Introduction

This is the 4th article in our series exploring how human memory works and how AI tools can either enhance or hinder our learning journey. We've seen that true learning isn't about having access to information - it's about being able to understand, retain, and use that information effectively.

In the first article, Learning with AI - Introduction, we explored the three phases of learning: encoding, storage, and retrieval, described the forgetting curve, and discussed how overuse of AI tools can cause cognitive atrophy. While these tools offer access to information and learning support, they also present new challenges.

Our second article, Learning with AI - Encoding, focused on the encoding phase, particularly on elaboration techniques that help us understand and connect with new information. We discovered that over-reliance on AI tools can lead students to skip crucial elaboration steps. When students let AI summarise, explain, or solve problems without engaging deeply with the material themselves, they miss out on the essential mental work that builds strong neural connections during the encoding phase.

In our third article, Learning with AI - Storage, we explored how information moves from working memory to long-term memory and why the timing of our study sessions matters more than their duration. 

👩‍🎓 Quick Recap - Storage
The storage phase is where our brain consolidates information for later use. Successful storage depends not just on how we initially encode information, but also on how we reinforce it over time. Through spaced learning, we create stronger neural pathways that make information more accessible when we need it. This process can't be rushed or automated - it requires deliberate practice and time for our brains to build and strengthen these connections.

Now, in this fourth article, we turn our attention to retrieval. This is a phase when we actually need to access and use the information we've stored. While AI tools promise instant access to information, they can't replace our need to develop strong retrieval skills. Just as a calculator doesn't teach us math, having AI at our fingertips doesn't automatically mean we've learned something. We'll explore how retrieval practice strengthens our learning, and how we can use AI tools to enhance rather than replace this learning stage.

Understanding Retrieval in Learning

Our memory system works in two main ways: recognition and recall.

Recognition is when we can identify information when we see it - like knowing the right answer when we see it in a multiple-choice test.
Recall, on the other hand, is more demanding - it's when we need to bring information to mind without any external cues, like explaining a concept from scratch.

Science says: B)

Recognition is easier because the information is right in front of you, like when you're nodding along while reading your textbook. Recall, on the other hand, requires you to actively retrieve the information from memory, like when you're explaining a concept to someone else without any prompts. When you struggle to recall information during practice, you're actually identifying gaps in your understanding - gaps that you can then address before facing high-stakes situations like exams or real-world applications.

When you are trying to recall information from our brain, you're literally strengthening your memory by pulling information from your mind rather than simply reviewing it.

Let's compare recall with some of the commonly used techniques:

Recalling or rereading? 

A study by Roediger and Karpicke (1) compared two learning approaches: rereading versus retrieval practice. Participants either read a text twice or read it once and then wrote what they remembered. The researchers tested retention at three intervals: 5 minutes, 2 days, and 1 week after learning.

Initially, rereading showed slightly better results when tested after 5 minutes. That's why many learners believe rereading is an efficient learning strategy.

However, after 2 days and even more notably after a week, students who used retrieval practice significantly outperformed those who simply reread the text. 

Recalling or concept mapping? 

In a study by Karpicke and Blunt (2,) the researchers compared three different study techniques: repeated reading, concept mapping, and retrieval practice. They had participants learn about sea otters using one of these methods and tested them a week later.

In the study, participants either:

• Read the text four times
• Read once and created a concept map (a diagram showing relationships between concepts)
• Read once, wrote everything they remembered, then reread and recalled again

Interestingly, before the final test, participants predicted their performance. They expected:

• Best results from repeated reading
• Medium results from concept mapping
• Worst results from retrieval practice

The actual results showed the opposite.

Retrieval practice led to significantly better performance than both repeated reading and concept mapping, despite being perceived as more difficult.

This demonstrates what researchers call a "desirable difficulty" - while retrieval practice feels more challenging, this very challenge leads to better learning outcomes.

After seeing how retrieval practice consistently outperforms other study methods, the question becomes: how can we effectively implement these techniques in our own learning while navigating the AI-enhanced educational landscape?

Retrieval Practice and AI

The key to effective retrieval practice is maintaining active engagement with the material - let's explore specific techniques and how AI can support (rather than replace) your learning process.

Technique: Free Recall (Brain Dump)
Procedure: Take a blank piece of paper and write everything you remember about a topic, as if explaining it to someone who's never heard of it before. Once you've written your explanation, you can use AI to review your work, identifying any misconceptions or suggesting important points you might have missed. Remember: write your explanation first, then use AI for feedback - not the other way around.

Technique: Teaching Others
Procedure: Explain concepts aloud, organising the information in your own words. While traditionally done with study groups or (very patient) pets, AI can serve as an interactive student. After you've explained a concept, ask AI to pose follow-up questions that challenge your understanding or highlight connections you hadn't considered. The key is to formulate your explanation first, then let AI help deepen your understanding through targeted questions.

Technique: Concept Mapping
Procedure: Create diagrams from memory that show relationships between ideas. Draw your concept map first, based on what you recall. Then use AI to suggest additional connections or concepts you might have overlooked. You might ask: "I've created a concept map about (topic) - what key relationships might I have missed?" This way, AI helps expand your understanding rather than replacing your active recall.

Technique: Self-Quizzing
Procedure: Write your own practice questions first - this itself is a valuable form of retrieval practice. After creating your questions, use AI to generate additional ones that target areas where you feel less confident. AI can also help vary the question types (e.g., application, analysis, synthesis) to deepen your understanding. The idea is to start with your own questions before seeking AI-generated ones.

Conclusion

As we've seen throughout this, and previous articles, effective learning isn't about finding shortcuts - it's about engaging our minds in ways that create lasting knowledge. Retrieval practice, while more challenging than passive review or AI-generated summaries, is proven to lead to better long-term retention and understanding.

The challenge isn't whether to use AI in our learning, but how to use it wisely.

When used thoughtfully, AI can enhance our retrieval practice rather than replace it. It can serve as a feedback tool, a question generator, or a practice partner. However, the key to successful learning remains active engagement with the material. Just as a fitness coach can't exercise for you, AI can't do the crucial mental work that builds strong neural connections and lasting knowledge.

Read more in:

1) Roediger, H. L. III, & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249–255. https://doi.org/10.1111/j.1467-9280.2006.01693.x

2) Karpicke, J. D., & Blunt, J. R. (2011). Retrieval practice produces more learning than elaborative studying with concept mapping. Science, 331(6018), 772–775. https://doi.org/10.1126/science.1199327