The Frog and Crane: Folding Metacognition into Learning with GenAI Origami
I've written about how Joon Nak Choi and I are co-teaching a first year undergraduate course at 香港科技大学 before. The course, AI and Society: Ethics, Cognition, and Critical Analysis, is a new common core course with the aim of raising students' metacognitive awareness and critical thinking skills in a world where they will interact and collaborate with smart machines. My last article explored how we might integrate AI into the classroom for enhanced metacognition and self-regulation.
This article describes an in-class activity and showcases some results where we wanted to facilitate metacognitive thinking and self-regulated learning while using GenAI tools. A quick confession: the learning task we designed also had the purpose of creating a fun environment during a time when students are bombarded with midterm exams.
An Experiment in Teaching Metacognition with AI
Last week, we devoted an entire lesson with a workshop to raise students' awareness of their self-regulatory learning and problem-solving skills when using AI. Actually, the workshop had multiple purposes: we wanted students to think about their thinking as they were learning something new, while also thinking about how AI tools might influence their learning strategies, while further developing their understanding of how GenAI tools can help or hinder different learning tasks.
After assigning students to pairs, we gave them the following instructions:
Task 1: The Frog
Learning Origami Objective: Each pair is tasked with creating an origami model of a frog, but you are not provided with any instructions. Instead, you must search for resources within the HKUST library, YouTube, or other non-AI sources to learn how to create the model.
Roles: One member actively searches for resources and attempts to create the origami model based on your findings, while the other observes and notes the problem solver's approach using the metacognitive and self-regulatory checklist.
Process: The problem solver should verbalize their thought process as they search for resources, interpret the information, and attempt to create the origami model. They should discuss how they evaluate the credibility and usefulness of the resources they find. The observer should note how the problem solver manages their time, selects and uses resources, adapts to challenges, and reflects on their learning process.
Unsurprisingly, students were able to create an origami frog rather quickly. Watching YouTube was the main go to tool, enabling many students to thoughtlessly (more on this later) fold the paper into a jumping frog. Once we saw that the most students had a frog, we asked them to go through a checklist and talk about their thinking process and learning strategies. Below are just a few of the questions and answers they discussed:
How were self-regulatory strategies (e.g., persistence, adaptation, stress management) employed in the task?
no stress during the task
She displayed persistence initially as she did not give up at the start. However, as time went on, their enthusiasm waned, and she eventually decided to give up. Their ability to adapt was evident when she switched from searching for images to utilizing YouTube for visual aids. Throughout the process, she experienced considerable stress and struggled to manage the challenges they faced. In terms of overall management, it could be considered average since she did not achieve the desired end result.
What strategies did they use to overcome difficulties?
He switched methods, trying to recall memories, self analyzing where the problem is
Following the wrong the steps and then rewinding the steps to make sure whatever the tutorial is doing is done properly. Fixed the jumping mechanism by fixing the edges and closely repeating what the video does.
Task 2 meets frustration, failure, and learning
For the next step, we gave students a new task to learn, but with a twist: this time they had to learn how to fold the paper into an origami crane, and they could only use GenAI tools to teach them how. Unsurprisingly, students found this task much, much more difficult. In fact, very few were able to learn how to make an origami crane. Below are the instructions given to students:
Task 2: The Crane
Learning Origami with the Assistance of GenAI Objective: Pairs are given a new origami model to create a crane. This time, you can only use GenAI tools (such as ChatGPT or AI-powered tutorials) to find instructions and guidance.
Roles: The roles remain the same, with one member responsible for creating the origami model and the other observing.
Process: The problem solver should interact with the GenAI tool to search for instructions and tips on creating the new origami model. They should verbalize their thought process as they evaluate the AI-generated instructions, integrate the information, and apply it to their folding technique. The observer should note any changes in the problem solver's approach, such as increased efficiency, reliance on AI suggestions, and modifications in their learning strategy.
Below is a screenshot of one pair's outcome as they processed why Task 2 failed:
As with the first task, we asked the students to go through a checklist and talk about their thinking process and learning strategies. This time we asked additional question about how adding GenAI impacted their learning outcomes.
Did the use of GenAI influence the application of these strategies?
Yes. When texts were unable to clearly explain, turned to an image generator to generate visual steps
How were self-regulatory strategies (e.g., persistence, adaptation, stress management) employed in the task?
领英推荐
flexibility: he did not blindly follow rules but think out of the box and search for more effective solutions for higher productivity. stress management: he stayed calmed altho he is MIND FRIED, good stress coping
Did the use of GenAI impact the ability to regulate emotions or motivation?
GenAI give unclear instructions, partner couldn't visualise the steps, he's unmotivated. So he switched to youtube. We did not have motivation to make it after looking at the instructions that GenAI provided.
Given the limitations of LLM chatbots, it's not surprising that most were unable to learn how to make an origami crane. And, while on the surface it appears that they failed a goal, they achieved the learning outcomes of the workshop.
The crane was never the end goal. By giving this new task and requiring them to use GenAI to complete the task, we were forcing them to think: think about why AI was not helping, why its generated output was often useless, and, perhaps more importantly, think about thinking when trying to learn with AI and then explain the outcomes.
We wanted to encourage students to engage in deeper thinking about their learning process, particularly in the context of using AI.
To help facilitate this, we had one final task for everyone. We wanted them to present to their peers via Miro, what happened and why. Here are the instructions we gave students:
Final Report Instructions:
After completing both tasks, discuss your experiences, focusing on the differences between learning with and without the assistance of GenAI. You should reflect on how the AI tool influenced your ability to find and understand instructions, as well as its impact on your problem-solving and learning strategies. Consider the following:
Final Objective: Prepare a visual report on the Miro board summarizing your overall conclusions from Tasks 1 and 2, focusing on your learning processes, the impact of GenAI, and your reflections on the workshop.
Instructions: Create a section titled "Introduction" on your Miro board. Briefly outline the objectives of Tasks 1 and 2 and the main focus of your report. Key Observations: Create a section titled "Key Observations." Use visuals (bullet points, icons, images) to summarize your observations from each task, focusing on strategies and problem-solving approaches. Comparative Analysis: Add a section titled "Comparative Analysis." Use a comparison chart or Venn diagram to visually compare your experiences in Task 1 (without GenAI) and Task 2 (with GenAI). Conclusions: Create a section titled "Conclusions." Draw overall conclusions based on your observations and analysis, answering questions about the impact of GenAI on your learning and problem-solving strategies. Reflections: Add a section titled "Reflections." Share your personal reflections on how the workshop has influenced your understanding of metacognition, self-regulation, and the role of AI in learning. Recommendations: Create a section titled "Recommendations." Provide suggestions for future learners or ideas for effectively integrating GenAI into the learning process. Visual Design: Ensure that your report is visually appealing and easy to follow. Use colors, shapes, and layouts to organize information and emphasize key points. Presentation: Once your report is complete, you will present your findings to the class. Be prepared to discuss your conclusions and answer questions from your peers and the instructor.
The Presentations and Reflections
The outcomes from this exercise were great. One group expressed how relying on YouTube to show them how to create the frog required little thinking, and, as a consequence, little to no deep learning. They were just going through the process to achieve a goal (the frog) without understanding why each folding of the paper worked as it did. Meanwhile, the crane experience forced them (some frustratingly so) to think about strategies for learning how to create the origami bird.
I observed some knowing students understand a text-based chatbot would not suffice and desperately tried to think of other ways to utilize the tools to achieve their goal. I also observed some students discouraged by a chatbot's instructions to fold the paper into what seemed to be infinite triangles.
There was a mix of students being persistent, others giving up on using the tool and trying to figure how to fold an origami crane on their own, while others left the goal entirely and talked about why GenAI was useless for this task. The whole while, they were talking through how they approached the problem. In relation to learning how to utilize GenAI tools for learning and problem, I observed that some students were applying different prompt engineering techniques, like step-step, chain of thought, setting context, or a combination.
A couple of groups used multiple GenAI tools to help them, such as getting one tool to write a prompt for a text-to-image tool. Unfortunately, the image generators are still lacking sophistication and are terrible at creating visual instructions for origami. One day, however, they might be able to do this.
The final Miro presentations were great, and students could see what others did, felt, and accomplished. Below are some examples of what students wrote or produced on Miro.
Concluding Thoughts
The workshop on metacognition and self-regulated learning using GenAI tools provided valuable insights into the complexities of integrating AI into the learning process. While the tasks of folding origami frogs and cranes may have appeared simple on the surface, they served as powerful tools for encouraging students to engage deeply with their own thinking processes and the potential and limitations of AI assistance.
Upon my own reflection, I might do some things differently next time. For example, I'm not certain it was clear to students that the purpose of the experiment was to facilitate their development in metacognition and SRL, while, at the same time, have them think about how interventions like introducing AI as a problem-solving collaborator impacts their thinking process.
I might scaffold the lesson better by referring to Winne and Baker's (2013) definition of learning: “the process of learning is a temporal success of states" where learners:
Knowing this we can ask the following:
While the experiment during our lesson might not answer those questions, it is my hope that it sparks some curiosity in them. I might also give them a task that text-based GenAI tools can help students' learning something new. But the origami was fun, nonetheless.
Deputy Associate Dean (Academic) & Associate Professor of Educational Psychology
5 个月This is super interesting Sean. Thank you for sharing.
Lecturer (Assistant Professor), Chemistry at the University of Wollongong Australia
6 个月I really love this! We don’t have the capacity (I think) to run a similar subject at UoW however there are several elements of this that can be incorporated into teaching-focussed workshops to promote awareness of AI amongst academics.
Building better futures.
6 个月Very cool. Would be neat to introduce a 3rd condition where students are asked to "learn" to fold a model (meaning be able to eventually recreate the origami on their own without referring to any instructions). They could use any combination of tools they like to help them learn the steps and articulate how they get from condition A (production w/o learning) to condition C (skill acquisition).
With so much worry that Gen AI is making us dummer, I thought it was interesting that the students found that the YouTube instructions made the task so easy they didn't learn from it, but because AI was not as useful for this particular task, they were frustrated at having to figure it out on their own. I love that the point was for them to understand how they learn a new task and what role AI may play in that process. Great stuff!
FICPA Scholarship Foundation Board of Trustees
6 个月This is fantastic!