Experiential Versus Instructional Approaches for Eliciting Metacognitive Awareness in AI-Assisted Learning: A Short-Term Longitudinal Study

Imagine you are learning to ride a bike. You could sit in a classroom while a teacher draws a perfect diagram of gears, pedals, and balance on a whiteboard, explaining exactly how the physics works. That is instructional learning: getting knowledge from someone else, step-by-step, without getting your hands dirty. Or, you could hop on a wobbly bike, wobble, fall, get back up, and figure out how to balance by feeling the wind and the road. That is experiential learning: learning by doing, making mistakes, and reflecting on what happened.

Now, imagine that instead of a bike, you are learning how to use a super-smart robot friend called Generative AI (or GenAI) to help you study. This robot can write essays, solve math problems, and explain complex ideas in seconds. But here's the catch: if you just let the robot do all the work, you might stop thinking for yourself. This is where metacognition comes in. Think of metacognition as your brain's "dashboard" or "inner coach." It's the ability to think about how you are thinking. A good inner coach knows which strategies work (like asking the robot to challenge your ideas) and which ones are traps (like letting the robot write your whole homework). The big question educators are asking right now is: When teaching students how to use this powerful AI tool, is it better to give them a lecture about it, or to let them play with it and learn from the experience?

The Experiment: Lecture Hall vs. The Playground

Researchers at Universitat Pompeu Fabra in Barcelona decided to settle this debate with a real-world experiment. They gathered 126 first-year engineering students and split them into two teams. Both teams spent two hours learning about how to use GenAI effectively, but they learned in very different ways.

The first team, the Instructional Group, sat in a traditional lecture. They listened to experts explain the "right" and "wrong" ways to use AI. They heard stories about how using AI too easily can make your brain lazy (a concept called "cognitive offloading," where you outsource your thinking to the machine). They learned the theory, but they didn't actually do anything with the AI during the session.

The second team, the Experiential Group, got their hands dirty. They didn't just listen; they played. They were given four specific challenges. In one, they had to write a summary of an article before asking the AI to check it, forcing them to think first. In another, they had to argue a point and ask the AI to play "Socratic tutor," challenging their arguments with questions. They also tried the "wrong" way—asking the AI for a solution immediately and copying it—to see how it felt to skip the struggle. After each task, they paused to reflect on what happened.

The researchers wanted to see how these two approaches changed the students' "inner coaches" (their metacognitive awareness) right after the session and five weeks later. They measured two main things:

1. Knowledge: Did the students understand what strategies work?
2. Regulation: Could the students actually use those strategies to control their own learning?

The Results: The Fast Start and The Slow Burn

The findings were like watching a sprint race that turned into a marathon.

Right After the Session (The Sprint):
The Experiential Group (the players) got a massive head start. Immediately after their hands-on session, they knew much more about effective AI strategies and felt much more engaged with the tools than the lecture group. Their "inner coaches" were wide awake. They understood the difference between using AI as a partner and using it as a crutch.

The Instructional Group (the listeners), however, showed a mixed picture right away. While they did report feeling slightly more positive about AI and more engaged with the tools after the lecture, their understanding of how to use AI strategies (Metacognitive Knowledge) did not significantly improve. The lecture didn't instantly wake up their inner coaches regarding the specific strategies needed for effective learning.

Five Weeks Later (The Marathon):
Here is where the story gets interesting. Five weeks later, the gap in engagement closed up. The students who had just listened to the lecture had caught up to the players in terms of how engaged they felt with the tools. They had figured out, through their own daily use of AI, that the lecture's advice was actually true. The "lecture group" had slowly learned what the "play group" had learned instantly.

But there was a twist. While the engagement gap closed, the Experiential Group started showing a new, special superpower: Regulation.

Over those five weeks, the students who had done the hands-on exercises didn't just know the strategies; they started using them better and better. Their ability to monitor and adjust their own thinking kept growing. They were getting better at controlling how they used AI. The lecture group, on the other hand, plateaued. They knew the strategies, but they didn't seem to get better at applying them over time.

What This Means for the Future

The paper suggests that if you want students to quickly understand how to use AI, let them play with it. Hands-on activities create an immediate spark of understanding and excitement that lectures just can't match.

However, the paper also warns us not to expect miracles overnight. Just because students know a strategy doesn't mean they will use it perfectly right away. The "play" group showed that the ability to control their learning (regulation) takes time to develop. It's like learning to ride a bike: you might understand the theory of balance immediately after falling off, but actually staying upright while pedaling takes a few weeks of practice.

The researchers suggest that teachers should prioritize these hands-on, "try-it-and-see" activities over long lectures when introducing AI. But they also advise patience. If students don't immediately start using AI perfectly after a fun exercise, it doesn't mean the lesson failed. It might just mean their "inner coaches" are still waking up and learning how to steer the ship. The benefits of doing the work yourself might not show up in the final grade immediately, but they build a stronger, more self-regulated learner in the long run.

In short, you can tell someone how to ride a bike, but you can't teach them to balance without letting them wobble a little first. And sometimes, that wobbling is exactly what helps them stay upright weeks later.