AI Literacy in the Classroom: How Robotics Makes Intelligent Systems Visible
AI literacy starts the moment a student can watch an intelligent system work instead of guessing what it did. When a robot detects an object, shows how sure it is, and then moves because of that reading, the whole perceive, decide, act process stops being a mystery and becomes something you can point at on a screen. That is the shift our team cares about most: moving students from typing a prompt into a tool and accepting whatever comes back, to observing an AI process where every step is open to inspection. This article is for the teachers and curriculum leads who want that shift in their own classrooms, and it shows how a robot with a vision sensor gets you there.
Key Takeaways
- Watching a robot perceive, decide, and act is what turns AI literacy from a slogan into a process students can inspect, question, and change for themselves.
- The AI Vision Sensor reports what it sees as readable data: a 320 by 240 pixel view, a confidence score that can read 99%, and a bounding box around each object.
- VEX teaches AI with vision sensors, not chatbots: no personally identifiable information is collected, and images or video never leave the student device.
- The concept scales from Kindergarten up: VEX 123 introduces sensing, and by senior grades students compare how two pre-trained AI models behave on VEX EXP and V5.
- Change one rule in the code, a confidence threshold or a target colour, and the robot behaves differently in front of the class, proving a person shaped the system.
- Book an AI Vision Sensor demo or request a quote from our team, and we will map VEX AIM to your grade band, budget, and program goals.
What is AI literacy, and why does “black box” thinking hold students back?
AI literacy is the ability to understand how an AI system reaches an outcome and to judge that outcome with a clear head. UNESCO puts it plainly in its AI Competency Framework for Students, published in September 2024, which describes AI literacy in terms of knowledge, skills, and values, and stresses that AI should support human decision-making rather than replace it. When students only ever meet AI as a box that gives answers, they learn to trust the box. They stop asking how it decided, whether it was sure, or what would change the result. That habit is the real risk, and it is the one we set out to break.
AI literacy defined: an inspectable process, not a magic trick
VEX Robotics defines AI as a branch of computer science that lets computers do things that would count as evidence of intelligence if a person did them, drawing on David Touretzky’s work in “Teaching AI in K-12.” We like that definition because it keeps AI grounded in ordinary computer science, built on data and algorithms, not on mystique. UNESCO’s student framework organizes the same idea into four areas that give teachers a checklist worth keeping nearby:
- A human-centred mindset. Students understand and hold on to their own agency around AI.
- Ethics of AI. Responsible use, safe practices, and ethics considered from the design stage.
- AI techniques and applications. The foundational knowledge and skills to understand how AI works.
- AI system design. Problem-solving, creativity, and design thinking applied to building AI-driven solutions.
From using AI to observing AI: the shift this article is about
There is a real difference between a student who uses an AI tool and a student who can watch one work. The first types a question into a chatbot and reads the reply, with the reasoning hidden the entire time. The second sees the sensor register an object, reads how confident the system is, follows the rule that fires next, and watches the robot act. Everything is exposed, and everything is open to change. That second experience is what the rest of this article builds, and the graphic below sets the two side by side.
Using AI vs. Observing AI
| Using an AI tool | Observing an AI process |
| A student types a prompt and reads the answer. | A student watches the sensor register an object. |
| The reasoning stays hidden inside the tool. | The confidence score and object data are on screen. |
| The student is a passive user of a result. | The student is an active inspector of each step. |
| Hard to ask why the output came out that way. | Change one rule and the behaviour visibly changes. |
How does a robot make an AI process visible in the classroom?
A robot with a vision sensor makes the process visible because each stage of intelligent behaviour becomes a physical event students can watch in the room. The camera takes in a scene, the system works out what it is looking at, it plans a response, and the robot moves. The Carnegie Mellon Robotics Academy, which trains educators on the VEX AI Vision Sensor, describes this as a cycle of Sense, Perceive, Plan, Act, and it is the backbone of everything that follows. The value of a robot here is that none of these steps stay abstract. Students see the input, the reading, the rule, and the movement, in order.
The intelligent-behaviour cycle: Sense, Perceive, Plan, Act
Carnegie Mellon frames modern robot intelligence as four steps, and each one lines up with something a student can observe on VEX AIM and its screen. The diagram below shows the cycle with the classroom artifact that goes with each stage.
The Sense → Perceive → Plan → Act cycle in the classroom
1 Sense: The AI Vision Sensor captures the scene in front of the robot. Students see the live camera view.
↓
2 Perceive: The system decides what it is looking at, for example “this is a sports ball,” and how sure it is.
↓
3 Plan: The code chooses a response based on that reading, written as a rule the student can read.
↓
4 Act: The robot moves. It drives toward the ball, kicks it, or stops, and the loop begins again.
Source of the four-step cycle: Carnegie Mellon Robotics Academy.
What the AI Vision Sensor actually shows students
This is where perception stops being a word and becomes numbers on a screen. The AI Vision Sensor sees the world in a frame that is 320 pixels wide and 240 pixels tall, so the exact centre of its view sits at coordinates 160 by 120. Every object it detects comes with a set of measurements students can read directly in the AI Vision Utility, the live view they use while setting the sensor up. The table below lays out the data the sensor reports and what each field tells a student.
| What the sensor reports | What it tells the student |
| Confidence score | How sure the system is about an AI Classification, shown as a percentage. VEX’s own example reads 99%. |
| Centre and origin coordinates | Where the object sits in the 320 by 240 view. The origin marks the top-left corner of the object. |
| Width and height (pixels) | The size of the object’s box. A barrel reads taller than a sports ball, and a bigger box usually means closer. |
| tagID | For an AprilTag, the actual ID number. For a pre-trained object, its proper name, such as a sports ball. |
| Angle | For colour codes and AprilTags, whether the detection is turned at an angle. |
The data the VEX AI Vision Sensor reports for each detected object, drawn from VEX’s AI Vision Utility documentation.
None of this asks students to take perception on faith. They can hold up a barrel, watch the confidence climb, move it farther away, and watch the width shrink. The reading changes because the input changed, and that cause and effect is the lesson.
Detection types: AI Classifications, AprilTags, and colour signatures
The sensor perceives the world in three ways, and each one gives students a different window into how machines recognize things. Knowing the difference helps a class pick the right tool for a task.
- AI Classifications. Pre-trained objects the sensor already knows, such as sports balls and barrels on VEX AIM. This is the closest thing to “the robot recognizes an object.”
- AprilTags. High-contrast markers that carry an ID number and orientation. Carnegie Mellon calls them QR codes designed for robots, and students can use them as signposts, telling the robot to head for tag 3 and ignore the rest.
- Colour signatures and colour codes. Specific colours, and ordered patterns of colours, that the students themselves configure. This is where a class defines what the sensor should care about.
Why VEX AIM is built to make AI visible
VEX AIM was designed to bring computer science off the screen and into active learning, and its AI Vision Sensor sits at the centre of that. AIM can drive in any direction, kick and catch a ball, show data on its touchscreen, and message a second AIM robot over Bluetooth. Those features matter here for one reason. They give the AI process a body, so perception and decision play out as movement students can see, not as output buried in a console. We carry and support VEX AIM for schools across Canada, and we point teachers to the parts of it that make AI teachable.
Perception at the centre, not bolted on
On many devices, AI is an add-on. On AIM, the AI Vision Sensor is the heart of the robot’s sensing, which is exactly why it works for a lesson about seeing AI think. VEX describes AIM as turning abstract computer science into visible, tangible outcomes, and that is the whole point of the exercise. When perception is central, a student’s attention goes to the reading the sensor produces and the decision that follows, which is where the literacy lives.
Low floor, high ceiling: drive it, then code it
A mixed-skill class does not have to start with code. Students can drive AIM with the One-Stick Controller, then use the on-screen Button Coding to make it move, turn, and kick, all before writing a line of anything. VEX makes a smart point about why this order works: driving the robot to solve a task first gives students a physical model to build their computational model from. From there they move into VEXcode Blocks, and then into Python, on the same platform. The robot grows with the class instead of leaving anyone behind.
How do students see that human decisions shape the AI?
Students see it because the decision is a rule they wrote and can rewrite. The robot is not choosing on its own. It is following instructions a person set, and the code makes those instructions plain. In VEXcode, the perceive-decide-act cycle runs as a short loop: take a snapshot, check whether the object is there, then respond to the data. Once a class can read that loop, the idea that AI is authored by people stops being a talking point and becomes obvious.
The snapshot → check → act loop in VEXcode
Every interaction with the sensor follows the same three beats, and each one maps to a step of the cycle:
- Take a snapshot. This captures what the sensor sees right now and pulls the data from it. VEX is firm that you take a fresh snapshot each time, because an old one leaves the robot acting on stale information. That alone is a good discussion about how AI can be wrong when its data is out of date.
- Check that the object exists. Before acting, the code confirms the last snapshot actually found the target. Skipping this step means the robot might “decide” based on an empty view.
- Respond to the data. Now the rule fires. In VEX’s own examples, the robot drives forward when it detects a blue colour signature and stops when it does not, or it keeps approaching an object until the width reaches a set size, or it turns until a detected object is centred in the view.
One distinction worth flagging before a coding lab. The sensor works with two different kinds of targets, and it helps to name which one a snapshot is looking for. A Colour Signature is student-defined. Your class picks a hue, tunes it, and the sensor tracks anything that matches, so “blue” means whatever blue the students set. An AI Classification is pre-trained. The model already knows objects like the sports ball and barrels, and it returns their proper names along with a confidence score. Colour is a rule students author. Classification is a model they use. When a class knows which one they are coding against, the snapshot behaves the way they expect instead of surprising them mid-lab.
Read that loop aloud with a class and the arrows in “perceive, decide, act” each land on a line of code. Perception is the snapshot. The decision is the rule. The action is the robot moving. There is no hidden step left to imagine.
Change one rule, watch the behaviour change
Here is the moment that makes the whole idea stick. Ask a student to raise the confidence threshold so the robot only acts on very sure readings, or to switch the target colour, or to change the width at which the robot stops approaching. Download the change and run it. The robot behaves differently, right away, in front of everyone. Nobody has to be told that a human shaped the system, because they just watched a human do it. The panel below shows how a single number produces two outcomes.
One rule, two behaviours
| The rule the student sets | What the robot does |
| Act only when confidence is above 50% | Reacts quickly, but sometimes acts on a shaky reading and gets it wrong. |
| Act only when confidence is above 90% | Waits for a sure reading, so it is more reliable but ignores borderline objects. |
An example with two confidence thresholds. Same robot, same scene, one changed rule. The trade-off between speed and certainty is a human choice, and students make it themselves.
How do AI concepts scale as students grow? (The VEX Continuum)
AI does not have to wait for high school. VEX teaches it on a continuum that begins in Kindergarten and carries through senior grades, with one idea threaded all the way through: perception, or how a machine senses and interprets its world. For a curriculum lead planning a multi-year pathway, that thread is what lets each grade build on the last instead of starting over. The table below maps the progression, and the sections beneath it fill in what happens at each stage.
| Stage | Grade band | AI concept students meet |
| VEX 123 | Kindergarten to Grade 2 | How humans and computers sense differently, using the built-in Eye Sensor. |
| VEX GO | Grades 3 to 5 | What data is, how it is collected, and how it drives a decision. |
| VEX AIM | Grades 4 and up | Built-in AI Vision that recognizes pre-trained objects and AprilTag IDs. |
| VEX IQ (2nd gen) | Middle school | Choosing the right sensor, and logging sensor data to a CSV to graph it. |
| VEX EXP and V5 | Senior grades | Comparing how two pre-trained AI models behave in the real world. |
How AI concepts build across the VEX Continuum, from primary sensing to senior-grade model comparison.
Primary grades: VEX 123 and VEX GO (perception and data)
In the earliest grades, the Eye Sensor built into VEX 123 introduces the idea that machines sense the world differently than we do. VEX built its AI literacy activities for these grades with the CS Everyone Center at the University of Florida, and they teach the concept of perception through hands-on tasks, including a Little Red Robot lab where students build a wolf-detecting algorithm. VEX GO takes it a step further into data, with a Data Detectives bridge challenge where students use Eye Sensor readings to spot a crack in a bridge. The message at this age is simple and true: data is something you collect, and it helps you decide.
Middle grades: VEX AIM and VEX IQ (applied AI and data logging)
This is where our anchor robot lives. VEX AIM brings built-in AI Vision to grades 4 and up, sharing what it detects with students through Blocks or Python. Alongside it, VEX IQ has students work open-ended challenges with data from several sensors, which teaches a real skill: choosing the right sensor for the problem in front of you. IQ can also log sensor data to an SD card as a CSV file, so students graph the numbers and see exactly how a sensor perceives its surroundings. Perception becomes a dataset they can study.
Senior grades: VEX EXP and V5 (comparing AI models, real missions)
By senior grades, students move from using one model to comparing two. The AI Vision Sensor on VEX EXP and VEX V5 carries two pre-trained object detection models, so a class can watch how different AI models behave on the same real-world scene, which is a genuine step toward evaluating AI rather than just running it. VEX frames this work inside missions like the EXP Clean Water Mission, where students automate a portable water-treatment plant using sensor data. For classes heading toward career and technical pathways, this is also where AI connects to industrial automation and the kind of work students may do after school, which is the focus of the VEX CTE range.
Will it fit my classroom? Setup, support, and student privacy
The questions we hear most from teachers and buyers are practical: Will it work in a real room? What support comes with it? Is student data safe? Here are straight answers to all three, because a confident yes depends on getting them right.
Classroom-ready: lessons, PD, and a low barrier to start
AIM has a compact design that fits an ordinary classroom, and it arrives classroom-ready with free lessons and activities. VEX provides a free, integrated computer science curriculum for AIM, an introductory course, an API reference, and professional development through its PD+ community. The Carnegie Mellon Robotics Academy also runs educator training specifically on the AI Vision Sensor. When our team sets a school up, this is the support we lean on, so a teacher is never handed a box and left to figure it out.
Student data privacy: vision sensors, not chatbots
This is the point that often settles a decision, and it is worth stating clearly. VEX teaches AI with vision sensors rather than large language models like ChatGPT, and that choice carries real privacy protection. No personally identifiable information is collected. Images or video from the sensor never leave the student device. Students work with pre-trained models, so there is nothing to upload to a cloud server. For a curriculum lead who has to answer to a privacy review, that is a clean answer. The diagram below shows where the data stays.
Where the data stays
1 On the device: The AI Vision Sensor captures an image on the robot itself.
↓
2 On the device: A pre-trained model reads that image locally. Nothing is sent out to be processed.
↓
3 Nothing leaves: No images or video leave the student device, and no personally identifiable information is collected.
Aligning VEX AIM to your program and grade band
For program alignment, we map the lesson to UNESCO’s student competencies, especially the human-centred mindset and the AI techniques and applications areas, since watching a robot perceive and act speaks directly to both. It sits naturally inside computer science and STEM courses, and it extends into career and technical pathways as students grow. We keep the alignment honest and do not invent a match to a specific provincial outcome code. Tell us your grade band and program, and our team will help you map it properly.
A first lesson: making one AI decision visible in 30 minutes
If you want a way to try this that fits inside one period, here is a sequence we use. It needs a VEX AIM robot, one of its objects, and about half an hour. The goal is narrow on purpose: make a single AI decision visible, then change it.
- Drive it first (5 minutes). Let students drive AIM with the controller so they see how it moves and interacts with an object. No code yet.
- Take a snapshot (5 minutes). Point the sensor at a sports ball or barrel and take a snapshot. The robot now has a reading to work from.
- Read the data aloud (10 minutes). Have students call out the confidence score, the object’s position, and the width of its box. Move the object closer and farther and watch the numbers change.
- Change one rule (5 minutes). Set the robot to act only when confidence is high, or to chase only a blue object. Download the change.
- Observe and discuss (5 minutes). Run it, watch the new behaviour, and ask the class what they changed and why the robot responded the way it did.
At the end, students have seen a machine perceive, seen the data behind the perception, and changed the decision themselves. That is AI literacy you can do before the bell.
Bring visible AI learning to your classroom
If this is the kind of AI literacy you want your students doing, we can help you get there. Our team will walk you through the AI Vision Sensor, help you map VEX AIM to your grade band and program goals, and put together a quote that works for your school. We support purchase orders and institutional buying, so the procurement side is straightforward. Reach out to request a quote or book a demonstration, and we will help you take the next step with confidence.
Have a question or want a quote? Our team helps schools, businesses, and makers across Canada find the right setup and get the most out of it.
About the author
Andy Fenos is the President and Founder of iDESIGN 365 Ltd. and iDESIGN Solutions, an educational technology company based in Toronto that has spent more than two decades bringing hands-on coding, robotics, drones, AI, and design into K-12 classrooms across Canada. He also serves as President of the CanSTEAM Foundation, a not-for-profit advancing equitable access to STEAM education, and works closely with partners such as VEX Robotics to connect classroom learning with real-world, competition-based experiences.
Sources and further reading
Frequently asked questions
Do students need internet access for the AI Vision Sensor to work?
No. Because students use pre-trained models that run on the device, the perception work does not depend on a cloud connection, and no images are uploaded for processing. This is worth confirming with your IT team early, since it often removes a network hurdle that slows other AI tools down.
How is this different from teaching AI with a chatbot our students already use?
A chatbot hides its reasoning, so students can only judge the output. A vision sensor exposes the reading, the confidence, and the rule, so students judge the process. If your goal is literacy rather than productivity, the visible process is what builds the skill.
Can one AIM robot serve a full class, or do we need a set?
A single robot works well for teacher-led demonstrations and small-group rotations, which is how many classrooms start. If you want every student coding hands-on at once, a set or the AIM Classroom Bundle fits better. Our team can size this to your class and budget, so ask us before you order.
What happens when the sensor gets a detection wrong?
A wrong or low-confidence detection is one of the most useful teaching moments you will get. It shows students that AI is not certain by default, and it is why the confidence score and the object-exists check exist. Turning a mistake into a lesson about reliability is easier here than with almost any other AI tool.
