Failure-mode role-play

Play the system or person at the moment it fails and narrate the cascade, so you learn failure as a process, not a moral punchline.

When to use this: When you want to go deeper
Learning impact: Reliable gain
Learning evidence: Credible with caveats
Where you are with the topic: Building understanding
How often to use it: Worth repeating often

When to use it

You understand a system when it works but not how it fails.
You need to narrate a cascade from inside the failing component or role.
Post-mortems and failure-mode thinking are on your course or job.

What to ask AI

Type something like: “Failure-mode role-play: you are the primary key under a thundering-herd retry storm. Narrate the cascade from inside. Pause for me to state what I would change. Do not jump to the post-mortem summary first.”

You propose the fix; AI stays in the failure voice until debrief.

How it helps

Failure-mode role-play builds causal tracing through a cascade, not only a happy-path model.

Speaking from inside the failing part makes early neglected signals harder to ignore.

How to run it

Pick a system and a failure seed.
Let AI narrate from inside the failing element.
Interrupt to name what you would change and when.
Debrief into a short causal chain with owners.
Compare to a real incident report if you have one.

Example exchange

You

You are a bridge bearing during overlooked corrosion. Narrate toward collapse, but pause after each stage so I can intervene.

Salt, micro-cracks forming, an inspection window skipped… Your move at this stage?

You

I’d have required hands-on inspection at year five instead of drive-by checks.

Too late in this timeline. What signal already existed at year three that your year-five plan still ignores?

Copyable prompt

Failure case: [ ].
I will role-play as [VANTAGE] from T0 of failure. Ask me what I perceive, what
I cannot know, and what I do next, turn by turn. Stop me if I jump to hindsight
blame. After the cascade, ask me for one interrupting change.

The Tell

You know this method has flipped when you enjoy the drama and never name a controllable early signal.

Spectacle is not analysis.

Every cascade needs an earlier observable you could have acted on.

Principle evidence

Strength of the underlying learning idea, not a claim about AI products.

The underlying learning idea is rated moderate. Failure-mode role-play relates to failure analysis, FMEA-style thinking, and learning from errors.

AI delivery evidence

Whether an AI tutor delivers this method well is a separate question.

Principle moderate as causal tracing practice; AI cascade stories may omit real failure physics, speculative.