A Lego PBL Unit That Fosters Collaboration

“Mrs. Dailey, it doesn’t know what Minecraft is!” The shocked voice of a fourth-grade boy rang out in surprise and frustration. All around the room, eyes widened as students waited for my response. They were part of an after-school Lego club where I had decided to try an experiment: Could I introduce elementary students to artificial intelligence (AI) in such a way that they would come to see it as a partner in project-based learning (PBL)?

I carefully selected students who had been loath to work with a human partner in years past, the students who would rather finish the entire project alone. After receiving proper consent, I had a group of lone wolves from first through fifth grades. Perfect!

Getting Started

We dipped our toes into the AI world with Brickit, a free app that uses AI to analyze a pile of Lego bricks and generate builds with step-by-step instructions. During this phase, my students came to see AI as a tool to help them in the building process.

My students soon realized, however, that the AI did not understand their questions. Not only did the AI not know what Minecraft was, it did not recognize Harry Potter or the Transformers, either. We quickly moved to an AI art generator at neural.love.

Trying Another Approach

Using neural.love, we worked to engineer good prompts, but we were also designing the framework for our AI-assisted Lego team build.  After a few trials with generating a solid prompt and generated image, my students built close approximations of the vision they had cocreated. We then decided to use the skills we had acquired to do a group project partnering with ChatGPT.

The first stage of our framework was to establish a base vocabulary for the project. Using the questions below, we generated a list of words we felt AI would need to know to understand our prompt.

1. What vocabulary will best communicate our idea?

2. If each word is perceived separately, will the question convey the wrong idea?

3. What vocabulary could be misleading, and why?

After determining that our project would be a cityscape, we added to our vocabulary list. We also started a collected vocabulary list, which included words generated by ChatGPT that the students could reuse, as well as new words they needed to define.

I left our list by the computer, and as the students collaborated, they brought words to me for the list. I would ask them why we needed to collect each of the words, help them define unfamiliar words, and even retire redundant words.

Moving to the Prompting Phase

In the prompting stage of our framework, we began to include ChatGPT as an actual partner by asking it the same questions we were asking each other. Here, we focused heavily on prompt design.

1. Is this prompt too wide/narrow?

2. If so, how can I widen/narrow it?

3. Does it assume the AI has background knowledge?

Eventually, we submitted the prompt, “We would like to build our city with five large sections that are connected. What do you suggest for each section?” ChatGPT gave us a list with a brief description of each area.

Going on to the Next Step

Curating, which is the process of selecting, organizing, presenting, and looking after items in a collection, was the next stage of our framework. To curate our collection of ideas, we needed to dialogue with all of our partners, human and AI.

Each builder drew a design based on their collected ideas. The prompt “What buildings would be in the downtown district?” gave one student a clear picture that he could work with, while another student spent a significant amount of time prompting and re-prompting for the details of realistic train tracks.

When their designs had been finalized, we moved into the creating stage, and afterward we moved fluidly between prompting, curation, and creation toward our final iteration of the build.

Encountering an Obstacle

Suddenly, we ran into a significant problem. Each student was building their section on an entirely different scale. The train tracks were larger than the houses, and the skyscrapers were shorter than the statues. The roads varied in size and configuration. One student noticed these discrepancies and designed this prompt, “How do I make the residential neighborhood to scale?”

ChatGPT’s answer produced more questions, which caught the attention of the other team members. Soon, they were all crowded around the computer with partial builds in their hands comparing and contrasting. Ultimately, they decided on microscale and adjusted their builds to comply.

Collaborating to a Conclusion

During this stage, I watched my individual creators become a team. Prior to this, the students would each gather materials and work alone at different tables in silence. After the scale disaster, they moved their builds to the same table so that they could compare sizes. This led to discussions and adjustments being made, more prompting to ChatGPT, and even co-building. We asked ourselves the following questions:

1. Does the build resemble the original idea, or has it changed?

2. If so, whose input brought the change?

3. Do you like the changes?

4. If not, what can you do to shift the project back on track?

My students confessed that they had all changed their build based on the input of their human and AI partners, and as they moved forward, they were pleased with the changes.

After we reached a consensus, the final stage of our framework was feedback. In this stage, we helped ChatGPT grow as it learned about the outcomes we achieved. After we fitted the completed sections together and made a few final tweaks, we described our finished product to ChatGPT and received its congratulations.

The students were elated with this final bit of dialogue, and they all expressed the idea that their AI partner was valuable and had a large body of knowledge to bring to the table. They concluded that consulting with AI was a starting point for a project, rather than a source for a finished product.

Finally, they reported that their creativity was enhanced by all partners. As a team, they worked with several different types of AI and helped create a replicable framework for partnering with AI. In the end, they completed an innovative project that enhanced their learning, technological, and interpersonal skills.

For me as an educator experimenting with new technology, it’s gratifying to facilitate a group of students coming together to complete a lengthy project. Watching this group of solitary builders solidify into a team working with their peers and a sophisticated artificial intelligence to create a superior product was truly rewarding. I consider this experiment a success, as the group clearly came to regard AI as a partner in their project-based learning, and I am proud of their effort, their final product, and the framework we created together.