Editorial Gearhead 2 plot generation talk

[Saint_Proverbius]

Tags: Gearhead 2; Joseph Hewitt

<b>Joseph Hewitt</b> has posted a little background info and other such on the plot generation of <a href="http://gearhead.roguelikedevelopment.org/">Gearhead 2</a> in the mailing list which I thought was moderately interesting. So, here's the post:
<br>
<blockquote>Right now there are six tasks which can be used to
<br>
select plot components. A task describes roughly what
<br>
I expect the PC to be doing at this point in the core
<br>
story. It isn't nessecarily a quest assigned by a NPC,
<br>
though it could be. The tasks are "no task",
<br>
"investigate encounter", "speak socially with target",
<br>
"speak business with target", "locate target NPC", and
<br>
"guard scene".
<br>
<br>
The character I'm playing with doesn't belong to any
<br>
faction, so coming up with adventure ideas is more
<br>
difficult than if he did. I can make certain
<br>
assumptions about a character with a faction: Knights
<br>
and Privateers want to fight someone, Pirates want to
<br>
steal something. I can also legitimately use NPCs to
<br>
boss them around. I can't do either of those things
<br>
with unaligned PCs.
<br>
<br>
Currently I'm building up the "attack by unknown
<br>
enemies" plot state. The components involved in this
<br>
state should be slightly different depending on the
<br>
nature of the PC-aligned faction. If the PC-aligned
<br>
faction is the government of the city, the unknown
<br>
enemies are attacking the city. If the PCAF is a
<br>
corporation, they're attacking the corp's holdings. If
<br>
the PCAF is the police, these mecha attacks will be
<br>
handled as a criminal matter.
<br>
<br>
Creating a XRAN episode for GH1 usually took 2-3 days
<br>
of work. Creating a simple XXRAN component for GH2 can
<br>
be done in about 20-30 minutes. This is because the
<br>
new components are individually much smaller than full
<br>
adventures and can take advantage of persona fragments
<br>
and random scene content.
<br>
<br>
I'm somewhat concerned that the components I've
<br>
written so far aren't dramatic enough. The PC isn't
<br>
being given enough choices, and the PC/NPC actions
<br>
don't have great enough consequences. I blame this on
<br>
the current lack of decent giant robot anime on both
<br>
Tooniverse and AniOne.</blockquote>
<br>
You're probably better off looking at some decent sci-fi for plot ideas anyway. Fundamental plot point ideas probably don't rely on there being giant robotic vehicles anyway. Even plots in more standardized sci-fi that do revolve around something dealing with technology can be hammered to fit in to the stompy robot theme.

 

[bryce777]

Like fighting off an alien supervirus...with your giant stompy robot.

 

[Saint_Proverbius]

Basically, yeah. Most decent sci-fi plots have very little to do with the technology around them in a direct fashion anyway. Typically the technology is just a tool to resolve some part of the conflict, but it doesn't really have to be. Or course, it could be a means to generate conflict, but something new introduced to the world in terms of technology that has unanticipated consequences could work in any sci-fi setting. The same thing goes for the inevitible breaking down of much used technology causing conflict and so forth.

Hell, look at the original Star Trek series. How many of those plots delt with Federation technology? Not too many.

 

[bryce777]

Actually, I think you have a goos point, but I was taking it to an absurd extreme because that is the vision that popped into my head. Like shrinking your giant robot and shooting it into the infected russian president to fight off his alien superdisease.

I find the idea of randomly generating plots to be extremely interesting. Even before this article it is somethign I have been thinking of a lot - I think this is pretty simplified, but I think if you could make some complexity and some dependency between plot points then you could have some very interesting replayable scenarios that pop up.

One of the biggest things that I dislike about games today is theya re either too simple or else they are not replayable enough and something like this could really spice it up...but only if done very well.

 

[Saint_Proverbius]

I tend to agree. I fail to see why more games aren't using randomly generated events to progress aspects of the game when most of the hand scripted events are just as menial as something even a crude random generator can spit out. At least the random generator will give you something new each time as opposed to "Same old shit, different playthrough."

As long as you have snippets of plot points that mesh together and a system that knows which points can mesh up to what came before them, I'd say you could do a pretty decent job of generating stories. It sounds like GH2's generator will already know what the player is doing by tracking certain decisions he's made about how he's going to play the game, so the end result should be a pretty decent play experience based around how the player is playing. To me, that's just a damned fine idea and one that I'm amazed the commercial industry hasn't come up with.

 

[bryce777]

I think oblivion is a good example of what you are talking about. You could easily have the mage assassin sometimes be an ice mage or necromancer, and have it sometimes turn out to be the girl who attacks you and the guy she fingers is innocent. Any change forces you to actually read the text and wake up a little instead of waltzing through it by rote the second time you play a game.

I found it prety funny that a lot of the quests such as the mountainlion in basement thing were exactly the same as the random quests that could be thrown out in daggerfall...except scripted for some reason.

 

[Saint_Proverbius]

Well, I didn't want to mentiont he M-word or O-word in this thread, but honestly the M-word game is the one that actually got me thinking about why random generation was tossed in favor of hand scripting which is just as bad as what the random generator can toss out.

Random generation has a great deal of possibilities even beyond just making each play through mostly unique. In fact, you can have randomly generated overlapping and multiple plots if the player is shying away from one method of getting through the game and favoring another.

When you think about it, random generation of plot point is fantastic for getting around a lot of the problems that heavy story CRPGs face, like killing vital NPCs and so forth. Whoops, you broke a plot point by killing Lord Douchbag! Well, the game just has to recognise that broke that generated point and then generate a plot point that doesn't have Lord Douchebag in it.

 

[Gwendo]

I don't think plots that touch your emotions and that feel rewarding, could be randomly generated.

But for little side-quests, I think it's interesting.

 

[bryce777]

I was also thinking about having multiple paths at every point, and some things more rpessing than others - such as one plot point may lead to a sort of bad result if you ignore it long enough such as some bandit raids for example. Nothing gamebreaking, though.


As for plots touching you, I think that so far procedurally generated stuff has been very generic, but it does not necessarily have to be. Whole plots can be done manually and thrown out with some variation, once in a while. You can also use characters from plot to plot to develop them a bit - and if they die then hopefully the player will notice it, but at the same time it will not impede the game if so.

Like I said, we need more complexity,a nd that is possible, just difficult. Being able to string plot points together semi cohesively and overlap them is not easy, but it's possible.

For example, the game itself could be a random theme with random variations on a theme. This goes being merely generating out a quest per se as generating out the whole game's plot. You will have to write a lot of material to do this to act as 'seeds' and do a lot of work to keep things coherent, but it is possible and it is something that interests me a lot both as a player and a programmer....

 

[spacemoose]

I don't think plots that touch your emotions and that feel rewarding, could be randomly generated.

I agree. I don't think it will ever be possible to automate quests without having them sound repetitve after you've done a few. Why? Because the human brain is great at spotting patterns, its what helps us survive. Once you've seen one template quest, replacing the nouns, placenames and verbs to make it 'new and fresh' would not work.

 

[pyrrho12]

I'm somewhat concerned that the components I've
written so far aren't dramatic enough. The PC isn't
being given enough choices, and the PC/NPC actions
don't have great enough consequences. I blame this on
the current lack of decent giant robot anime on both
Tooniverse and AniOne.

You're probably better off looking at some decent sci-fi for plot ideas anyway. Fundamental plot point ideas probably don't rely on there being giant robotic vehicles anyway. Even plots in more standardized sci-fi that do revolve around something dealing with technology can be hammered to fit in to the stompy robot theme.

I wasn't really being serious there, just looking for scapegoats. I'd say that the honest reason why the components aren't very good so far is that I'm just getting used to the new plot system. Writing bits for a random plot generator is much different than writing a regular story, or even a regular game.

One of the biggest things that I dislike about games today is theya re either too simple or else they are not replayable enough and something like this could really spice it up...but only if done very well.

Replayability is the main reason for adding random plots to GearHead. Like the creators of the original Rogue, I want to be able to play my own game once it's completed.

 

[Saint_Proverbius]

I wasn't really being serious there, just looking for scapegoats. I'd say that the honest reason why the components aren't very good so far is that I'm just getting used to the new plot system. Writing bits for a random plot generator is much different than writing a regular story, or even a regular game.

That reminds me. It's nearly about time we did another interview about the crazy fun new stuff in Gearhead 2. You game?

 

[pyrrho12]

I wasn't really being serious there, just looking for scapegoats. I'd say that the honest reason why the components aren't very good so far is that I'm just getting used to the new plot system. Writing bits for a random plot generator is much different than writing a regular story, or even a regular game.

That reminds me. It's nearly about time we did another interview about the crazy fun new stuff in Gearhead 2. You game?

Sure.

 

[Gwendo]

I was thinking: if we studied the human thought process when making a small story or even a quest for a game OR if we studied writecraft techniques for making stories/plots, etc... Could we make a really good algoritm for a random plot generator?

 

[bryce777]

I was thinking: if we studied the human thought process when making a small story or even a quest for a game OR if we studied writecraft techniques for making stories/plots, etc... Could we make a really good algoritm for a random plot generator?

There is no way to emulate the human thought process and it is really not understood. A human brain has more computing power than you would possibly believe, as well; it is like how people say 'well why don't you simulate an artifical universe and then get the random dice rolls from that?'.

I think I might be trying to push it a little further than pyrrho - well, my thoughts about it...I have not done any real work yet. I think you can have a great deal of randomization, but the basic elements have to be written beforehand. You can probably get some very signifigant differences, though. Much more than random quests at certain points - I do believe with some effort the whole story arc can be largely randomized. It will have to work from material you actually write, but it can shoot out the combinations in much much different ways that could be very interesting.

 

[Section8]

I don't think plots that touch your emotions and that feel rewarding, could be randomly generated.

I agree. I don't think it will ever be possible to automate quests without having them sound repetitve after you've done a few. Why? Because the human brain is great at spotting patterns, its what helps us survive. Once you've seen one template quest, replacing the nouns, placenames and verbs to make it 'new and fresh' would not work.

I'm not convinced that it's impossible, and one major factor here is contrasting random generation against the average quality of scripted quests. I know for me personally there aren't too many game plots that are emotional and rewarding.

I think that random quests based upon an obvious template could be supplanted with enough effort, there just has to be a willingness to have major dynamic shifts within the game world, a greyscale of success and failure, and a wide variety of motivations from the NPC giving the quest.

[edit] Don't think I fully got my point across here, on re-reading. I think the biggest thing lacking in procedural quests is effective motivation through differentiated, less tangible rewards. Progressing along a linear quest line is rewarding by the very fact that it "unlocks" further quests and narrative, so it provides rewards beyond a bit of XP and a bit of material gain. So, with regard to my previous statement:

• Major Dynamic Shifts - if successful (or failed) player actions have a visible impact on the gameworld itself, then that is as much of a reward as narrative advancement. If I assassinate a druid, then to see decay and discord in the natural surrounds he previously protected is a great intangible reward, especially if that then triggers other world events (famine, plague, etc.)
  
• Grayscale Success and Failure - Most of the time, quests only have a single outcome - success. Sadly becoming rarer, are quest failures. But, what can make a single quest script shine, would be a much broader set of variable resolutions.
  
• Variation of Motivation - The lack of narrative hooks in most procedural quests is largely due to realisation of repetition. It's hard to care about someone's plight when it's so overtly presented as a generated template. Mount and Blade's "Kidnapped Daughter" quests are a good example. All it does is serve to unite merchant NPCs as a collective entity.
  
  And let's face it, most quests in RPGs are redressings of the same sort of ideas, and the stylistic wrapper represents a lot of the appeal. Just look at the reactions to Oblivion's "FedEx: Oils on Canvas" quest. It's a simple matter of go here, kill that, and yet it seems to be garnering undeserved praise for having a great deal of style in the place of substance.
  
  Now, I'm not suggesting that style over substance is in any way acceptable, but a simple redressing of a trite idea can easily improve and disguise repetition. Combine that with a more complex quest script than "oh noes the <monster> stole my <item>" and you're on track for some decent procedural quests.

 

[Araanor]

Of course a player *will* pick up patterns after a while, but that's still better than knowing just about everything on the second playthrough as far as replayability goes.

 

[Gwendo]

Maybe reading this site might be useful.

http://writingcraft.deep-magic.net/

 

[bryce777]

Of course a player *will* pick up patterns after a while, but that's still better than knowing just about everything on the second playthrough as far as replayability goes.

Again, that just depends on how much basic content you have. If you come up with 1000 different basic things that can be varied in dramatic ways, then you can effectively avoid this.

I think the actual generation of plot is a lot different that generating a cookie cutter quest; you are generating a lot of the motivation and goals...actual quests that make sense might be a totally different issue or in a freeform game something you figure out on your own.

 

[Saint_Proverbius]

Of course a player *will* pick up patterns after a while, but that's still better than knowing just about everything on the second playthrough as far as replayability goes.

Players pick up on patterns of things regardless of wether they're generated by a series of templates or crafted by a human with a cut and paste editor.

In fact, the templates don't have to be that complex to invoke an emotional reaction from the player. You don't really need thousands of them. In order to do that, the templates just need to make use of people and events that have already happenned to the player, much the same way TV series revisit old episodes and have re-occuring guest characters.

For example, say that your character, early in the game, saves some randomly generated kid from an evil gang - or encourages the kid to join the gang for his own good. Have the game state save the information for that kid as to his new status as well as the the reaction adjustments for the player from the kid and from the gang faction.

Later on, the player might be contracted to help the town clear out that gang because they are annoying the citizens of the town as a random event. Well, that randomly generated kid is associated with the gang and that information is saved, right? Well, you can just have the template plop that kid in to the scenario. If your player got him to join the gang, you might have to kill the kid - or the kid might escape the battle. If the kid dies, he's removed from the gamestate info. If he's not in the gang because you saved him, he might show up to help you fight them. Then save all that information.

Even later, the game might toss up an ambush scenario where the player is confronted with a consequence of his action. If he told the kid to join the gang, then fought the kid, that kid would be in the list of possible characters to base the ambush plot around rather than randomly generating a character for an ambush. Instead of a simple ambush, it becomes a matter of personal vendetta.

If you want emotional attachment from a player to things, you have to use them more than once. If the information is stored and the game knows how to retrieve that info, then it's just a matter of the templates knowing to check that info to insert a re-occuring character for those scenarios.