#43. When the Wi-Fi Dies Mid-Lesson: Graceful Degradation

The moment every teacher using EdTech dreads...

It is Monday afternoon. My gifted Grade 6 extra-curricular activities group is gathered for Lesson 2 of our New Horizon Island Project — a browser-based AI simulation world where students interact with twelve AI "citizens" to solve a mystery across the island. Previously,  Lesson 1 has done its job: the students now understand the world, know a  little about the AI citizens, and have been introduced to how to communicate with them. Lesson 2 is the payoff — the first live, whole-class demo of messaging the AI citizens: how to phrase questions, how the chat works, how to get useful answers out of a character who only knows what the island knows. I am excited. My students are intrigued. Excellent!

About ten minutes later, the school wi-fi goes down.

Let's move from the inner teacher rage of the moment, the sarcastic mental comment to one's self, "Well, it was all working fine when I tested it before the lesson, of course it was, bloody typical..." to a past tense reflection from distance. 

The Internet in the classroom briefly went down. Then it came back. Then it went down again. Each time, we watched the spinning progress wheel, and we waited, and we lost the thread, and we reconnected — and then we lost the connection again. Anyone who has taught with technology has lived some version of this moment. The question is never if it will happen. It's when. This raises the question: Can your lesson survive it?

Mine did. And the reason comes down to a concept borrowed from engineering that I think every educator using technology should steal: graceful degradation.

__________________________________________

What graceful degradation actually means

The term comes from systems engineering and fault-tolerant design. A gracefully degrading system is one that, when parts of it fail, doesn't crash completely — it drops to a simpler mode of operation and keeps functioning, albeit with reduced capability. Think of a phone that switches to emergency calls only when it loses signal. The fancy features are gone, but the core purpose survives.

The key insight for education is this: degradation is inevitable, but gracefulness is a design choice. The wi-fi failing isn't the failure of the lesson. The lesson failing is the failure of the lesson. Those are two separate things, and only one of them is within your control.

Building the island twice

Some context here, because the most important part of this story happened before the students ever walked in.

New Horizon Island runs across multiple sessions. The students aren't just chatting with chatbots — they're collecting thirty-two clues scattered across four sectors of the island, questioning AI citizens, and slowly uncovering who built the island and why. It runs on a live backend with an AI model doing the talking, over school wi-fi, on a class set of iPads. Every one of those is a potential failure point, and I knew that before I built a thing.

So when I designed the project, I built the island twice. Once digitally — the platform the students will use from Lesson 3 onwards, logging in with the class code and directing their own questions to the AI citizens. And once physically: a set of printed cards. Not one for each AI citizen, carrying the exact same content the digital citizens hold, but an 'emergency' set for when the wi-fi goes down. Amongst these physical, laminated small cards are some of the information, but also some 'fake news' cards. The cards are the analogue shadow of the digital world created, but with a twist. Same characters, same clues, same communicative purpose, broadly speaking, but with some red-herring cards thrown in. When the school wi-fi goes down, I have a back up activity— just running on paper and student talk instead of servers and screens.

This matters for a simple reason: in this project, talking to the citizens is the learning objective. The platform is one delivery method for that objective. The cards are another. Once that's true, a network failure stops being an existential threat to the lesson and becomes a change of delivery mode.

Why this lesson should have collapsed

Lesson 2 had one critical job: show the students, as a whole class, what successful communication with an AI citizen looks like before they do it independently from Lesson 3 onwards. I would model the chat on the big screen, the class would suggest questions, and we'd see how the citizens responded. That job depended entirely on a live connection to a live AI.

When the wi-fi started dying, the tool was unusable. The lesson's core activity was unreachable. By the usual standard, that's a dead lesson.

What graceful degradation looked like in practice

Here's the thing: the cards weren't a fire extinguisher behind glass that I smashed in a panic. Switching to them was always an option on the lesson plan. When the connection dropped for the third time and the reboot cycles started eating the lesson time, we put the iPads down and dealt out the cards.

Each small group got a card, and the task inverted beautifully: instead of me showing them how to question a citizen, they had to decide what they would ask the character in front of them once they got back online. Then we asked students to reason about what the citizen might answer, based only on what little they already knew about the island. The modelling lesson activity became an interrogation-planning lesson. Same skill, same content, different medium.

What I couldn't fully plan for was the atmosphere: watching my thirteen students watching a loading progress wheel, waiting for a network that kept betraying them. Before we moved to the physical cards, we speculated and we folded the failure into the fiction. Instead of pretending the delay wasn't happening, we made predictions about why the AI citizens were taking so long to respond. Maybe there was an emergency on the island? Maybe an incident in one of the sectors? Maybe the citizens were busy dealing with something we didn't know about yet?

Think about what that did. The students were still using exactly the skills the project targets, namely hypothesizing, reasoning, and speculating in English, while waiting for the technology to catch up. The downtime stopped being dead time and became part of the mystery. Nobody was off-task, because the task absorbed the failure.

And the next day, the IT department announced the whole school's internet would be down for some time. Which was the universe confirming what the previous afternoon had already demonstrated: this wasn't just my bad luck with the wi-fi gods. It was Tuesday.

What made the degradation graceful

Looking back, four things separated this from a lesson that merely survived:

1. The fallback existed before it was needed. Not an idea of a fallback. Printed cards, in the room, content-matched to the digital citizens and the simulated AI world and mystery narrative, designed alongside the platform rather than after the first disaster. A backup you have to invent during a crisis is not a backup.

2. The learning objective outlived the tool. The objective was never "students use the platform." It was "students learn how to communicate with the citizens." When you define objectives around the tool rather than the learning, any tool failure is a lesson failure. Design it the other way around.

3. The failure was folded into the narrative. The single best decision was treating the delay as content. Students speculating about island emergencies were practicing the exact language and reasoning skills the project is built on. Graceful degradation isn't just about having a Plan B — it's about making Plan B feel like it was always Plan A.

4. The students never really experienced a 'broken' lesson. From their side, the lesson had a mystery, a discussion, cards, and characters. The infrastructure drama was invisible. In graceful degradation, the user shouldn't have to know that anything degraded.

The practical takeaway

If you're building technology-dependent lessons, do a simple exercise: for each planned activity, ask "what does this look like with no internet?" If the honest answer is "nothing," that's your priority redesign target. It doesn't mean abandoning the digital version — it means building the non-digital shadow of it in advance, so that when the network dies (and it will), the lesson loses features, not purpose.

Lesson 3 is coming up next, and with any luck the wi-fi will hold as the students log in and start questioning the AI citizens by themselves. But luck was never the plan. If the wi-fi goes down again in Lesson 3, I will pull out the physical cards and tell the students: "Card Number 12 is 'fake news'. There are 4 other fake news cards to throw you off the scent and never learn the truth behind the island. Shall we try and figure out which cards might also be fake news?"