Artificial intelligence: Use in education

In the TED Talk below, Sal Khan shares the opportunities he sees for students and teachers to collaborate with AI tools – including the potential of a personal AI assistant for every student and every teacher – and demonstrates how he has implemented AI within his Khan Academy.

Read also: Should we use AI for guidance? | Wilfred Rubens

When you, as a student, choose to have AI complete your study assignments and submit them as if you had done them yourself, the use of AI becomes harmful. However, when you employ AI as a tool to support your own learning process, AI can indeed have a positive influence.

Below, you will find various ways in which you can use AI as a personal study coach (Last & Sprakel, 2024; O&T, n.d.). Look at the Prompt engineering tab in this library guide for specific prompts related to this type of AI use and under the Tools tab in this library guide for an overview of specific tools that can be employed.

• To summarise: summarising is an essential task, often in preparation for exams, to organise, describe and remember information. However, writing a good summary requires practice. You can, for example, use the Cornell method (University of Utrecht), a methodology for effective summarising. Add the text to be summarised to your prompt (or upload it in a separate Word document) and make it specific enough - for instance, focused on the Cornell method - to produce an effective summary. For summarising video or audio, AI typically needs the transcript.
• As a translator: AI is increasingly being integrated into language applications. Try out different translation apps and tools and compare them.
• As an aid to revising the material: an effective learning strategy is to test yourself on the things you need to learn. For simple facts and definitions, you can use flashcards, but when it becomes more complex, you need to devise questions yourself. AI can be employed in both creating flashcards and generating revision questions. AI tools can automatically generate flashcards based on the content of a webpage, a section of a book, a scientific article, YouTube video or your own notes. For generating revision questions using AI tools, it's important to provide good prompts. You can use the learning objectives of the course or the book.
• To convert complex content into a clarifying image: for yourself or for your presentation. AI image generation tools can also help design a coherent structure for your presentation and generate appropriate, creative images to elevate the design of your presentation. Use Richard Mayer's principles of multimedia learning, on the one hand as a framework for yourself, and on the other hand to adjust your prompts to improve the image suggestions provided by AI.
• As a writing aid: ask AI for an outline or structure of your text, or for simple or alternative wording, or for feedback on your writing style and grammar. Use the responses as advice in the writing process.
• Upload a PDF export of your presentation and ask an AI tool to generate speaker notes for each slide.
• To find an internship: use AI as a writing aid (see previous point) for your written internship request. Never use personal or company data as input for your prompt. Include tone-of-voice instructions in your prompt. Practice your internship interview with AI. The AI chatbot can take on the role of interviewer. Indicate this in your prompt along with information about your internship position and assignment. Do use fictitious names and company details.
• As an aid to self-reflection: keeping a diary is a powerful method to promote self-reflection. The use of AI in this context can enrich this self-reflection. You can have an AI tool analyse the text from your diary. However, be aware of privacy and confidentiality issues as you often share personal information in a diary. At the very least, use an AI tool that does not use the input data for training purposes.
• As a planner: you can use AI as a virtual assistant (thus using the output as advice) in planning and organising your study work. Have AI create a schedule by including in the prompt all your deadlines, available time and activities. A study plan for an exam is also possible by including in the prompt additional information such as the exam date, your current knowledge of the subject and available study time. By providing information about changes in study work and deadlines, AI can help you adjust your schedule.
• As a motivator: the use of AI can increase your motivation due to the interaction you have with it. However, be aware that the learning curve required to learn how to use an AI tool also takes extra time. Additionally, there is the time needed to critically process the AI's output into your study assignment. Nevertheless, AI can inspire you and help you get back on track if you get stuck.

Teachers can use AI tools as valuable teaching assistants, for example in (Ding et al., n.d.):

• Answering students' questions, for example in the form of an FAQ list (+ answers!).
• Providing tailored education and suitable learning paths for each student.
• Making education more accessible for students who could benefit from extra support.
• Using ChatGPT to summarise or simplify texts.
• Enriching your teaching materials with a storyline about the learning objectives, a number of engaging case studies or worked-out examples.
• Offering students the opportunity to experiment with AI.
• Allowing students to use AI to generate first draft versions of essays, texts, projects, etc.

Points of attention in this regard are that, on the one hand, the information that generative AI tools convincingly present as complete and entirely accurate often is not, and on the other hand, privacy and confidentiality issues may arise. Extensive testing with these considerations in mind and, above all, a strong focus on responsible use and prompts (see the responsible use tab in this library guide) are therefore essential! The step-by-step plan below can serve as a guideline if you, as a teacher, want to start using GenAI in your teaching:

University of Utrecht

A recent framework titled 'The AI Assessment Scale' distinguishes five levels of AI use in assignments (Last & Sprakel, 2024):

You can assign one of these five levels to each assignment (or part of it), making it clear to students to what extent a GenAI tool is allowed. More information about the AI Assessment Scale can be found in this blog by one of its creators and the accompanying article.

Below are specific ways in which AI can be used in teaching (Last & Sprakel, 2024). Check the Prompt engineering tab in this library guide for specific prompts related to this type of AI use and under the Tools tab in this library guide for an overview of specific tools that can be used.

• When designing a lecture, you must consider various factors such as your intended learning outcomes, student needs, available resources, teaching methods, and the learning environment. GenAI can facilitate this process by providing you with a basic structure that you can then adapt to your own preferences. Ask the tool to generate multiple proposals for structures, allowing you to vary and combine. Pick out the ideas that appeal to you, and continuously refine. If you are accustomed to working with templates, first feed these to the tool. This way, the tool learns your way of working and can then display the output in the desired format.
• Ask an AI tool for ideas for an activating assignment that promotes your students' engagement and understanding, such as a discussion task, brainstorming exercise, or icebreaker activity. Don't ask for just one idea, but have it develop 50 straight away. Let yourself be inspired! When a particular idea for an assignment appeals to you, have the AI tool help you develop this activating assignment in more detail. Include principles that you consider important, such as the classroom layout, the level of difficulty, or how much preparation the assignment might require.
• AI can assist in implementing direct instruction by generating detailed steps and examples of modelling, which you can use to refine your instruction. It may sometimes be necessary to first 'feed' the language model the steps as an example, so the tool understands which steps you want to go through. Ask the tool to explain the suggestions for each step in more detail, or to make new suggestions. Engage in dialogue and refine your output. You can also use AI to differentiate your instruction. There are different variants of the direct instruction model. Choose the variant that best fits your context. Before using methodologies, align with your language model what you specifically mean by them.
• AI can help transform complex content into a clarifying image: for yourself or for your presentation. AI tools for image generation can also help design a coherent structure for your presentation and generate appropriate, creative images to elevate the design of your presentation. Use Richard Mayer's principles of multimedia learning, on the one hand as a framework for yourself, and on the other hand to adjust your prompts to improve the image suggestions provided by AI.
• Harness the power of AI to create efficient knowledge clips, from working out your preparation, writing a script, devising the b-roll (supplementary images or video that can be used to clarify and reinforce the message), to analysing the quality based on Mayer's multimedia principles. For scripts, it can be useful to work with chunking, dividing long texts into smaller pieces. For example, first work out the structure, and then ask the AI tool to develop a script for each section. Ask the AI tool to develop ideas for the b-roll, purely based on your preparation. Let the system be very creative, so you get unexpected results. Ask the AI tool to analyse a particular elaboration using Mayer's multimedia principles.
• When using an AI tool to generate examples or case studies, you can adjust the input parameters to the complexity and relevance. For instance, ask the tool for simpler examples or additional case studies. AI can also help generate images, graphs, diagrams, or other visual representations that can be used as visual aids for the generated examples and case studies. Use AI to explain concepts in different contexts, such as practical application in daily life or professional practice.

The rise of GenAI compels us to evaluate and adapt our approaches to writing assignments. Use the following strategies for this:

• Establish clear guidelines, with AI in mind.
• Redesign writing assignments: have students make connections between knowledge and their own experiences. Or have them apply concepts to very specific situations.
• Integrate AI tools into writing assignments.

By introducing students to AI in their studies, they not only learn to use AI effectively but also develop the ethical and philosophical insights necessary to deploy these technologies responsibly. This provides them with the essential skills and competencies to navigate an increasingly technology-driven society, ensuring that higher education remains not only current but also guiding future developments. We share activating teaching methods below for inspiration.

Teaching Method: Reverse Engineering

Provide students with a text or an image, either authentic or generated by GenAI, and then ask them to create a prompt that produces a result as similar as possible to the original (Last & Sprakel, 2024). Ensure that students have access to an AI tool to test their prompts, allowing them to continually revise and improve them. This reverse thinking approach encourages students to closely examine the various aspects and specific characteristics of the text or image, and how to refine them to achieve a particular output.

Through this playful process, students learn how GenAI generates output and what constitutes an effective prompt. They also come to understand that the quality of GenAI results is directly proportional to the input it receives. This exercise enables students to better utilise AI tools themselves as they enter the professional world.

Teaching Method: AI-Assisted Introduction Writing

Have students write an introduction on a topic using GenAI. Provide specific instructions on the use of the AI tool. For example: You may only adjust the prompt twice. This encourages students to think carefully about the modifications they wish to make. After receiving the AI tool's output, the student may further refine it themselves. This teaches students not only to adopt AI output verbatim, but to view it as input upon which they can expand. The final version of the introduction is then discussed. Is this introduction usable? Have students choose a topic for the introduction about which they are knowledgeable. This enables them to effectively evaluate the AI output and stimulates their critical perspective.

Here and here you can find more ideas for AI-based teaching methods. You can also find open educational resources involving AI on the Edusources platform (scroll down or click on this link for the latest publications):

Do you have any interesting AI-based teaching methods of your own? Let us know! You can find contact options on this page. We'd be delighted to add them to this list. Alternatively, you can add them yourself as open educational resources to Edusources and include the keyword AI. Your teaching method will then automatically appear in the results list above.

The possibilities offered by generative AI call for a more robust reflection on the role of assessment in various degree programmes and a shift from (summative) control towards (formative) development of students. When assessment more closely mirrors the demands of students' professional practice, two key questions emerge:

1. Is AI an integral part of that professional practice (legal, financial, or policy practices)?,
2. What is then required of these professionals: specialist knowledge to "evaluate the merit" of AI-generated output, or other types of professional activities that are independent of AI output?

To make assessments AI-proof, a critical examination of learning objectives is necessary. Learning outcomes can be personalised in the execution of studies, tasks, and assignments in such a way that the student's motivation is to conduct their own research and focus on their personal development. This will require a different perspective on the content and coherence of learning objectives or learning outcomes in the degree programme profiles.

Focus on narrative feedback and development in the process. Explicitly inquire about how AI was used and what was learned from it.

Below are several ways in which AI can be employed in assessment, evaluation, and feedback (Last & Sprakel, 2024).

• Use language models to transform existing assessments or let them inspire you with possibilities for alternative, authentic assessment forms or the design of new assessment methods that place greater emphasis on deep learning and applying knowledge in practice. Support alternative assessment forms with clear rubrics or success criteria.
• Provide an AI tool with a student's writing assignment along with a prompt focused on assessing spelling, grammar, and structure or other parameters such as the use of prepositions, agreement, rubrics, etc. Have it write out improvement points as constructive feedback. Ensure random checks of the AI-reviewed texts. Use the AI as a second marker, and compare the results with your own. This still saves considerable time while ensuring 'the human in the loop'.
• Use a language model to create detailed and growth-oriented rubrics for feedback and assessment, and use these as instruments for evaluation and self-assessment. If you already know which criteria you want to evaluate, include these directly in the initial prompt. This also applies to the scale, weighting, and assessment method. Refine iteratively until you are satisfied. Apply a created rubric to a written assignment by inputting it as data, and then asking for an assessment based on the rubric. Have the language model write out narrative feedback based on an assessment, and send it to students. Also mention the steps they can take for improvement. Be transparent if you use a language model as a tool. Share your process and prompts with students and encourage them to do the same.
• Make clever use of a language model to help you draft a concept evaluation plan to measure the effectiveness of the impact of your teaching, so you can see what works and what can be improved. Test the evaluation plan with the help of dummy data. This allows you to refine your approach and anticipate certain challenges. Ask the tool to generate dummy data for your evaluation plan, and then use this as input for your test.

Automated feedback and assessment appear to be an attractive application of GenAI in education. It offers the potential for faster, more personalised, unique, and perhaps more objective feedback. For teachers, automated feedback and assessment could mean significant time savings. Additionally, feedback generated by GenAI can contribute to more learning path-independent education and a richer dataset for learning analytics. However, GenAI can produce inaccurate and sometimes simply erroneous output when providing feedback or assessment. Moreover, the GDPR contains a prohibition on "solely automated processing-based decisions that have legal or otherwise significant consequences for the data subject". Automated assessment in education undoubtedly has significant consequences for students (Ding et al., n.d.)!

A balanced approach can lead to greater effectiveness (Last & Sprakel, 2024). Use AI tools to provide rapid and objective feedback on lower-order aspects of writing (such as spelling, grammar, structure), whilst you, as the teachers, concentrate on giving qualitative feedback on higher-order aspects (such as argumentation, reasoning, and creativity).

AI can be employed not only in the delivery of education and the development of educational materials but also in the development of education itself (Last & Sprakel, 2024).

• The process of vision development and concretisation begins with exploring and understanding the educational philosophy underlying the educational design. Subsequently, the vision is formulated and further elaborated into learning and design principles. Concretisation then takes place. Useful methods for this include brainstorming sessions, focus groups, and the use of design techniques such as design thinking. It is important to test and adjust the vision based on feedback from stakeholders, such as teachers, students, and educational experts. Cleverly employ a language model to enrich this process with new ideas, concept texts, or feedback on existing texts. Use AI to analyse insights and identify patterns that can contribute to strengthening or concretising your vision of education and learning. Generate alternative scenarios or future images of your educational practice and investigate, with the help of AI, how these can influence the vision of education.
• Use AI to analyse and improve the coherence between intended learning outcomes, activities, and assessment to promote constructive alignment in your educational design. For example, you can ask an AI tool to analyse the constructive alignment between different subjects or courses, with suggestions for improvement. Have the tool elaborate this in a spreadsheet or table. Make use of existing educational models and frameworks to further refine the constructive alignment. It is important, however, to provide the chatbot with examples of good and poor output and how it should apply the chosen model.
• It is quite a task for a teacher to formulate good learning outcomes, but fortunately, you can support this process with the help of GenAI. Note: it takes some practice to arrive at well-described learning outcomes. When working with a chatbot, it is advisable to refine your learning outcomes iteratively. When formulating multiple learning outcomes, work with a template. This way, you can more easily have each learning outcome generated by a chatbot, having it fill in the same fields each time. When you work with a certain fixed language in your learning outcomes, such as always starting with 'Students are able to', include this as a parameter in your prompt. Link your own framework with the AI tool by feeding it to the tool and indicating that the tool should always apply this framework before elaborating learning outcomes. You can also give the AI tool examples of what you consider good learning outcomes, and what not.
• AI tools can provide ideas or a concept structure for a lesson plan or subject/course structure. If you have a certain fixed format for a lesson plan or work with a specific methodology for designing your education, provide this as additional input or a step-by-step plan for the tool to follow. Language models can sometimes make mistakes when it comes to time notations. Pay close attention to this, and do not blindly adopt the output. Work through iterations. Be very specific in your instructions, and always indicate exactly in the tool how you want the output to change.

How to Achieve AI-Ready Education: A Guide.
(Aarts & Themmen, 2024)

1. Analyse the learning outcome/learning activity​
   1. Analysis of the assessment/learning outcome
   2. Analysis of the learning activity
   3. Estimating the impact of technology​
   4. Supporting usage​
   5. Adjusting the assessment/learning outcome
   6. Adjusting the learning activity
2. What is Intended to be Learned from this Learning Outcome/Learning Activity? This can be explicitly described, unspoken, or even unconscious.​ 
   1. Short-term (learning outcome/goal)
   2. Long-term (professional performance/competencies/national standards)
3. How do this goal & activity align with your team's core values and educational vision?

Various aspects that should be addressed within the institution so that an educational team can make their teaching AI-ready (Harmsen, 2024):

• Vision and strategy
• Policy (ethical, legal, educational)
• Culture
• Competence (knowledge, skills & attitude)
• Governance
• Technology
• Other

Teachers can find ideas and inspiration at aiforteachers.com.

Additionally, there is the English-language website AI for Education with free trainings and downloads for use in education.

If you, as a teacher, wish to delve deeper into the responsible use of GenAI in education and its potential impact on your own teaching, you'll find two comprehensive, free online training courses below. These are offered by other institutions but are also accessible to external participants. Although the KU Leuven training also indicates students as a target group, the content and scope of the training make it more suitable for teachers.

• The e-module "Impact of GenAI on Higher Education" from the network of UvA Teaching & Learning Centres. This English-language e-learning consists of six parts: 1. What is GenAI and how does it work? 2. What are the limitations and ethical issues of GenAI? 3. What is the impact of GenAI on teaching and assessment? 4. How can you help your students use GenAI responsibly? 5. How do you formulate effective prompts? 6. Which GenAI tools can be useful in higher education? Estimated total completion time: 2 hours.
• The online training "AI in Education" by the Itec research group at the Catholic University of Leuven. This English-language e-learning consists of 4 parts: 1. The relationship between AI and education 2. The black box of AI: all about data 3. The challenges of AI in education 4. Navigating AI in education. Estimated total completion time: 5 to 7 hours.

• Personal assistant
• Study help
• Teaching
• Teaching methods
• Assessment and Feedback
• Developing Education
• AI-ready education
• Inspiration and Deepening