AI in the Classroom: What to Allow and What to Stop
A blanket ban on AI is easy to get around and teaches students the rule is arbitrary. A specific policy, with one no-AI zone and one disclosure line, actually changes what happens in the lab.
Where AI earns its place
Three uses work well in a school. A patient tutor: a student can ask the same question eleven times without embarrassment, which is the single biggest advantage over a teacher. A feedback generator: it comments on a first draft in seconds, so students write more and revise more. And a data tool: it turns a week of sensor logs into charts, so a Class 8 project can include analysis instead of a table of numbers.
Three uses do not work. Marking anything that is supposed to measure unaided understanding, because the tool removes the thing being measured. Letting a model write the report, because the report is where the thinking is supposed to happen. And pretending the model is always right, because it is confidently wrong often enough that students have to learn to check it.
Rules that keep it honest
The rules that work are short and specific. Declare a no-AI zone, which is any assessment that measures a skill you need students to own. Require disclosure, in the form of one line in the report saying where AI was used and for what. Assess the process and not just the output, by marking the data, the code, the error log and the explanation. And require a human check on every fact a model supplied.
There is one more that is not about teaching. Anything you upload may be stored and may be used to improve the service, so student names, photographs, marks and records stay out of public tools. If a project needs AI at scale, run it on school hardware or an approved account, and check your own data protection rules before you start rather than after.
Step by step
-
Write a one-page policy with the students
Three sentences on what is allowed, what is not and how to declare it. A policy students helped write is one they follow.
-
Choose the no-AI zone deliberately
Decide which assessments measure unaided skill, such as the timed practical, and protect them without exception.
-
Set the disclosure format
One line: AI used for X, verified by Y. It costs a student ten seconds and it makes the rest of the work trustworthy.
-
Pick one project where AI is the tool
A class dataset is ideal. Every student logs sensor data and the model helps turn it into charts and questions, while the electronics stays theirs.
-
Verify AI-sourced facts by hand
Pick three claims from any AI-written text and check them against a book or a datasheet. Do it as a class, and it teaches the habit faster than any policy.
-
Review after a month
Look at what students actually did, change the policy, and write down why. A policy that has never been revised is a poster rather than a rule.
What you'll need
| Item | Qty | Why |
|---|---|---|
| Laptop per pair of students | 1 per pair | The tool itself. A shared device per pair is enough for the tutor and drafting uses, and it keeps the licence cost down. |
| Raspberry Pi 5 with 8 GB | 1 per lab | Runs a local model so nothing leaves the building and there is no per-token bill to worry about each term. |
| Arduino kit with Uno R3 and sensors | 1 per pair | Where the data comes from, and where the assessment is unambiguous because the wiring is physical. |
| Ethernet switch and cables | 1 set | Keeps lab traffic off the home internet connection and off public wifi, which matters for anything you do not want to upload. |
| USB to serial adapters | 1 per bench | Lets you log sensor data from every board at once without unplugging anything and losing the session. |
| LED bar graph module | 1 per pair | The visible output for the class dataset project, so the data on the chart came from a real circuit. |
Practise it before you build it
Every step above has a matching circuit in the ThinkEdge Learn simulator. The simulator runs a real
Arduino sketch against a modified-nodal-analysis solver, so digitalWrite actually drives the
LED on screen and the readings change when you change the wiring — no hardware, no burnt components.
The Mega has 4 hardware serial ports (Serial, Serial1, Serial2, Serial3). That means you can talk to 4 different devices at the SAME time — like a GPS, a Bluetooth module, an XBee radio, and the USB. Use 3 of those ports to read 3 different sensors in parallel and stream their data to the Serial Monitor as a CSV row every second. This is the foundation of every real Arduino data logger.
A 10-segment LED bar graph is the easiest way to visualize an analog input: spin a potentiometer and 1, 2, 3 ... 10 LEDs light up in sequence. This is a "VU meter" — the kind you see on stereo equipment. The Mega is ideal because it has 10 free digital pins without stealing any from the analog or serial pins. The same code can be extended to 20, 30, or 100 LEDs on a Mega.
The DHT11 is a small blue module that measures temperature (0-50°C) and humidity (20-90%). Inside it has a tiny microcontroller, a humidity sensor, and a thermistor. It speaks a custom one-wire protocol that the Adafruit DHT library decodes for you. You just call dht.readTemperature() and you get a number back.
Mount an HC-SR04 ultrasonic sensor on top of a servo. Sweep the servo from 0° to 180° and back, and at each step measure the distance. Print the results to the Serial Monitor as "angle,distance" pairs. Open the Serial Plotter in the Arduino IDE to see a live radar-like graph of the room. With Processing on your computer, you can build a full visual radar.
You have three ways to summarise a set of numbers, and they are not interchangeable. The MEAN is the total divided by the count — it uses every value, which sounds safest but lets one huge or tiny value drag the whole answer. The MEDIAN is the middle value once sorted — it ignores extremes, so it survives bad data. The MODE is the most common value — the only one of the three that can describe a non-numerical set like shirt sizes. They agree on roughly symmetric data and disagree wildly on lopsided data. A class where one student scored 95 and the rest scored around 40 has a mean that describes nobody. Real world examples: the average income of a country is pulled up by a handful of billionaires. Average house price in a city is meaningless because of one very expensive house. Picking the wrong one is how statistics misleads people.
A real traffic light runs a state machine: GREEN (cars go) → YELLOW (slow down) → RED (cars stop) → PEDESTRIAN WALK (cross the road) → RED → GREEN. The Mega has enough pins to build a full intersection: 3 LEDs for cars in one direction, 3 LEDs for cars in the other, 2 LEDs for the pedestrian signal, and a button to request to cross. Use millis() for non-blocking timing so the button check is responsive.
Common questions
Should schools ban AI?
No. A blanket ban is easy to get around and it teaches students that the rule is arbitrary. What works is a ban where it matters, which is the assessment that measures unaided skill, plus disclosure everywhere else. The aim is a student who knows when to use it and can check its output, not a student who hides it.
What is the risk with student data?
Anything you paste into a public tool may be kept and may be used to improve that service, and you cannot check what happens to it afterwards. Keep names, photographs, marks and any record out of it, or run the model on school hardware with the Raspberry Pi AI HAT. Treat an uploaded photo of a student as a photograph of a child, because that is what it is.
How can I tell if a student used AI?
Usually you cannot, and chasing detection is a bad use of your time. Assess the parts that cannot be faked: the raw data, the circuit, the debug log and the viva. A student who submits a perfect report and cannot explain the wiring has told you what you needed to know, and a good rubric catches that in the room rather than in an argument afterwards.
Related guides
Try the full simulator with your class
103 guided circuit projects, 21 lessons and an in-browser Arduino simulator. The first 15 projects are free, no account needed. NCERT Science mapping on 47 projects, with Arabic and English copy in the regional builds.