reader, scribe, timekeeper and three stage checkers; double up roles in groups of five.
The shared brief
How it runs
- Order the map — 5 minutes. Each group writes one exact stage name on each of six slips, mixes them, then arranges them in order. Check all maps together before continuing.
- Challenge the need — 5 minutes. Groups write one simple rule-based baseline, such as planning from the fixed weekly menu, and one reason patterns in past demand might add value. A group may recommend testing the baseline first.
- Plan the work — 12 minutes. Divide the large sheet into six boxes. Under each exact stage heading, write one concrete team action. Require an unseen test set at Stage 4 and a named canteen user at Stage 5.
- Audit another team — 6 minutes. Pass each sheet clockwise. The receiving group circles one weak or missing action and writes the stage that should repair it.
- Release the monitoring card — 6 minutes. Write on the board: “After Saturday clubs begin, the system repeatedly predicts Regular when demand is High.” Groups draw a feedback arrow from Stage 6 to an earlier stage and write the corrective action beside it.
- Debrief — 6 minutes. Ask three groups to defend different return paths. Returning to Stage 2 for representative Saturday records is sensible; returning to Stage 1 is also sensible if Saturday users were outside the original problem. The evidence must determine the route.
Project brief & rubric
This is a planning project, not a software build. It assesses whether students can make connected lifecycle decisions before tools distract them.
Submission
Two A4 pages or one chart, plus a two-minute team viva. Every member must be able to explain the AI-versus-automation judgement and one feedback route.
| Criterion | 4 — Exceeds | 3 — Meets | 2 — Approaching | 1 — Beginning |
|---|---|---|---|---|
| Problem and AI judgement | Precise user, outcome and success check; compares a credible baseline and gives a conditional, evidence-based verdict | Clear problem and user; compares automation with AI and gives a sound reason | Problem is broad or comparison is weak; verdict has limited support | Starts with AI without defining the need or considering a fixed-rule option |
| Lifecycle order and completeness | All six exact stages in order, with connected actions and a justified feedback route | All six exact stages in order, with a relevant action at each | One stage is missing, misplaced or only named | Several stages are missing, reordered or confused |
| Data, training and evaluation | Relevant fields, preparation and labels; clearly separate sets; metric and error analysis fit the prediction | Suitable data and preparation; training and unseen test cases are separate; percentage correct is planned | Some useful data, but preparation or honest testing is unclear | Data is unrelated, or training and evaluation are treated as the same task |
| Deployment, monitoring and maintenance | User action, failure checks, changed conditions and corrective update are specific and linked | Real user and use are clear; performance is checked and an update is planned | Release is described, but later checking or maintenance is vague | Stops at deployment or assumes performance will remain unchanged |
Suggested total: 16 marks. Record the four criterion scores separately so the feedback shows which lifecycle decisions need work.
Evidence record
Keep one completed record per class for this chapter, with the selected samples named below. This page records what was taught and assessed; it is not a review or endorsement.
| Field | Record |
|---|---|
| School | |
| Class & section | |
| Chapter taught | AI Ch. 1 — AI Project Lifecycle |
| Periods used | |
| Dates | |
| Teacher | |
| Activity conducted | Lifecycle Control Room (unplugged group simulation) |
| Assessment used | Canteen Demand Planner, rubric-scored |
| Students assessed | |
| Common misconception noticed | |
| Follow-up taught | |
| Samples retained | ☐ 3 marked proposals ☐ 3 completed worksheets ☐ 1 group lifecycle sheet |
| Teacher’s note | |
| Teacher signature & date |