Saturday, 27 June 2015


This is a short game about processes and building upon what has already been stated: in it, the player's collectively take on the role of a crashed AI trying to diagnose what went wrong with it and complete its function, whatever that is.


In order to play this game, you'll need a large sheet of paper (graph paper or plain paper work best), a sharp pencil, an eraser and a coin: the larger the sheet of paper you pick, the longer the game may take. Start by drawing a 4cm x 4cm box in the upper left corner of the sheet (it doesn't matter what shape the paper is or whether you orient it in portrait or landscape): the overall aim of the game is to reach the bottom right corner by drawing a series of boxes and arrows to represent the decisions the AI is faced with, the data is gathers and the actions it takes.

At the start of the game, as the AI reboots and goes through a check-list of procedures, none of the players will know what the AI's purpose is: this will only be learned through play, as the AI accesses its database and sensors. The story of the game will told through the data the AI gathers as it restores itself to functionality.

It isn't necessary to detail every function and process of the AI as it reboots, only the interesting ones: the game is about finding out what the AI's purpose is and whether it can complete that successfully, so it doesn't need to be 'fully programmed,' a rough sketch of its programming is sufficient. The starting point for the game is to choose an ACTION or DECISION for the first box on the chart: how does the AI respond when it reboots after crashing? What is the first thing it does? Since the first box on the flowchart potentially determines a lot about the story, you might want to make this choice as a group, before proceeding to play in turn order.


The AI may begin its reboot by taking  an ACTION: each ACTION the AI takes involves it activating a part of the physical system it is operating and has access to. Some potential ACTIONS include:
  • Begin warm-up sequence for main thruster.
  • Send crash report to mission HQ.
  • Engage active defence systems.
  • Power-up drone workforce.
Whenever the AI takes an action, draw a 4x4 box on the flowchart: this should be connected to at least one prior box by an arrow pointing to it. Note the type of ACTION taken and leave space to draw arrows leading out of this box to subsequent ones. Having described the ACTION, complete your turn by tossing a coin to see what the outcome of that action is: if the result is heads, the ACTION is successful and the AI proceeds to the next point on it's check-list; if the result is tails, the ACTION is unsuccessful, so the AI must attempt to rectify the failure before it can proceed. In either case, draw an arrow leading from the box to the next one, but try to be consistent, so that all successful tasks completed are in a continuous line, whereas unsuccessful ones form small loops attached to this line.

If an ACTION fails, it fails: don't simply repeat an ACTION on subsequent turns, as the game will be dull if it features too many back-ups, fail-safes and auxiliary functions. A failed ACTION can be corrected by drawing an arrow back to it after a loop that contains a successful DECISION or ACTION: the AI can gather data and perform functions that allow it to repair or work around failed ACTIONS, so the second time any ACTION is taken, it succeeds, as long as the arrow that is drawn back to it originates from a successful ACTION or DECISION.


The AI may attempt to gather more data on its situation, in order to assess what steps it needs to take to resolve a problem; DECISIONS are taken by accessing memory banks or live sensors, interrogating them for whatever data is needed in order to determine the correct course of action. In order to make the DECISION, the AI must apply a binary choice to the data, usually phrased as an IF/THEN statement; some typical DECISIONS might include:

  • If the perimeter is breached, then seal all internal doors.
  • If the drive plasma is below 10%, then change course for the nearest star.
  • If no confirmation is received from mission HQ, then prepare missile launch sequence alpha.
  • If total profits are less than total outgoings, then dismiss the least productive staff member.
When you phrase a DECISION, draw a box on the chart as for an ACTION and describe how the AI obtains the data required, by accessing stored data or collating live feed from sensors, then toss a coin: if the result is heads, the question is answered affirmatively, but if tails, it is answered negatively. When a DECISION is answered affirmatively, the AI takes the ACTION described: add the box for that ACTION to the flowchart but do not toss a coin for it, it succeeds automatically. When a DECISION is answered negatively, the AI must seek further data, therefore the next player must also frame a DECISION for the AI.

DECISIONS are often triggered by failed ACTIONS or negative responses to prior DECISIONS: for example, if an ACTION like 'Begin heating of incubation chamber' fails, then the AI will most likely seek a reason for the failure and take the appropriate steps to rectify it.


The game requires the most narration where DECISIONS are taken, as these are at least a two-tiered process: the active player must both narrate the data being gathered, stating how it is obtained and then incorporating the result into their narration. For example, a player states on their turn that the AI must make a DECISION about this statement: "If there is vehicle activity in the city, send out a drone to investigate." They might narrate this by saying the following: "The AI signals the satellite surveillance network requesting the most recent images of the city at this location." The player then tosses a coin, with the result being tails, so the DECISION is answered negatively, like this for example: "12 microseconds later, the network sends a dossier to the AI containing images of the city over the last hour; comparison of these images shows no significant change, therefore vehicle activity is flagged as absent."


The game continues until the AI reaches a CONCLUSION, a point at which it has completed all current functions required of it and can go into hibernation mode until it receives more direct input, such as new instructions or urgent new data. The game should arrive at it's CONCLUSION by the time the flowchart has reached the bottom right corner of the page: the line of boxes may loop back and forth on the page before it reaches the CONCLUSION or it may take a more direct route if a shorter game is acceptable. The CONCLUSION may even be reached abruptly, if an ACTION is successfully taken that all players feel adequately represents a CONCLUSION.

No comments:

Post a Comment