Doing without turns

Back in my post on the Tyranny of the Turn I complained that campaigns with tuns depend on regular player input, and that is something that cannot be relied upon.

I propose to get away from that kind of turn dependency by bringing together three concepts:

• In most games, the player directly adjusts the model; instead, let the player set objectives and policies.  
• Use Artificial Intelligence, which is far less grand than it sounds, to execute the policies.
• Frequent feedback so that the player can adjust his policies when he needs to.

Each of these ideas brings with it problems along with possibilities.  However, I don't see "problems" as bad things -- more as opportunities to do something interesting.

Problem: If the player is to set policy to be implemented by an AI, and the player is encouraged to be open-ended and in-character in his direction, the referee takes on the role of input translator. 

• The shear labor of this is mitigated by good design; I don't see it as any more obnoxious than having to read player orders anyway.
• The possibility of misinterpreting the order is a real risk.  This is a possibility in any written-orders campaign; the risk can be mitigated by having the AI "explain" what it intends to d back to the player in "its own words".  The bits in quotation marks are the interesting part.
• The player needs to feel connected to the environment. I think this is best achieved by implementing the AI interface using Agents.  This, too, is less grand than it sounds. If you have ever set up a table of actions a non-player commander can take and rolled a die to pick one, you have implemented an agent.  We will need one a bit more clever (or at least labor-intensive) but the principle is not fancy.  The agent lets us make the situation appear life-like to the player:  in the real world the King of France did not "move" his armies, whatever that means, he directed the minister of war to make appropriate arrangements.  The minister reported on his success.

Problem: Creating AI to execute policy.  This is an interesting intellectual exercise, but not as big a problem as it might be for a few reasons

• I'm a computer programmer.  I also happen to find AI interesting; this is just a good chance to play with it.
• This is not a video game, so hyper-fast optimized heuristic algorithms are not needed.  I can just be patient instead.
• The end-user interface will be me, reading e-mail.  I can put together something that emphasizes accuracy and simple data entry without having to program something bulletproof.
• I'll be there to hold its hand.  If the AI produces implausible results, or is faced with a new problem that drives it outside its sweet spot, I can directly adjust the model to bridge the problem until I can code a fix. 

Problem: Frequent feedback.  We want to be free of "turn tyranny" but at the same time people will not be on hand 7/24 to reply to game events.  My first thought on interaction frequency are:

• I have to figure out my own cycle.  How much time will I give to this, and how will I reserve and allocate those hours?  The following points must live in that context.
• When a player's orders are read and fed into the system, get an immediate report back from the minister explaining the result (Majesty, our forces will take 6 months to march to Moscow, and logistics are tenuous at best) to the player can rethink if needed. 
• When the player gets feedback or results, there is a deadline and silence is consent.
• There should be some sort of routine and fixed progress of time so players can plan their own lives.  In that respect, it is hard to do away with turns.  What we can achieve is a lowering of player stress by making it possible to skip a turn occasionally without disaster.

Time and space are interesting; maybe we can do something less turn-like.  It is something I will consider in the next post.