Vocational Innovation – Teaching
Interactive learning material · M3-T1-MLC2
Check, interpret, move on – AI suggests, the teacher decides
Learning objective
Performance assessment is one of the most sensitive areas of teaching and learning. Teachers, parents and students have been debating its purpose and methods for decades.
This material shows how AI can help teachers plan assessment and identify common mistakes. It also shows how to choose next steps that match learners’ different starting points. Practical examples show how generative AI can help teachers improve assessment.
Basics · Learning outcomes
Modern teaching focuses on assessment linked to learning outcomes. Learning outcomes, or learning goals, describe what learners should know and be able to do at the end of a learning period. They include the skills needed to use this knowledge to solve problems and complete tasks. Assessment must follow these outcomes, which are set in advance. The outcomes must be observable and measurable.
A learning outcome for a trainee electrician: “The learner can independently build and check a simple lighting circuit using a circuit diagram. They follow workplace safety rules and can explain how the circuit works.”
The next question, task or learning path depends on the learner’s current result. Experienced vocational teachers do this naturally. For example, they watch a learner wire a socket. They then decide whether to explain again, set a new task or let the learner continue.
AI can help teachers do this with a whole class. It quickly analyses responses, identifies patterns and suggests different next steps.
Basics · Summative and formative assessment
In vocational education, assessment usually takes place at the end of a learning period. Tests and grades come at the end of a topic, semester or school year.
Summative assessment mainly shows the learner’s level of knowledge at that point. A grade alone gives little guidance on what to improve or how to make progress.
Learning signals · Snapshots
Diagnostic checks, response summaries, self-assessment and progress indicators provide evidence that AI can help teachers interpret.
None of these signals is a grade. They are snapshots that help answer three questions: Who is ready to move on? Who needs more support? Which misunderstanding is common enough to discuss with the whole group?
Ask a few well-planned questions at the start of a topic. During a lesson, use a “hinge question”. At the end, use an exit ticket. Good questions include diagnostic distractors: wrong answer options that represent common mistakes. The option a learner chooses helps explain why they got the answer wrong.
Use summaries showing how many learners chose each answer. Include the most common wrong answers and short written explanations. Replace names with codes, such as Learner A or T07. Group patterns often tell us more than individual scores.
Confidence ratings (“I am sure / I am not sure”), traffic lights or “I can…” statements. Compare confidence with correct and incorrect answers. A confident learner who makes a mistake needs different support from a learner who is unsure but answers correctly.
Completed workshop tasks, number of attempts, items ticked off on a practical checklist, or progress in an online module.
Learning signals · Differentiated next steps
AI can help us create test questions at different levels of difficulty for a topic. It can also create short practical tasks and problem-solving situations. We can then use the results to give learners different tasks that match their level of knowledge and skills.
Quiz tools such as Microsoft Forms, Google Forms, Moodle and Kahoot can export results to a spreadsheet. AI assistants such as Copilot, ChatGPT, Gemini and Claude can summarise the results and group similar written explanations. They can link common wrong answers to likely misconceptions and highlight wrong answers marked “sure”. Use a school-approved account where possible.
Once the teacher sees the pattern, AI can suggest different next steps: questions that target a specific mistake, a worked example or a practical task. One path explains the concept in a new way. Another provides practice and helps learners strengthen their knowledge. The third increases the difficulty of the task.
Try it out
Choose a path to see what learners could receive after the check in the next example on series and parallel circuits.
Revisit
Learners who think that “the first lamp uses up the current” get a measurement task: they use a multimeter to measure the current before and after each lamp, then explain the result.
Practise
Learners who think that resistances in parallel add up practise calculating equivalent resistance with targeted questions and a worked example.
Extend
Learners who answered almost everything correctly continue with a mixed circuit and justify the steps of their calculation.
All three paths lead to the same learning goal – after the next check, learners can switch to another path.
Worked example · Trainee electricians
A teacher gives 24 second-year trainee electricians a six-question check on series and parallel circuits in Microsoft Forms. Learners use codes instead of names. After each question, they mark how sure they are of their answer. The teacher exports the results and deletes the column linking codes to names. The teacher then shares the data with the AI assistant.
Example prompt: interpreting a diagnostic check
I teach second-year trainee electricians. The table gives an anonymised summary of a six-question diagnostic check on series and parallel circuits (learners T01–T24).
The learning goal: “calculate current, voltage and resistance in simple series and parallel circuits and explain the result”.
[pasted table]
Worked example · Interpretation and decision
Chose “the first lamp uses up the current”. AI suggests a measurement task: measuring the current before and after each lamp.
Thought that resistances in parallel add up – most of them marked “sure”. AI suggests targeted questions about equivalent resistance.
Answered almost everything correctly. For them, AI suggests a mixed-circuit task.
Feedback · The adaptive assessment cycle
Adaptive assessment follows a cycle: check, interpret, adapt, learn, then check again. Effective feedback answers three questions:
Where am I going? Where am I now? What is the next step?
AI can help write feedback and suggest next steps, but learners need to understand this feedback.
Show the class an anonymous summary, for example: “Many of you think the current gets used up. Let’s test it!” Let learners compare their self-assessment with the results. Where possible, let them choose their own next learning path. This strengthens self-regulation, the ability to manage their own learning, which is also essential for lifelong learning alongside work.
Feedback · Assessment criteria
Assessment based on learning outcomes needs clear criteria set in advance. AI can then help create a rubric that describes what performance looks like at each level.
Example – Catering
The teacher describes the expected learning outcomes and asks AI to create a four-level rubric. The criteria cover ingredient preparation, correct use of cooking methods, food safety, time management and presentation. The levels for one criterion:
| Level | Correct use of cooking methods |
|---|---|
| Excellent | Follows the required cooking method correctly and independently. |
| Satisfactory | Makes minor mistakes that do not affect the final result. |
| Needs improvement | Needs the teacher’s help several times to follow the steps of the cooking method. |
| Unsatisfactory | Makes basic mistakes in the cooking method and cannot complete the task correctly. |
Example – Mechanical engineering
At the start of a CNC programming module, the teacher uses AI to create a short diagnostic check on coordinate systems, reading technical drawings and basic machining operations. AI can identify recurring errors in the answers and suggest tasks for extra support and more advanced practice.
Feedback · Vocational examples
One of the most promising uses of AI is supporting formative assessment. As teachers, we provide the assessment criteria and our observations of the learner’s work. We then ask AI to write short, clear and constructive feedback.
Vocational teachers often need to assess complex work processes. AI can help organise assessment records, summarise the teacher’s observations and identify patterns in learners’ progress.
Example – IT
Learners create a simple web application. The teacher can use AI to analyse the code and identify possible errors. AI can also suggest improvements that match the learner’s knowledge level. Rather than asking AI to fix the program, the teacher can ask:
“Do not give the complete solution. Write three guiding questions to help the learner find the error in the program on their own.”
Example – Construction
Trainee bricklayers are building a wall. The teacher notes how accurate the measurements are and whether the wall is vertical. The teacher also records how learners use materials and organise their work, and takes photos.
AI can use these notes and photos to summarise each learner’s strengths and areas for improvement.
Feedback · Self-assessment and peer assessment
Assessment supports learning best when learners can identify the strengths and weaknesses of their own work. AI can help by asking questions, providing checklists and discussing the work with learners.
Example – Commerce and marketing
Learners design an advertising campaign for a local business. AI uses agreed assessment criteria to ask questions: How well does the message match the target audience? What data supports the chosen solution? Which alternatives did you consider?
Learners first assess their own work. They then compare their views with AI’s suggestions and feedback from their peers.
A prompt for any vocational field
I am a vocational teacher. Create an assessment rubric for the following learning outcome: [Learning outcome]
The learners’ task: [Description of the practical task]
Create a four-level rubric with 4–5 assessment criteria that can be observed. Describe each level in specific terms that a vocational teacher can check. List critical safety requirements separately. Then suggest three feedback questions to help learners assess their own work. Do not invent requirements from laws or standards. Mark anything that is uncertain.
Fill in the parts in square brackets to match the specific task.
Risks · Teacher responsibility
Using AI for assessment requires care. Language models may misunderstand a learner’s answer. They may reach incorrect technical conclusions or sound too confident in their assessment.
Even small changes in a task’s wording or the information provided may lead to different assessments of the same performance.
Use AI to plan learning. Do not use it to grade or rank learners, or decide whether they can move on.
AI may make mistakes, find patterns that are not there or suggest tasks with technical errors. Check the technical content of every question, especially any workplace safety rules.
Watch for suggestions that lower expectations for some learners. Differentiation changes the learning path and the support, but keeps the same goal.
Enter only the data you need. Replace names with codes and keep the code key separately – pseudonymised data is still personal data.
AI systems intended to evaluate learning outcomes are classified as high-risk (Regulation (EU) 2024/1689, Annex III, point 3).
Tell learners what data you collect and how you use AI. Make clear that AI never makes the decision.
Takeaways · Trust and assessment for learning
Effective learning depends on trust, mutual respect and a shared learning goal between teachers and learners. When cheating is suspected, learners may feel they must prove their innocence with every piece of work they submit. Constant suspicion can easily damage this trust. It also weakens learners’ motivation, which is already becoming harder to sustain.
Because AI can generate content, essays and take-home exams are more vulnerable to misuse. Projects, portfolios, oral presentations and interactive simulations let learners show their understanding by applying what they know.
AI can help develop assessment criteria linked to learning outcomes, support personalised feedback and make learners’ progress easier to track.
The goal is to strengthen assessment for learning, rather than automate assessment.
AI can free up time for teachers to focus more on each learner’s development. Teachers still make the professional and educational decisions.
Practice · 45 minutes · Individual, then small-group work
Learners complete a short diagnostic check, compare their self-assessment with the correct result and choose a learning route. The aim is to involve learners in using assessment results to plan their own next step.
Materials
A five-question check with confidence rating in a quiz tool (learners use codes); three route cards prepared in advance with the help of AI (revisit, practise, extend); projector.
Takeaway
Assessment results are information for the next step, not a verdict. Comparing confidence with correctness strengthens self-assessment.
Self-check quiz
Each question has one correct answer. Answer all six questions, then check your result and the explanations.
Sources · Related materials
The Disturbing Secret Behind AI in Education: How It’s Crushing Student Trust
The illustrations come from the images supplied with the learning material.
VITA · Contributors and attachments
This work may be reused and adapted under the terms of the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0).
Funded by the European Union. Views and opinions expressed are those of the author(s) only and do not necessarily reflect those of the European Union or the National Agency.