I’ve noticed a trend in recent games, a trend I’m calling “de-automation”.  From another perspective this name may be inappropriate, but I’ll get to that later.

Computers Are Good At Some Things But Not Others

Computers can keep track of vast amounts of information and manipulate it quickly.  Consider a “game state” in Dominions 3: A world made up of 200 ‘provinces’, each of which may be affected by various magic effects, inhabited by different numbers of population, hosting various army units which themselves may be suffering from effects, etc., etc.?  For a modern computer it’s easy to keep track of it all.  And then processing a turn where any and all of those units may move and interact?  Done in seconds.

However, this is possible because all the rules by which the game elements interact are clearly spelled out.  Someone had to decide what would happen in each and every case.  They worked for years to get it up and running.  Most people don’t want to spend the time to work out all that programming.

Consider, also, interactive fiction.  One of the hardest tasks is writing a parser to figure out just what the player is trying to do.  When a good parser is created, other authors will happily write hundreds of games that use it.

Humans Are Good At Some Things But Not Others

Most of us can’t keep track of a hundred provinces or armies.  However, we can weigh possibilities and make decisions in a heartbeat; decisions that it would take months to program a computer for.

The Obvious Conclusion

Recently some people have been taking the idea of a computer game and letting people make the hard decisions.  One example is Parsely.  Parsely brings to mind the environment and interaction of an interactive fiction game, but the parser is replaced by a human.  Humans are, of course, notoriously good at interpreting language.  Therefore, the game plays quite smoothly.  The experience becomes much closer to improvisational theater than game playing, since the emphasis is on humans interacting.

Along the same vein is Sleep is Death.  One player enters a computer-graphics world and interacts with it.  The consequences of the interactions are decided by the second player.  The game affords the ability to rearrange and manipulate game objects as if they were scenery.  The computer doesn’t have to interpret or decide anything; that’s all left up to the players.

Yet another example is How to Host a Dungeon.  This “game” is notable for several reasons.  First, it doesn’t include any meaningful choices, thus it may be inaccurate to call it a “game”.  It’s meant to be a solo activity of dungeon creation; through the course of How to Host a Dungeon, you’ll create a series of underground rooms with a history of habitation by various fantasy creatures.  Play consists of actually drawing rooms on a sheet of paper.  This puts a number of interesting decisions into the player’s hands; If dwarves tunnel near a cave containing water, do they dig into it or leave it be?  How close do they have to be to tunnel into it?  These are decisions the player weighs as they go through the game.  A computer might have to be programmed with dozens of complex rules to make good decisions; a human can make them in a moment.  Not only does this save time, it also lets the player guide the dungeon’s development according to their own preferences, thus making it uniquely their own creation.

Taking this concept a bit further, I could mention Dwarf Fortress.  Of course Dwarf Fortress is known for spending a lot of effort to simulate a plausible fantasy world in real time.  However, as Josh Diaz’ master’s thesis “Dwarf Fortress Gathers At The Statue And Attends A Party” notes, much of the “play” occurs when Dwarf Fortress players interpret, repurpose and guide the events occurring within the game.  The game takes care of what it does best—keeping track of an entire world’s worth of information, while the humans do what they do best—constructing meaning and narrative out of chaos.

Another Perspective—Is This Really So Impressive?

From another perspective, this phenomenon is nothing new.  Parsely is a lot like improvisational comedy, where the players are encouraged to portray interaction with an IF game interface (making use of the associated tropes as well).  Sleep is Death is like roleplaying, only you have scenery on the computer.

I suspect the presentation is what’s new here.  There are lots of people who assume everything in the game has to be calculated by the computer, but this recent crop of games shows that leaving it up to the humans can lead to interesting possibilities.  If nothing else, it lets the players see that they really can do interesting things on their own, and it might even encourage them to explore their creativity even more.

The Future, Part A—Computer Training

Neural networks are computer models for making tough decisions.  They are adaptive, in the sense that they go through a learning phase before being used.  In the learning phase, you give the neural network a lot of data to train it; “This is a possible input, and this is what should be output”.  Once training is over, the network knows how to make those tough decisions (in theory, anyway).  De-automation shows us that some areas of decision-making are still difficult; these might be good candidates for neural networks.  (Assuming we want to re-automate everything; I suspect some people will go that route even if not everyone does.)

The Future, Part B—Composition

Three of the examples I’ve mentioned—Sleep is Death, How to Host a Dungeon and Dwarf Fortress—are all designed to assist in the process of creation.  They explore the division of labor between humans and computers.  What other creative tasks can computers help us with?