Some Interesting Tidbits About Fable 3

Peter Molyneux has a reputation for delivering games that don’t live up to what he says about them – a reputation he fully deserves.  Case in point: the dog in Fable 2.  I liked my dog, but he did not set a new standard in emotional impact in video games.  That said, his company actually does deliver good games, and Molyneux has some good ideas.  He just hypes his ideas…a lot.  He’s a hyperactive man…seriously, have you seen the man chatter?

That said, the things he presents about Fable 3 in an interview with Joystiq are interesting.  It looks an awful lot like Molyneux has a core experience he’s trying to get across, and while that exists in the prior Fable games, he’s still iterating on it.  One of those things is figuring out a granular but compelling means of having the player interact with the world, while giving players the opportunity to feel like this really is their character, for whom they can make real decisions and to whom they can emotionally connect.

It’s actually pretty interesting to compare Lionhead’s take to Bioware’s.  Bioware tries to provide players with narrative control via dialog state machines, which usually consist of series of small, disconnected dialog state machines, along with some global state information that gets set at certain narrative turning points.  The result is that Bioware’s writers have an awful lot of granular control over the delivery and content of dialog.  Subtle narrative points and characterizations can be delivered, because there are relatively few paths that will be taken at any point.

The flip-side is that there’s a limit to how much personality the player gets to imbue into their own character.  Broad decisions are absolutely under their control, and some broad conversational direction is available, but at any one point a player character really only has 2-3 possible personalities.  As a player, you tend to either stick with one direction or another, or you schizophrenically bounce between the two.  Heck, in Mass Effect 2 I’ve been weirded out more than a few times when I’ve come out of dialogs and gained paragon points and renegade points in equal measure, and I have no idea what it was I did…or thought I was being a renegade.  I’m not even entirely sure what those two personality types mean in some cases.

The Fable games have opted for a sandbox, akin to the Sims.  Since dialog options are non-existent for players, they can project whatever reactions they want on their character.  The character is a toy, a tool for moving about a world where a narrative is happening and the player moves with it.  Within those confines, the player can do whatever the hell they want.  While in a Bioware game the results of player decisions tend to be reflected in the way dialog plays out, in Fable it’s shown via the way the world reacts (generally in how the people run screaming from you or come gathering around to fawn over you).  Because the visual cues of the world tend to have higher simultaneous information content than dialog trees, and because sandboxes tend to be better at representing complex world states than hand-crafted state-machines, Fable delivers a more complex set of interactions between player and the world.

Of course, it loses out dramatically on the ability to deliver a subtle, well characterized narrative.  A conversation consists in an exchange of, effectively, state information, where each new entry in the exchange is contextualized on prior entries (at the simplest level).  Attempting to do this inside a sandbox, where the same AI machine drives the characters, limits the amount of interesting conversational options.  It’s precisely for this reason that Fable doesn’t let your main character talk: if they did, they’d lose some of the focus on the world interaction options, and they can’t make talk work well in a sand-box.

Don’t get me wrong, I think these approaches could be combined…but not easily.  Sandboxes don’t sit well with top-down narratives.  To some extent you have to limit the impact of the narrative on the sandbox, and if you let the sandbox dictate the narrative…well, it’s not exactly top-down then.

To some extent, it sounds like Lionhead is shifting to focus more on the sandbox.  The broad narrative consists in a goal: overthrow the tryant and rule the land; beyond that, the world is your playground.  To stitch this together, they can create a series of set-piece side-quests and included cutscenes which help propel you to that final goal.  The way you work through those, and the way you interact with the sandbox, set the tone of the game narrative for your individual playthrough.

So that’s the experience Molyneux looks set to convey.  Fable 2 did an adequate job of delivering it – certainly I felt like the world was my playground.  However, it appears they want to expand on the ways the player has to make their character impact the world.  Fable’s world is mostly people; the land, flora, fauna, and buildings are typically static.  So the logical jump is to expand on how the player can interact with other people, without it involving dialog.  How do you give the player a chance to tell the world something?

Molyneux apparently went to ICO for an answer, and yanked out dragging people around by their hand.  I have to admit, this is a brilliant mechanism for delivering conversation in a sandbox: let the environment provide the contextual clues, which is already an inherent part of a simulation.  Agents in simulations are generally driven by reaction to the immediate environment, built to exhibit self-contained behaviors when reacting, and then given a global method for moving around the environment and selecting between immediate options.  That lets you relatively simply piece together a personality exhibited by behavior, simply by tinkering with broad motivations and immediate reaction choices.

That means that all the important bits of your environment already contain important contextual clues, both for your simulated agents and for the player.  Once you add in the option for the player to order an npc to do something (by dragging them somewhere), you’ve begun something of a conversation, where context exists to help drive it forward.  Consider: I have my character grab another by the hand and drag them to a diner.  The context is simple: I want them to do the thing that they do at diners, whatever that is.  They can then react to that, however their motives please them.  While some added decision-making code is needed to drive their reaction (more motives get drawn into this, because not only do they react to the diner but they also react to your request), it’s not a very big deal…and it requires no dialog trees.  The npc can, in the simplest case, agree or disagree.  Depending on their response, the player can react, which in turn adds additional context (it’ll affect motive levels for the npc, changing their future responses), leading to a very primitive conversation (akin to pointing and grunting and hand waving and shouting).

The other thing Molyneux is attempting to get at is the character’s reflection of their impact on the world around them.  Hence we get things like “followers” as levels (a dubious notion from the standpoint of communicating character package to the player), weapon morphing (a relatively simple notion that flows pretty naturally from character morphing), and even more character morphing.  It looks like we’re getting more paths along which our action can have an impact, meaning the results may vary dramatically between characters.  This may be cool…or it may lead to some really fugly characters.

Regardless, Molyneux hasn’t disgusted me yet…as long as he keeps Natal out of my Fable 3, I’ll probably be happy.