Artificial Intelligence Curriculum for Primary and Secondary Schools
AI in Computer Science in Elementary Schools
Hoo and Ray
in the Universe of Words
Language Model
Lesson 08
Language Model
Today, Hoo and Ray find themselves in a huge and rather mysterious universe of words. Each word here has its own place and is surrounded by many others that are somehow related. And as it turns out, even ordinary word-balls of string can really tangle up their circuits! Their mission? To help Kitty the cat, who looks very sad, and figure out what might cheer her up. And maybe they’ll discover that what truly matters isn’t just picking the right words – but also how and to whom we give them.
Lesson Overview
The lesson begins with a story in which the robots Hoo and Ray find themselves in a "universe of words" – a vast space where each word has its own place and is connected to other words. Their mission is to find out why their cat Kitty is sad, and they try to solve the mystery using a language model. The teacher leads a discussion about how language models work – that they are trained on large amounts of text and generate responses based on the probability of word combinations. Students play a simple word-completion game, learning how different word choices can lead to different meanings. In a hands-on activity, students create their own "universe of words" by connecting images and words based on semantic similarity, helping them understand how language models organize information. If time allows, the lesson continues with experimentation using a chatbot, where students test how models generate responses and discuss why they sometimes produce nonsensical or inaccurate sentences. The lesson concludes with a reflection, as students share what they’ve learned about language models, their capabilities and limitations, and the importance of critical thinking when working with generative AI.
Lesson Overview
Recommended Age, Lesson Length
Children aged 8-11, 45—90 minutes.
Tools
Teacher: Projecting device, presentation slides, account in an app
that generates text.
Students: Writing supplies, glue, colored paper, printed worksheets.
Building Blocks
Language model.
What Are the Students Learning?
A language model is a program trained on a large number of texts. It can produce texts that sound like they were written by people, but sometimes it generates incorrect content.
Why Are They Learning This?
By understanding how language models work, students can critically evaluate the outputs of generative AI.
How Do We Know They Have Learned It?
In their own words, students describe how a language model works and explain why it might generate incorrect information.
Digital Competence
Facilitating Learners' Digital Competence.
Bloom's Taxonomy
Understanding: Students describe how a language model works
and explain why it might generate incorrect information.
Creating: Build their own simplified model.
Five Big Ideas
4-B-I Commonsense Reasoning.
