Build an AI.
Fly the R101.
In this workshop you’ll build a real AI agent, use it to investigate a historical crash, and then refly the airship with your fix. Here’s how it works:
Build Your Agent
You choose its name, personality, and the rules it follows — exactly how real AI assistants are configured
Investigate
Scan your team’s component, read the evidence, then chat with your AI agent to dig into what went wrong
Discuss as a Group
Share your findings with your teammates, agree on a fix, and write your group’s answer
Fly the R101
Apply your fix and refly the airship — the deeper you investigated, the better your chances of reaching Karachi
For facilitators & researchers
Resnick’s 4 P’s of Creative Learning — Projects, Passion, Peers, Play. Each stage maps to one P. Students construct understanding of AI systems by designing one, not by being taught about one.
Can constructionist design tasks — naming an AI, writing a system prompt, choosing a constraint — scaffold meaningful AI literacy in secondary learners?
Inquiry depth (0–5 turns), system prompt choices, rule specificity, group answer quality, and flight outcome provide naturalistic behavioural data without explicit assessment.
Each personality maps to a distinct system prompt sent to the LLM. Students read their own prompt before it is used — making the hidden layer of AI systems visible and discussable.
Child-centered AI · Resnick’s 4 P’s · Airships XR Workshop, Bedford
Britain’s R101 — the world’s largest airship — departs tonight on its maiden voyage to India. 237 metres long. Built to connect an Empire.
54 people aboard. Hydrogen keeps it aloft. Tonight, you’re at the controls.
Lift
Descend
Incident Report — October 1930
Your Mission
The R101 crashed on its maiden voyage. Three components are suspected. Your team investigates one, builds an AI agent, then reflys the airship with your fix.
Name it, set its personality and rules
Point the camera at the physical model to unlock your evidence
Ask questions, push back, work out what went wrong
Compare findings and agree on a shared answer
Your investigation depth determines if the airship makes it
Three suspected components — your facilitator will assign yours
Hydrogen bags that kept the ship airborne
Five Beardmore diesel engines — speed and control
Full silver-service cabin — added weight nobody accounted for
Stage 1 — Projects
Build Your AI Investigator
You’re writing the system prompt — the hidden instructions that control every response your AI gives. Name it, set its investigative style, then write a rule it must follow. This is exactly how real AI assistants are built.
Give your agent a name before choosing an investigation.
Write a rule. It becomes part of your agent’s system prompt.
Your rule changes how the AI replies — try it and see.
Pick one — the text shown is the actual instruction sent to the AI.
▼
—
—
Never answers directly — always responds with a question.
Select a personality above to preview the full prompt that will be sent to the AI.
Stage 2 — Passion
Scan & Investigate
↓
② Read evidence
→
③ Answer questions
↓
④ Chat
Engine Cabin
Dining Hall
Hold card flat in the box
Hold your laminated card flat in the frame — the AI will identify it.
Hold your laminate card flat in the frame.
Identified — evidence unlocked below
Choose a component to unlock.
Evidence locked
Scan a card above to unlock
Brief your AI
Added
Added — launching investigation
AI brief assembled
0 / 5
Replies are shaped by your system prompt from Stage 1 — not free thinking. Tap Instructions above to see what the AI was told.
Stage 3 — Peers
Group Answer
Talk first, then type. Compare what everyone found.
Group question
In one sentence – what is your strongest redesign idea?
Write your group’s answer before launching.
Your group said
Stage 4 — Play
Fly the R101
You’ve investigated the failure. Now refly the R101 with your fix — does it make it to Karachi this time?
Lift
Descend
(keyboard: ↑ / W / Space)