Learning about AI does not require starting with code
Teaching artificial intelligence in secondary school does not mean turning every class into an introduction to programming. Students can start by identifying where AI systems are used, describing what tasks they perform, and discussing how they can fail. Understanding a system and its effects is an educational goal distinct from building one from scratch.
A useful way to organize the topic is around three questions: what information a system receives, what result it produces, and what consequences that result has for people. This avoids reducing AI to machine learning alone or to a list of eye-catching apps. The AI4K12 framework groups related ideas under perception; representation and reasoning; learning; natural interaction; and societal impact. It can serve as a map, not a required curriculum.
Day of AI says its AI literacy activities can be completed without using AI tools. This offers one way to explore AI literacy without using AI tools during the activity.
Observable goals for a group with no prior experience
Before preparing materials, decide what students should be able to demonstrate by the end. In an introductory sequence, the goals can be simple but verifiable: distinguish a rule written by a person from a pattern learned from examples; explain why training data matters; and justify a decision about using a system in a specific context.
Everyday examples—such as classifying images, recommending content, or recognizing speech—can help frame discussion without presenting the result as human intelligence. The learning goal is to explain and question, not to master technical vocabulary or claim that every digital tool uses AI.
A five-minute opening assessment can help gauge the group’s starting point. Ask students to write down a situation where they think AI is involved, what they think goes into the system, and what comes out. Save their answers to compare with a final reflection. There is no need to mark them right or wrong immediately: they reveal prior ideas and help clarify terminology.
A flexible six-session sequence
Day of AI offers formats ranging from a 30- to 60-minute lesson to longer courses, while AI4K12 organizes its framework for different age groups. Use the available class time to choose a pace for the activities below.
Sort everyday examples: In teams, students compare a spam filter, a calculator, and a recommendation system, then explain what makes them think AI is involved. Conclude by distinguishing rule-based automation from systems that identify patterns or generate outputs. The aim is not to force a single answer for every example, but to explain what information would be needed to classify it.
Explore data and patterns: Without computers, hand out cards with fictional examples of objects or animals described by two or three features. One group proposes a classification rule, and another tests it on new cases, including ambiguous examples. Then discuss which examples are missing and how a different selection might change the rule. The activity introduces training, features, and generalization in everyday language.
Compare perception and representation: Show a simple drawing or verbally describe a scene, then ask pairs to identify which features they would use to recognize it. Compare what information is kept and what is lost when a scene is converted into categories. This connects to perception and representation without suggesting that a computer sees or understands in the same way a person does.
Test a rule and examine errors: Give teams new cases to test the rule from the data-and-patterns activity: common, unusual, and difficult-to-classify examples. Have them record correct classifications, errors, and uncertainties, then suggest what additional examples might help. Testing on new cases can reveal limitations in a proposed rule.
Discuss impact and decisions: Each group analyzes a fictional scenario, such as a system that recommends educational resources or helps sort applications. Students identify who benefits, who might be disadvantaged, what data would be reasonable to use, and who should review a disputed decision. AI4K12 includes societal impact among its five core ideas; this activity uses that idea to prompt discussion of specific decisions.
Explain an idea, then revise it: Students create a fact sheet, short presentation, or poster describing a system’s task, the data it would need, one way to test it, and an important limitation. They can then revisit an earlier response and add what has changed. The activity emphasizes reasoned explanation and willingness to revise an idea, rather than familiarity with commercial tools.
Data, bias, and privacy without personal information
An activity about data does not need student profiles, photos of classmates, or personal histories. Use invented examples, teacher-prepared cards, or educational datasets whose terms of use explicitly allow the planned activity. If examples are fictional, say so; this helps ensure no one mistakes a simulation for a real diagnosis or prediction.
To show how an incomplete selection changes results, prepare two groups of cards with different distributions of examples and apply the same rule to both. Ask which categories are better represented, which cases are missing, and who defined the labels. A result can seem consistent while still reflecting the limits of its examples.
Privacy can also be discussed without using online platforms: ask students which data would be necessary for a task and which would be excessive. As a classroom rule, do not enter identifiable information into external services, and do not require students to create accounts for an introductory activity. If a tool is essential, first review its terms, age requirements, and the data-protection rules that apply at the school.
Accessibility and adapting to the classroom context
The activities below include card-based exercises and discussion prompts. Check what materials and access your classroom has when choosing how students will participate; written, drawn, or verbal responses can be used for the activities described here.
Adjust the complexity of examples to the students’ age group and experience. In earlier grades, it is enough to compare examples and explain a rule; in later grades, add questions about representation, evaluation criteria, or the consequences of automating a decision. AI4K12 organizes its ideas for different school levels, while Day of AI offers activities and courses of different lengths. These are references for adaptation, not guarantees that a resource will automatically suit every classroom.
If you use external materials, check the specific resource page for its language, registration requirements, and reuse terms before sharing or adapting it. Describing a collection as free does not establish that every file can be copied, translated, or redistributed without restrictions. When the terms are unclear, you can design your own activity around the concept and cite the framework as a reference without reproducing its materials.
Assess understanding and choose resources carefully
A short rubric can assess four aspects: whether the student describes the system’s task; identifies why data matters; proposes a test using unseen examples; and points out a relevant effect or limitation. The scale might be “not yet,” “partly,” and “with explanation,” with a sentence of feedback. Students are not being assessed on guessing which commercial tool was used.
To wrap up, ask students to answer these questions individually: What claim about AI can you explain now? What data or test would you need before trusting a result? Which decision should remain subject to human review? AI literacy includes understanding uses and impacts, not just producing code.
As starting points, Day of AI offers activities and learning paths for different amounts of teaching time; MIT RAISE describes its work supporting K–12 education; and AI4K12 provides five big ideas and reference materials. Check that each resource is current and appropriate for your students before using it, and do not assume that access or reuse terms apply unless they are stated on the specific resource page.