[MUD-Dev] Procedural content generation

Hans-Henrik Staerfeldt hhs at cbs.dtu.dk
Fri Oct 12 11:25:47 New Zealand Daylight Time 2001

On Wed, 10 Oct 2001, Brian Hook wrote:
> As I've mentioned in the past, I'm doing procedural content
> generation for a game I'm working on.  A friend of mine made an
> interesting observation:
>   "You don't explore procedurally generated universes, you reveal
>   them.  And that's a different experience than exploration."
> Subtle distinction between the two, but he made a very good point.
> His argument against procedurally generated content is that, no
> matter how good you make the generator, in the end it's still
> random and will feel random.  It won't have the flavor, depth and
> cohesiveness of hand-made content.  The individual pieces don't
> fit together to form a higher level structure that the player can
> identify and relate to.

Define hand-made.

Its after all a gradual scale, where at one end, everything is
fixed, then you can have some generation. The reward for some quest
can either be the 'superfast spaceship Nuarh',100T platinum ore or a
lordly title, picked randomly at the time among options that has
been clearly defined 'by hand'. On the other end of the scale, all
aspects of an area are assigned from multiple-choice functions based
on some seed, but all the choices are still hand-made. What i'm
hinting at is that the fractal function to fill out a generated
game-world is itself a 'hand-made' thing. Sure, your game is boring,
if all the fractal generator has are rivers and trees. However
fractally generated quests could also be more narrowly defined
similar to what your 'keyframe' idea, and still have "the flavor,
depth and cohesiveness of hand-made content", since that will be
what it is.

To be able to generate a more diverse (more fractal) world that
still contain "flavor, depth and cohesiveness", you need to be able
to quantify these features statistically, or structurally. Also,
remember that some aspects of the world are dynamic, while others
are more static.

If youre able to define "key points of interest", you would be able
to include them also in the general ruleset as less diverse elements
in the fractal generator. Therefore option 2 and 3 are very similar.

What i think needs to be done is to identify what exactly in
hand-made contents makes it cognitively different from procedural
content, and then try and create procedural generators that support
these kinds of "flavor, depth and cohesiveness" that is detected.

Next step is ofcourse to create the tools that lets the designers
use the procedural programming instead of the static control they
usually prefer.  The new kind of development you can do is then to
build the world like fractal modules that is tied into their
surroundings with enough flavor, depth and cohesiveness that it
fools the player into thinking its hand-made. At that point the
player will be exploring, not just revealing.

One thing i feel is needed for generating successfull fractal worlds
are some very large scale structures. In your example, large
different alliances between many worlds, large scale conflicts (as
if we didnt have enough of those) and the like on a 'global' scale
(read 'universal') and their effects right down to the NPC
individual youre dealing with are some of the aspects that are
needed (i guess you would call that 'debth'). The important thing is
that on the local scale, the universal scale features have
distinguishable effects.

Making such systems however require long and hard research into
'what really goes into a story', what information to pass between
levels of description, how to depend on them, and most important of
all how to describe it to the computer in a way that the content
creators will be able to use :-)

Hans Henrik Stærfeldt   |    bombman at diku.dk    | work:  hhs at cbs.dtu.dk      |
Address:                |___  +45 40383492    __|__       +45 45252425     __|
DTU, Kemitorvet,        | Scientific programmer at Center for Biological     |
bygn 208, CBS.          |  Sequence Analysis, Technical University of Denmark|

MUD-Dev mailing list
MUD-Dev at kanga.nu

More information about the MUD-Dev mailing list