The clarification loop: how students use GenAI for deeper learning

Abstract glowing polygonal head background with neurons. Artificial intelligence and network concept. 3D Rendering

As a student at the University of Southampton, I often use AI to build a step-by-step understanding of complex topics. I do not use AI to get an instant solution, and I use it to clarify the confusion that I face while studying. I call it the ‘clarification loop’, as it is a back-and-forth conversation with an AI tool like a personal tutor or study partner. These are the steps that I follow to learn: 

  1. Explain what I already know about the topic. 
  2. Identify and explain what confuses me about the topic. All the module resources, such as lecture notes, can be added here. AI should know the level of complexity based on the module content.  
  3. Ask AI to explain the topic in a way to fill my knowledge gap and build my understanding from the basics when it’s required. 
  4. Respond to the AI answers with follow-up questions to resolve any remaining confusion. At this step, I usually rephrase the complex topic for the AI to make sure that I got all the points correctly about the topic. 

This loop can last for several messages until the AI reaches its limit. AI covers correcting misunderstandings and confirming accurate points, and I (as the student) cover questions and refining explanations.  

Sometimes, when the AI can’t explain everything properly due to its limitations and I still feel confused, I ask for real-world examples to understand deeply. When it’s not enough, I request guidance for self-study. For example, I ask, “What exactly should I search on YouTube to find good tutorials?”. Although AI can’t usually provide the most appropriate link for the related tutorials, it can always provide the best keywords to be searched on YouTube or LinkedIn Learning to find the most suitable tutorials.  

What educators should know   

Many students aren’t just using AI to shortcut learning. We are using it to strengthen understanding through questioning. This method can be more effective if lecturers encourage students to ask critical, follow-up questions rather than one-off AI prompts. Additionally, they can remind students to verify AI answers with credible sources.  

Bite-sized task 

Step 1 – learn 

Read this short blog post by Ethan Mollick. It is from 2023, but is still useful! It discusses uses of AI including as ‘coach’ and ‘tutor’. Read about ‘AI as coach’.  

Step 2 – do 

Open our UoS supported AI tool, Microsoft Copilot. Pick a concept that you recently taught and pretend you are a confused student. Now, follow all the mentioned steps:  

  • First, write what you know about the concept, then state a specific confusion. You can also upload your lecture slides there.  
  • Ask AI to explain the confusion from the beginning. Keep responding with sentences such as, “Do you mean that
?” or “Clarify with an example.” until you feel the concept is deeply clear.  
  • Continue this process until AI clarifies everything.  
  • You can also try asking for self-study guidance and evaluating AI responses. 

Step 3 – reflect 

Now think about the following questions. 

  • How did AI help with clarification? 
  • What worked well? 
  • When did AI give incomplete answers and fail to clarify? 
  • Would you recommend this technique to your students to support their studies? 
  • From the point of view of a University staff member, how would you refine this method to be more useful for the content that you teach? 
  • What kind of sources would you provide to your students to check the credibility of what they learn from AI? 

Contributor biography 

Fatemeh Sargazi is the Deputy Lead Intern in the Centre for Higher Education Practice (CHEP), at the University of Southampton. She has completed her Engineering Foundation year, progressing to the first year of BEng Artificial Intelligence at the Faculty of Engineering and Physical Sciences. She is involved in the University’s GenAI project, focusing on practical and ethical uses of AI in learning.  

© 2025. This work is openly licensed via CC BY-NC-SA