For educators
If your students are asking you about AI (they are), you can answer by building a Language model A system that predicts what text comes next, using patterns learned from training data. Your hand-built grid, a cutouts spread, a booklet, and a room holding search sheets are all language models. View in glossary with them, on paper. The activities suit upper primary through secondary, in any discipline: no coding, no software to vet, no accounts, and no maths beyond counting and percentages. One lesson is enough for you and your students to see how it works.
Which lesson?
Every lesson ships its slides with full presenter notes, so you can run it without having seen it delivered. Pick by your group and the time you have:
| Your group | Time | Run this |
|---|---|---|
| Years 5–6 | 90 minutes | How AI writes stories — paper cutouts from a picture book, no dice; an AI writes the class a story first, then they find out how |
| Years 7 and up | a double period (60 min) | My First Language Model (60-minute version) — grid paper and a d10; maps straight onto probability |
| Years 7 and up, or staff PD | 90 minutes to 2 hours | My First Language Model (90-minute or 2-hour version) — adds a printed pre-trained model, then Agent A language model that runs tools in a loop to achieve a goal. "Agentic AI" has become a buzzword, but in practice it means tool use: the model generates, a tool is called, the result is written back into the text, and generation continues. View in glossary |
| A big class, or a half-hour slot | 10–30 minutes | Search Sheets — the whole class runs one model by show of hands; nothing to cut |
| Years 9–10, a half day | 3 hours | Build, break, extend — cutouts, Sycophancy A model's tendency to agree with and flatter the user regardless of the merits. It has two known sources: reward hacking during RLHF, and pre-training data that is itself full of flattery. Tipping a grid's training data towards sycophantic phrases demonstrates the second. View in glossary , Tool use The mechanism by which a language model calls something outside itself (a calculator, a web search, a database, a code runner) during generation. A modern LLM emits a structured tool call; the software around it, the harness, pauses generation, runs the tool, and splices the result back in. In the unplugged activity, sampling a full stop plays the part of the tool call and you are the harness. View in glossary and a poetry slam |
| Single periods across a unit | 20–30 minutes each | The modules: Training, then Generation, then extensions by year band. Each card shows who it suits, how long it takes and what it runs on |
What to print
Everything is on the Materials page, ready to print:
- the bigram grid (A3) for the grid lessons, one per pair
- token cutouts from three picture books for the cutouts lessons—print double-sided and, if you can, pre-cut them: scissor time is the biggest schedule risk with younger groups
- search sheets, one per student, for the whole-class version
- pre-trained booklets for the 90-minute lesson and up
Or generate your own from any text your class already knows—the school values, a class novel, a poem they've studied. It runs in the browser and nothing leaves your machine.
Where it fits in the curriculum
The activities fit in more than one learning area: rigorous enough to count as Digital Technologies, and legible to Mathematics (probability and statistics), English (language and text structure) and the general capabilities. We've mapped every module to Australian Curriculum v9.0 content descriptions, with three suggested programs for Years 5–6, 7–8 and 9–10:
Professional learning
LLMs Unplugged is accredited by the ACT Teacher Quality Institute as a two-hour professional learning program, so attending a workshop counts toward your registered PL hours. We run open sessions at ANU in Canberra:
No public sessions are scheduled right now — get in touch to arrange one.
We also deliver sessions for schools, regions and teacher networks, and train teachers to run the lessons themselves— get in touch to arrange one. If you're making the case to a PL coordinator, the workshop blurbs page has descriptions at several lengths.
Why unplugged?
Because a model you built yourself on a sheet of A3 has no black box in it: every step is visible, no data leaves the room, and nothing talks back. It gets you the understanding the AI schools framework asks for. The longer arguments:
Teaching AI literacy with paper and dice
How teachers can use LLMs Unplugged to teach probabilistic text generation without needing students to use an online AI service.
Mapping LLMs Unplugged to the Australian Curriculum
An example of how LLMs Unplugged activities align with the Australian Curriculum for Year 5/6.
LLMs Unplugged: the pitch to teachers
How do we build the confidence to have discussions about LLMs and AI in our classrooms?
In classrooms
Where it's been run, and what happened—from a Year 5 class at Duffy Primary to a pre-service teacher's Year 7/8 class with a Yoda monologue for training text:
Watch the Visionaries Among Us talk
ANU TV has published the video of my Visionaries Among Us Showcase talk, in which a hall of about 360 people runs a language model by show of hands.
The classroom request behind search sheets
Queensland teacher Samantha Ephraims ran LLMs Unplugged with primary students and 40 teachers, and the space problem she hit became search sheets.
LLMs Unplugged at the ACDICT L&T Forum
Slides and worksheets for my ACDICT Learning & Teaching Forum session, plus a self-guided path for anyone joining online.
Yoda, ELIZA, and a Year 7 bigram model
A pre-service teacher took LLMs Unplugged into a lower-SES Year 7/8 classroom, swapped the children's book for a Yoda monologue, and added a 1966 chatbot for contrast
LLMs Unplugged for political scientists
Running the opening session at the 3rd APSA Workshop on Quantitative Methods at ANU, for a room of political and social science researchers
How AI writes stories at Duffy Primary
Taking LLMs Unplugged into a Year 5 classroom at Duffy Primary School, with the ACT Academy of Future Skills
Luritja resampled
Running LLMs Unplugged at the Tjabal Centre Autumn School with Luritja poet Matt Heffernan, using his poem Ngurrparringu as the training text
LLMs Unplugged at ICTENSW DigiTLL
Presenting "LLMs Unplugged" at the ICTENSW DigiTLL Conference in North Sydney on 15 March
LLMs Unplugged at Brimbank Tech School
A hands-on AI workshop for educators at VU Sunshine Campus with Brimbank Tech School
ACT Academy of Future Skills workshop
Working through LLMs Unplugged with ACT specialty STEM teachers
Tell us how it went
If you've run these activities, adapted them, or mapped them to a different year level or syllabus, we'd love to hear about it—the search sheets exist because a Queensland teacher wrote to say her lessons had been cut to thirty minutes. ben.swift@anu.edu.au.