I recently took what ended up as a 2 1/2 month break from interactive fiction development.  Most of that time got spent on Dragon Age, but I also played through Left 4 Dead 2 due to an offhand comment by Zarf on rec.arts.int-fiction.  In addition, over the Easter weekend I traveled to my in-laws for the holiday and also to celebrate the birthdays of a niece and nephew.  While at Chuck E. Cheese for the latter event, I found an interesting game called “Let’s Go Jungle”, which in addition to sounding like something my daughter might say, incorporated an interesting gameplay mechanic that adds interest to what is otherwise a garden-variety rail shooter.

So to rationalize to myself that I didn’t just flush the last couple of months down the toilet, here’s some of the musings I drew from my “research”:

Dragon Age uses exclusively tree-based conversations.  And if you look hard enough, you can see the seams: NPC responses that are phrased in such a way that they could be legitimate answers to every choice you were given, extra quips bolted on to the front of a generic response, etc.  But for the most part, they work well at giving you the feeling that you’re playing a distinct character, and that you are having a real conversation with someone.  Why is this?

I think it’s related to two different aspects of Dragon Age’s game environment:

  1. The extensive use of game state to modify the content of individual conversations, and
  2. The difficulty of reloading and reiterating over conversations (“lawnmowering”)

Many of the complaints with tree-based conversations in Interactive Fiction revolve around issues of immersion — a typical IF game presents a detailed world model that leads you to believe you have freedom of action, and accepts freeform text input, which reinforces that illusion of freedom.  The use of a conversation tree breaks this illusion and dumps you into a more choose-your-own-adventure (CYOA) model that explicitly and jarringly constrains your freedom of action.

This is much less of a problem in a graphical adventure such as Dragon Age, where the structure is set up from the beginning to be divided into four spheres:  exploration, management, combat, and conversation.  There’s no expectation that you should be able to say anything you want to NPCs, because the game makes that clear from the beginning, and there’s nothing in the input controls of the game (the XBox 360 controller) that leads you to believe you should be able to say anything you want.

So I can see why a naive implementation of conversation trees in IF might be a problem — the medium actively discourages it.  But I still think there’s a lot to learn from the way Dragon Age succeeds in creating a high degree of immersion with limited options.

Let’s look at the “lawnmowering” problem.  Simply stated, lawnmowering is the strategy of fully exploring a conversation tree through the use of the UNDO mechanic.  More generally, any save/restore system enables lawnmowering — it might be harder without direct use of UNDO, but it can be done.

The lawnmowering problem is minimized in Dragon Age primarily because it takes a substantial effort to prepare for it.  The acts of saving and restoring themselves take significant amounts of time, and it’s often not feasible to save immediately before a conversation.  Often you need to succeed in a difficult combat, with an immediate transition to conversation, and the only point at which you can save is before the combat.  So fully lawnmowering a given conversation tree is a daunting proposition.

Even if you were willing to plow in the time to do so, however, you’d run into the second issue:  important conversations are very dependent on the global state of the game.  Depending on who you are, what you’ve done in the game world, and what you’ve said to that character or other characters, you might get wildly different conversation options than you would have had you made different choices.  It’s simply impractical, and in some cases impossible, to unsnarl all the options completely.

The result of this is that you have to make irrevocable choices in conversation if you want to progress at all, or at least choices which are so painful to revoke that you don’t really want to try.  And since they do such a good job of making conversations dependent on game state, it provides a powerful illusion that your decisions matter — that your choices are affecting not just the game world, but other characters’ attitudes toward you.

In an IF context, we have many more options than a simple 1-2-3 conversational tree structure, and some of them wouldn’t lend themselves to the precise tricks Dragon Age uses.  But the general lessons of Dragon Age conversational structure still apply.  A controversial approach would be to strategically limit of save/restore or UNDO in certain circumstances to facilitate an increased investment in NPC interaction.  A more involved but less controversial plan might be to ramp up the differentiation of meaningful conversation based on game state, particularly game state separated in time from the conversation itself.  Even if the endpoints of the conversation were similar, varying the text based on game state would be a fairly easy way to differentiate player choice and increase the sense of immersion and agency.

***

On to Left 4 Dead 2!  I’m not sure the lessons here are directly applicable to IF games — basically what L4D2 does is to adjust the pacing of challenges to attempt to keep players engaged.  Effectively, it’s the “adaptive difficulty” feature made completely fluid.  The game analyzes how “emotionally intense” you are, and adjusts the flow of enemies to create a series of intensity peaks and valleys that both give you time to recover and provide a sense of impending doom.

The IF world lacks (for the most part) a real-time component, so direct control over pacing is not particularly meaningful.  But puzzles often serve as pacing mechanisms in IF, and we might be able to apply some of the L4D2 lessons here, at least in some form.  Aaron Reed has done a significant amount of work related to this; the Drama Manager in Blue Lacuna is a means to detect player boredom and prod the player with a relevant clue.  His Spin extension for Inform 7 is a player-directed mechanism that can be used to bypass a given puzzle in the interests of keeping the narrative flowing.

One of the problems with imagining how these mechanisms might work in an IF game is that there aren’t very many games (that I’m aware of, at least) with bypassable main-plot-track puzzles, let alone puzzles with multiple levels of difficulty.  Infocom’s Wishbringer provided a bypass mechanism in the form of the Wishstone, and Blue Lacuna has its story and puzzle modes.

I’m also not convinced that a mechanism under player control is good for immersion.  The magic of the L4D2 model is that the game itself figures out when it should ramp up the monster count and when it should back down; you don’t tell the computer “Hey, enough!  I need a break!”.  In that respect the behavior of the Blue Lacuna Drama Manager is really much closer to what L4D2 does than Spin or the Wishstone.

As a possible means of adapting the L4D2 design style to IF, perhaps puzzles in a game could be designed to emit in-context, in-character hints after a certain number of failed interactions, or otherwise lowering the difficulty somehow until the player solves the puzzle.  This would preserve the illusion that the player is in full control and overcoming the core challenges of the game, while allowing the author to keep the difficulty balanced to the capabilities of the player.  I doubt whether this could be systematized in any meaningful way, but it might make sense as a design consideration for story-based games that include potentially game-blocking puzzles.

***

OK, enough PC games.  Let’s head to the arcade!  Let’s Go Jungle is a fairly simple, corny rail shooter where you blast waves of oncoming bugs and other vermin with a machine gun that never runs out of ammo.  The gameplay is pretty forgettable, but it has an interesting mechanic:  it’s a two-player game, and each player takes the role of one member of a bickering couple.  You have to work together to survive, and if you perform altruistic or cooperative tasks, such as coordinating your fire or saving the other player from attack, your “heart meter” goes up.  At the end of the game (and I think after each level), you are shown a short scene of the couple interacting, which varies depending on how high the heart meter went.

This mechanic has very little to do with the actual gameplay, but I found it very interesting to see how the couple would react when I watched people playing the game.  I was far more interested in that than in the shooting itself or the other action in the game.  From a design perspective, what Let’s Go Jungle does seems similar to the now-ubiquitous “achievement” concept of putting a bunch of tangentially-related miniobjectives into an otherwise straightforward game.  Let’s Go Jungle doesn’t have multiple tiny achievements, however — they have one big one that is implemented with considerable detail.

The achievement mechanic is great for giving people choices — they can either play the game to achieve the primary objective, or they can choose a different paradigm and play to achieve the secondary ones.  Either way, the player is empowered to enjoy the game the way they choose.  What Let’s Go Jungle does is to make the alternate paradigm more than just a throwaway — there’s enough content that it is in some ways the more interesting way to play the game.

I haven’t seen much explicit use of achievement-style mechanisms in IF, possibly because those that support this mechanic do so at a more subtle level than a bunch of “Achievement Unlocked!!!1!” messages.  But providing multiple paradigms through which players can enjoy a game seems like a worthy goal, regardless of how exactly it’s accomplished.  And I tend to think that the Let’s Go Jungle style of deep implementation makes more sense for IF than the standard way achievements are implemented in other genres of gaming.

So that’s it!  Two and a half months of “research”, and now I’m ready to pull at least some of these concepts into my new IF WIP.  Back to the Inform 7 IDE!