In this activity, students build an AI assistant and investigate which tasks can be effectively delegated to AI. They learn about the benefits and risks of deploying AI and reflect on the societal, ethical, and organisational impact.
- Activity goal
- Assess | Practice skills | Reflect
- When
- In class | Post class
- Where
- Offline | Online
- Duration
- > 60 minutes
- Group size
- Small | Medium | Large
Step-by-step plan
Step 1
Show an example of an AI assistant that you have built yourself. See tips and tricks for an example prompt you can use for your AI assistant. Let students interact with this assistant and explain how you built it yourself: who is the target audience, what instructions did you give the assistant, which documents did you use?
Step 2
Have students think about the purpose and type of user of their assistant. Examples:
- Professional: helps with writing a job application letter.
- Academic: tutor for a course.
- Personal: looks up recipes based on the family's preferences.
Students can also choose their own challenge to learn to critically reflect on which type of problem an AI assistant is valuable for.
Step 3
Have students (in pairs) build an AI assistant. Have them define the purpose, target audience, sources and data used, benefits and risks of use (for example misinformation, bias, and hallucinations), and ethical and social considerations. Also have students think about what they can do to mitigate these risks. Possible reflection questions are:
- Why did I choose this target audience and these tasks for my assistant?
- Which choices in interaction style and instructions were difficult, and why?
- What risks are there? Which risks can I not yet fully mitigate, and how will I deal with them?
- How do my own assumptions or preferences influence the design?
By answering these questions during the design process, students learn to actively think about ethics, usability, and limitations of AI.
Step 4
Organise peer feedback. Have students assess each other's assistant using predefined criteria such as: interaction and user experience, reliability, safety, and risk management. They incorporate this feedback into their design and build in any ethical measures. The peer feedback helps students develop a more robust version of the assistant.
Step 5
Have students test their assistant with different scenarios. Discuss that AI output can vary. Students develop a rubric to assess performance in each scenario, for example on: quality of reasoning, accuracy, tone, style, and limiting risks.
Step 6
Have students document and submit their work: the link to the assistant, the full instructions for the assistant, a reflection on the design choices, the peer feedback, how the peer feedback was incorporated, and the test scenarios and their corresponding results.
Consider the tools and materials mentioned here as suggestions. In many cases it’s possible to use alternative tools. Please turn to the Learning & Innovation team of your faculty first to see which online and offline tools are available and how to apply them.
Offline / online
- For students: laptop with access to (the same) AI platform, (digital) log for design choices, test cases, and reflection.
- For teachers: pre-developed AI assistant as an example, rubric for peer feedback, possibly example prompts or test cases for inspiration.
Tip 1
- Ensure that all students have access to the same GenAI platform (paid or unpaid) for a fair basis.
Tip 2
- Ensure that there is enough time in each phase for students to reflect, for example on the choices made and the risks those choices entail.
Example prompt:
Build an AI assistant for my course [subject]. The assistant must help students with [e.g., practising assignments, explaining theory, feedback on essays]. It must:
- Provide understandable explanations in simple language.
- Offer examples and practice questions.
- Be interactive: ask questions to test understanding.
- Give personal feedback based on the student's answers.
Provide a concrete plan or prototype for how this AI assistant works, including possible dialogues, functions, and how students can interact with it.
