next »

[ஒழுக்கின்மை (Paul Eres)]

if you guys are really interested in this stuff, chris crawford has done a lot of research here (and has a few books on it, and about 3000 pages of material on his site); i've read through it all and used storytron quite a bit and remain unsatisfied that procedural stories as they exist today can be entertaining to anyone but nerds who like things just because they are procedural. perhaps the game that comes closest to being entertaining to a broader audience is facade.

[jpgray]

if you guys are really interested in this stuff, chris crawford has done a lot of research here (and has a few books on it, and about 3000 pages of material on his site); i've read through it all and used storytron quite a bit and remain unsatisfied that procedural stories as they exist today can be entertaining to anyone but nerds who like things just because they are procedural.

The closest Crawford ever came to good, procedural drama was Siboot.  Most of those innovations were anticipated in Captain Blood, but he really put them together in an interesting way despite the plot contrivance needed to support the mechanics.  He's been moving farther and farther away from what made Siboot a qualified success, and it's painful to watch. 

As for the rest, great procedural stories have always been generated by gaming rules and environments--see the treacly sport specials, the melodramatic write-ups of chess matches, etc.  As for computer games, everyone remembers their favorite soldiers (or first encounter with a Chrysalid) in X-COM, or their most dramatically thrilling game of Master of Orion (the first one).  Procedural stories in games that attempt to deliver procedural content with text, attempting to simulate interpersonal drama in a meaningful narrative sense, are mostly ugly and badly paced as you say.

perhaps the game that comes closest to being entertaining to a broader audience is facade.

I think this says a lot about what you view as procedural storytelling.  Facade is not what I would hold up as an example of procedural storytelling, and moreover it has several severe limitations:

-The parser is of the standard keyword variety, its only virtue being that its behavior is disguised.  In my view, a debug mode wherein the parser outputs its interpretation of input would reveal its inevitable shallowness.  That's no foul on the game, as every parser in existence is too shallow and stupid for dramatic purposes.

-The player has no real agency in the main drama, aside from typing an appropriate keyword that will set off pre-written beats in the real action, which is between the two NPCs.

-I don't think it is good to describe Facade as procedural, since nothing whatever will happen that the authors have not anticipated.  The main action is all determined, while the player can only control the order and flow of the content.

-It's an interesting game, but rather than revealing the possibilities of its approach it reveals instead the limitations, requiring: a bystander playable character; distracted/stressed main actors who can believably react too strongly, nonsensically, or not at all to the player's input; a severe length limitation yet also mountains of pre-written content required to fill it out; a "mellow" parser that simply hunts for keywords regardless of context; etc.

[s0]

(i) remain unsatisfied that procedural stories as they exist today can be entertaining to anyone but nerds who like things just because they are procedural. perhaps the game that comes closest to being entertaining to a broader audience is facade.

But the Sims is one of the best-selling games of all of time and a lot of people (nerds as well as non-nerds and even non-gamers) enjoy it precisely because of the "stories" it creates. It allows you to create your own cheesy soap opera plot as you go along (including the whole "multiple generations" hijinks) similarly to how Dwarf Fortress generates a generic fantasy plot. And yes DF's appeal is limited to nerds, but that's for other reasons.

This sort of thing is fun simply because you get to make up your own story through gameplay, it doesn't need to be great literature or be based on some nerdy fascination with all things procedural. It's the digital equivalent of playing with action figures and plastic swords as a kid.  

[Gnarf]

broader audience

Lol.

As for the rest, great procedural stories have always been generated by gaming rules and environments--see the treacly sport specials, the melodramatic write-ups of chess matches, etc.  As for computer games, everyone remembers their favorite soldiers (or first encounter with a Chrysalid) in X-COM, or their most dramatically thrilling game of Master of Orion (the first one).

So, people tell little stories about things that have happened and that they found exciting. So dynamic stories in games is a matter of making awesome stuff happen in games. Unlike game design in general which is all about making them boring. I can see the need for a distinction.

[J. R. Hill]

Here's an idea for y'all.

Procedural storytelling: Create seed events that occur and branch off to other events eventually connecting in a timeline to earlier and later seed events.  The branch events naturally would have to lead and be influenced by earlier and later events.

Example 1: An early event requires a player to do something good, and a later event requires a player to do something bad.  There have to be branch events between that lead to this change.

Example 2: A later event requires the player to only have one eye.  At some point during the event line the player either must lose an eye, or must never have had 2 eyes at the beginning of gameplay.

The challenge then would be to create a large number of general/branch events that lead to different developments, a fairly good number of key/seed events that will be memorable to the player.  It'd also help the uniqueness of a story to create them all in a madlibs fashion, so that even if you get robbed on a highway 5 times during a game, the thieves are different, the dialogue is different, the amount stolen is different, and most importantly the player can distinguish clearly between the different times it happened, but also see that the player's character is growing in distrust and dislike of thieves which leads to later event possibilities.  The player must also have either no control or limited control over the events' outcomes and occurrences.

[FatHat]

Am I in the minority in thinking that it wouldn't be that difficult to make NPCs talk procedurally? I mean, obviously it would be harder to make them talk with proper content, but given a rich enough representation of the semantics of what they want to say I don't think it would be that hard to come up with natural-sounding procedurally generated speech.

Well, there are conversation bots that can sort of work (MegaHAL) and such, so I don't think it's that unreasonable, especially since you'd be in a constrained world.

I feel like you could probably get away with blatant archetyping that way, because stories don't have to be as intellectually interesting if they're emotionally engaging

Man quoted for fucking truth -- that's like every Final Fantasy game ever made  

[gimymblert]

Am I in the minority in thinking that it wouldn't be that difficult to make NPCs talk procedurally? I mean, obviously it would be harder to make them talk with proper content, but given a rich enough representation of the semantics of what they want to say I don't think it would be that hard to come up with natural-sounding procedurally generated speech. Sure, you'd have to make a dictionary of words and their various forms, but creating sentences from semantics generally doesn't have much subtlety. It's going the other way that's hard.

You may check for "Chatterbot" (starting with alice the precursor) There is a whole contest about the "turing test"

You may check for the famous Emily SHORT's IF experiment "galatea", check also the conversation tips she gives on her blog (starting on gamesetwatch)

You may check the academic paper of "façade" (by michael mateas and andrew stern) and check the project OZ where it originate.

What you are talking about is one the oldest and longest studied topic in human computer interaction, it's a graal, and many experiment have been made, there is powerful community behind it and incredible amounts of efforts are made towards,and to essentially going nowhere  (for the moment)!

I always keep an eye on this and procedural generated plot

Emily's "game practical" conversation design series:
http://www.gamesetwatch.com/2009/05/column_homer_in_silicon_the_co_1.php
http://emshort.wordpress.com/2009/07/22/modeling-conversation-flow-types-of-npc-initiative/
http://emshort.wordpress.com/2009/07/24/modeling-conversation-flow-subject-changes/
http://emshort.wordpress.com/2009/07/26/modeling-conversation-flow-transitions-in-player-speech/
http://emshort.wordpress.com/2009/07/27/modeling-conversation-flow-interrupting-the-player/
http://emshort.wordpress.com/2009/07/30/modeling-conversation-flow-npc-repeating-information/
http://emshort.wordpress.com/2009/08/02/modeling-conversation-flow-open-ended-questions/
http://emshort.wordpress.com/2009/08/05/modeling-conversation-flow-actions-in-conversational-context/
http://emshort.wordpress.com/2009/08/06/modeling-conversation-flow-silence/
http://emshort.wordpress.com/2009/08/08/conversational-analysis-studies/
http://emshort.wordpress.com/2009/08/10/modeling-conversation-flow-beginnings-and-endings-using-scenes/
http://emshort.wordpress.com/2009/08/12/modeling-conversation-flow-multiple-people-conversing-some-closing-caveats/
http://emshort.wordpress.com/2009/08/26/the-conversational-script/
http://emshort.wordpress.com/2009/12/10/moods-in-conversation/

General set of link that IF wiki compile about the subject!
http://www.ifwiki.org/index.php/Past_raif_topics:_NPCs_and_AI#Conversation_interfaces

[SirNiko]

Here's an idea for y'all.

Procedural storytelling: Create seed events that occur and branch off to other events eventually connecting in a timeline to earlier and later seed events.  The branch events naturally would have to lead and be influenced by earlier and later events.

Example 1: An early event requires a player to do something good, and a later event requires a player to do something bad.  There have to be branch events between that lead to this change.

Example 2: A later event requires the player to only have one eye.  At some point during the event line the player either must lose an eye, or must never have had 2 eyes at the beginning of gameplay.

The challenge then would be to create a large number of general/branch events that lead to different developments, a fairly good number of key/seed events that will be memorable to the player.  It'd also help the uniqueness of a story to create them all in a madlibs fashion, so that even if you get robbed on a highway 5 times during a game, the thieves are different, the dialogue is different, the amount stolen is different, and most importantly the player can distinguish clearly between the different times it happened, but also see that the player's character is growing in distrust and dislike of thieves which leads to later event possibilities.  The player must also have either no control or limited control over the events' outcomes and occurrences.

Yoda Stories and Indy Desktop Adventures did this pretty successfully, I think.

The game pretty much picks a 'story', which is a single event (like rescue Han Solo from the carbonite chamber). Then the game pulls several events from a much larger pool of events, arranges them randomly about the map, and then makes the reward for completing each event the key for initiating the next event. For example, you might need the stun grenade to get past the guard blocking access to the carbonite chamber, but then it turns out the grenade can be obtained by doing a little sneak mission into an imperial warehouse. The warehouse key is obtained by giving the trader a sarlaac fang you got by killing a bounty hunter who had it, and so on. The events are a little disjointed sometimes, but every game (they are minesweeper sized) has a coherent story that moves toward progression.

This might not be a bad model to start with if you intend on generating something more robust, possibly by selecting a randomized villain who flavors each encounter in a game, or creating more continuity by making the specifics of one event alter the specifics of the next.

While not good for a Final Fantasy-esque game, it is fun for a quickie coffee break type of game.

-SirNiko

[Ben Kuhn]

Neoshaman, I wasn't talking about including all the subtleties of conversation, which is of course a really difficult problem. I was just saying that the act of transforming some computer representation of semantics into human speech is pretty easy. So if you were willing to "cheat" on the step where the semantics are generated - for example, using IF-style commands rather than natural language parsing, or just mixing together a large number of pre-generated sections of computer-representation-of-semantics instead of inventing new ones, or letting NPCs ignore conversation subtleties for the most part - I think you could create a pretty good illusion of procedural speech.

[gimymblert]

Well, i had worked on a secret procedural stories game project before my society get caught in side effect boom due to global financial crisis. According to my creative director, expert says that procedural speech is not quite there (dynamic tone adjustment control mostly).

BUT in a full constrain framework as you suggest (limited answer) i can see that maybe "Half procedural" speech can be done, with sample word/sentence recombine like sample instrument (like in mysims voice customisation)?
Well i have no idea i didn't investigate enough this technical aspect.

[droqen]

So dynamic stories in games is a matter of making awesome stuff happen in games. Unlike game design in general which is all about making them boring. I can see the need for a distinction.

I thought you were serious for a second but then I

However: while 'awesome stuff' plus freedom in gameplay makes for great stories about games, that's not what dynamic stories are about.
There is a huge difference between a story about something awesome and a really good story.

[Parthon]

There's also a difference between a story from a player about a game, and a story from the game about the player.

Players can turn events into words much better then a computer can. The tricky part is "Why?", and then find out a way to teach a computer to be able to do it.

[jpgray]

As for the rest, great procedural stories have always been generated by gaming rules and environments--see the treacly sport specials, the melodramatic write-ups of chess matches, etc.  As for computer games, everyone remembers their favorite soldiers (or first encounter with a Chrysalid) in X-COM, or their most dramatically thrilling game of Master of Orion (the first one).

So, people tell little stories about things that have happened and that they found exciting. So dynamic stories in games is a matter of making awesome stuff happen in games. Unlike game design in general which is all about making them boring. I can see the need for a distinction.

You've missed the point entirely.  "Game design" that multiplies meaningless, tedious and arbitrary interactions -does- make games boring.  One chief culprit is NPC conversation systems in narrative-driven games: the NPC has no agency or ability to alter the game world, and only exists as a vending machine for quests/info/baubles.  The only possible outcome of the conversation is that the player throws the boolean switches required by the game script, or he doesn't.  Failure is impossible, since the NPC just stands there for hours, weeks or days at a time until his script function is fulfilled.  These are lethal flaws for meaningful interaction, and what happens when "designers" try to make it interactive?

You usually end up with a needlessly complex system of personality traits and exhaustive dialogue trees, none of which require ability or creativity of the player.  The player just needs to throw the logic switch and move on, but no, the designer's ego requires him to mindlessly traverse a mostly barren tree of dialogue to get the one or two fruits necessary to advance things.  What happens if he does so in the improper way?  Failure, but not ultimate, interesting or meaningful failure--the game script can't generate a meaningfully new path on the fly to account for such, so the NPC just stands there until the player gets it right, or requires the player to reload.  If the writing is good, the player will happily reload or try out all the options for hours, to mow the lawn, in other words, because it is a beautiful lawn.  The actual interaction, however, is still mechanical and mindless.

Alternatively, in Master of Orion (the first), diplomatic conversation is ugly, abstract and simple, but has some actual meaning.  The player has much at risk, since the computer can alter the gameworld in every crucial and devastating way the player can.  The robust environment and game rules (game design) allow for a rich variety in variable states.  The conversations in the game are -all about- these variables--what's at risk, what might happen, what the player might gain/lose, what the NPC might gain/lose, etc.  Crucially, no two instances of conversation will ever be exactly alike, and most will be meaningfully different.  The effect of this is to require situational thinking, awareness and creativity from the player.  Over all the mechanics is the passive gloss (art, music, text, etc.) of warring space empires, which, while silly, does make the impersonal numbers of the thing more evocative in a narrative sense (game design again).

The problems this method runs into is that the narrative, while dynamic, is usually poorly paced and abstract or ugly.  However it bears more relation to "game design" than does writing reams of set-in-stone text and QTEs, resulting in 2-6 paths that never change.  A massive multimedia choose-your-own-adventure which forces the player to play tea party with inert mannequins is not much for interactivity, however good the passive content.  If the passive content is good, however, the experience may still be wonderful.

[Gnarf]

You've missed the point entirely.

Okay.

"Game design" that multiplies meaningless, tedious and arbitrary interactions -does- make games boring.

That doesn't really have anything to do with what I posted, but I'll disagree just because it's the internet: No, I feel that boring is more fun than fun while fun is more boring than boring.

However it bears more relation to "game design" than does writing reams of set-in-stone text and QTEs, resulting in 2-6 paths that never change.

That was the point I was making. By the way. That the reason why people write little stories about chess games is mostly that chess a pretty nice game where cool stuff can happen. It doesn't have a whole lot to do with storytelling, just game design in general. (Of course, the actual writing, that is done while no one is playing the game in question and no interactions or dynamic anythings are happening, happens to be storytelling.)