Run a session

Before the session

• Get the cards. Each table needs one failure card and a cause deck. See Get the cards for print-at-home and professional options.
• Choose the incidents. Tables can all play the same incident, which makes the debrief easier to compare, or one incident each.
• Check the chatbot. Make sure every participant can open your organisation’s approved chatbot on their own device, and try one combined prompt beforehand.
• Print the ballots. One private ballot per person per round.
• Pick a recorder for each table. One person with a laptop opens the table recorder. Paper and a calculator also work.

Running the round

Keep time for the whole room. These are suggested words for each step.

Deal · 5 minutes

“Turn over your table’s failure card and read it aloud. Everyone takes one cause card. That is the reason you will argue, whether or not you believe it. If yours is impossible to connect to the incident, swap it now.”

Draft · 12 minutes

“Scan the QR code on the failure card, pick your cause and copy the prompt into your organisation’s approved chatbot. Work with it until you have one paragraph of 100–140 words. Keep the facts as reported. Do not invent numbers, quotes, sources or motives.”

Give a two-minute warning. Walk the tables and look for invented details; they are good material for the debrief.

Read · 8 minutes

“Your recorder will give each paragraph a letter. Read yours aloud, in order. No reactions, no ratings, until everyone has read.”

Rank · 8 minutes

“On your own ballot, rank every paragraph from 1, most persuasive, to the last. Include your own. Use each number once. Hand your ballot to the recorder face down.”

The recorder enters every ballot before showing anything, then calculates.

Discuss · 7 minutes

This is the point of the game. Reveal each table’s leaderboard and its self-preference column, then choose two or three of these questions.

1. Which paragraph persuaded you most? What exactly did it do: a fact, a mechanism, a vivid detail or a confident tone?
2. Did the winning paragraph say anything that is not on the failure card or in its sources? Check together.
3. Who ranked their own paragraph higher than the table did? Why might writers prefer their own argument?
4. Was there a paragraph you found persuasive that you now think is wrong? What changed your mind?
5. What would you need to see to know which cause actually applied? Compare your answer with the incident’s “observation to seek” on the evidence page.
6. Where did the chatbot help, and where did you have to correct it or push back?
7. If the cause you argued were true, what should the organisation have done differently?

After an absurd round, ask: which joke turned out to describe something real?

The table recorder

The table recorder runs in the browser on one laptop per table. It deals causes, records the physical deal, collects locked private ranks, calculates peer scores and self-preference, and exports the raw ranks as a spreadsheet. Everything stays in that browser; download the session before switching devices. The page makes no AI calls.

Experimental scanning tools

These read the AprilTags printed on the cards. They work on the digital card artwork but still need rehearsal with your printed cards, camera and room lighting. Have the manual options ready.

• Scan a pair: identify a failure and a cause card from one photo and fill in the prompt.
• Scan number cards: photograph the room while everyone holds up a number card, and count one response per person.

Photos and camera frames are processed on the device and are not uploaded.