Quote:

Original post by Rosecroix I see where you're comming from, we're not really talking about the same thing. I can agree with everything you say and want, it seems interesting even if it wasn't my version of it. The reason I don't invite all those people is the same as one of your key points, the freedom to choose the place and time to play.

Good point. Touché.

Quote:

But since we're not talking about the same thing really anyway, what would drive your version? When you describe it I see for my minds eye a very advanced version of Carmen Sandiego, very much a thriller; a research-heavy detective story. I would belive constructing an engine to create one of those would be quite different from the soap opera emotion emphasized game I first envisioned you wanted, something much closer to Friends or Frazier than say CSI.

A detective story just makes for a simple example because there is a clear objective, there are clearly bad guys (even if we don't know who they are yet), and there is a clear protagonist. I merely chose that story to illustrate a point. The premise could just as easily have been that you're playing as Joey and you just kissed Rachel and need to find a time - and place - to break it to Ross (from the final season of Friends, IIRC). To improve your chances, you might confide in Chandler, but make him promise not to tell Ross - but since you didn't make him promise not to tell Monica, he tells her, and she comes to talk to you, Phoebe in tow, and "hilarity ensues."

Quote:

On one hand, the whole construct lends itself to te use of agents, and on the other hand, the many dramatic situation that must occur at a reliable and unrealisticly high rate immidatly calls for heavy scripting or algorthmic pruning of the starting conditions.

You're dead right about the unrealistically high right of dramatic interactions and tension builders, which means, as I alluded to above, that having a mathematical model for scene and overall tension is important, as it can help inform when to introduce new elements, and how.

Quote:

And how to choose a setting, and connect a setting to the engine? Not only must it realize captains order people around on boats, it must also be clear on what is kitch, what is gay, what is kosher, and what object/character is this as well as constructing sentences in eubonics.

This is an excellent question. One of the benefits, I think, of constraining a complete "experience" to a couple of hours and a fairly compact circle of characters is that it limits the boundaries of interaction. You could write hundreds of premises without having to utilize all of the above elements, and because the story is inscribed by the player actions and interactions, it would take a good while to get stale. By carefully selecting your cast of characters, you could avoid many of the ambiguous situations, at least in the first revision of this technology/

Quote:

It is a fascinating idea, but it gives me an overwhelming impression. Even if you don't got a working idea yet, what have you thought about yourself?

It sure is overwhelming! I actually took a class a few years back on Artificial Intelligence and Narrative Art, but we didn't get a whole lot done. I did get introduced to a few concepts and technologies here and there, such as Bayesian Inference, and I think it could be valuable in navigating the web of social interrelationships between characters and directing traffic, as it were, toward the player. But I'm trying to get a better grasp on the problem first, and think about implementation later. Once you have a concrete design and algorithm in place, the rest is easy. The hard part is performing the problem domain analysis to understand the problem as completely as possible, and then doing the creative thinking necessary to come up with an algorithm.

Quote:

Original post by Madster heck, i've even considered doing that... making a tv series out of demos of good multiplayer gaming ^_^ ok i'll shush now.

That just brings painful memories of Arena on G4TechTV to mind. I know you meant nothing like that, but...

Quote:

i had considered two types of NPC, primary and secondary. Primaries cannot be influenced by the consequences of the player (the player himself would be a primary... see? you can't change the player, only he can change himself), and secondaries can change allegiances and such. There should be a minimum amount of primaries, usually two.

I like the primary/secondary dichotomy (as you can see, I borrowed it above)! There is traditionally the protagonist and antagonist as primary characters, with all others being secondaries. In film noir, the femme fatale often serves as a third primary, usually forming a love/infatuation triangle of sorts. Nice idea.

Quote:

About NPCs and information: each NPC should have bits of knowledge, and they should be propagated according to allegiance at the end of each stage/scene (this way you have the opportunity to stop someone from talking by threatening them, etc before they leave). Then the allegiances would be recalculated with this new knowledge. There should also be some avaiability rules so that characters cannot always recieve the info at the end of a scene, and the NPC queues it, probably along with an urgency rating, to communicate as soon as the other NPC becomes avaiable again. If we go the symbolic language road, this is highly workable. how does this sound to you, Oluseyi?

It sounds great. It's a direct analogue to physical simulation systems such as ODE, so the macro architecture is a proven one for dynamic systems. (I mention this because I was writing some simple physical simulation code in the last few days and then I decided to try ODE yesterday, and because it provides a referrent for others).

Quote:

Primaries do not change because they drive the plot. As a counterexample: that guy in the train that gets down, he would be a primary. If he could change, he could decide he doesn't want to see you anymore, and the game becomes unsolvable. Primaries are there to drive the plot, think of em as markers of what the story is about. In a mystery, the plot could revolve around finding stuff about the primaries. heck, i guess you could define meta-primaries (dummy characters like... a big bag of gold in the treasure room) to drive other kinds of plots.

Actually, the guy who gets off the train is a secondary. He exists merely to facilitate the advancement of the plot. His actual interaction with you in the context of the narrative is limited. It's like every murdered teacher/master/buddy/partner/spouse in every trashy '80s action flick. But, yeah, that sounds like a fairly flexible system. Also, characters could move from primary to secondary focus between "episodes" or "chapters," or even between parallel narrative threads (we see this all the time in ensemble cast television shows).

[Edited by - Oluseyi on December 28, 2004 9:46:46 AM]

Quote:

Original post by seanw You should try out some recent text adventures from the IF Competition (the 2004 results came out a couple of months ago).

My biggest challenges getting into IF have been 1) the steep learning curve (which instructs us that anything that results from these discussions should be pick-up-and-play simple); and 2) the total lack of graphics. I know it's all about fiction, but that doesn't mean it can't have pretty background images to quickly communicate the environment, or portraits of the characters.

Quote:

It would be nice to see how text-adventures could be improved with current technology, and I don't mean adding graphics.

Hear, hear.

Quote:

Original post by Oluseyi Clearly, we have a miscommunication of what that term implies. To me, story-driven does not mean that there is a fixed narrative arc which the player explores or attempts to complete. To me, that can also mean that the "story," as it were, is generated by the player's actions. There is a premise, and when they game ends there is a conclusion. The former is fixed, the latter is open, and the sequence of events that bridge the gap is entirely player-dependent. It is this chain from premise through multiple events to conclusion that is a "story." I don't want a user to play through a story; I want her to create a story. ---- This is not a video game version of "Choose Your Own Adventure" books.

i believe we see the term the same way, but our aims are slightly different. I don't want the player to create a story, i want them to play trough a pre-written one in the most natural-feeling way possible, like one plays an action game and doesn't notice they're all corridors because they can go back and forth and do as they please. so in the end i want it to be like "Choose Your Own Adventure" but with content behaviourally spawned (w00t big words)
why? it's much easier to avoid deadlocks, and you need much less content. in my view, the set goals must be met, trough any means avaiable (the more the merrier) and following your idea, the epilogue would vary depending on how the objective was met. Someone mentioned a replay. This is a VERY GOOD idea (check it out, i managed to make friends with X and have him betray Z!!)

Quote:

Original post by Oluseyi Quote: Original post by Madster heck, i've even considered doing that... making a tv series out of demos of good multiplayer gaming ^_^ ok i'll shush now. That just brings painful memories of Arena on G4TechTV to mind. I know you meant nothing like that, but...

ew. i don't get G4, but i can imagine. I was thinking more along the lines of machinima. Good machinima. like Red Vs Blue, but maybe with a serious(er?) tone.

Quote:

Actually, the guy who gets off the train is a secondary. He exists merely to facilitate the advancement of the plot. His actual interaction with you in the context of the narrative is limited.

yes, i probably chose terms badly... in the game scheme i've been thinking about, a primary doesn't necesarily show up a lot (i mentioned a table as a primary in a previous post). What makes it a primary is that it doesn't change along the story (unless scripted to do so), and serves as an anchor for the plot. In this case you must go looking for your friend, so he anchors the story and becomes a primary.

Is there any reason why a primary would become secondary? maybe between chapters?

PS: good thread =D

besides the last question in my previous post, i have some more. I'll reiterate that one so it isn't forgotten ;)

-Is there a reason a primary (anchor NPC) would become a secondary (free NPC) or the other way around? are there restrictions to such change? i can imagine one would only change from primary to secondary but not backwards... since the secondary changes and you don't know beforehand what he'll be like before the end of the chapter, thus precluding him from anchoring the story.

-We've liberally used the idea of symbolic icons and responses that cause psichologycal effects on the NPCs, and that they react based on their traits and allegiances. I can easily imagine a way to represent allegiances (single float for each pair) but how would we represent traits? and how would we calculate the effects of each dialog? how would the NPC decide their replies and how would that affect our own player? how would you measure the internal state of the NPCs? would the player have an internal state?

-Oluseyi, aiming to your idea of a game, since the conclusion is open: how would you decide when the game has been finished? do you score the player?

-Tension control was mentioned. This is akin to scripting the scenes dinamically in order to maximize fun (heh... Your job in this company is to maximize fun!). With a behavioural narrative, how could this be achieved? i can imagine having characters 'stumble' upon the player (interruptions as seen in my thread) when the tension meter goes low (ratings!) but forcing interactions seems... well, forced, considering the world model. Also, how would Tension be measured?

Oh so many questions. let's break new grounds!

edit: forgot stuff. Again.

Quote:

Original post by Madster i believe we see the term the same way, but our aims are slightly different. I don't want the player to create a story, i want them to play trough a pre-written one in the most natural-feeling way possible, like one plays an action game and doesn't notice they're all corridors because they can go back and forth and do as they please. so in the end i want it to be like "Choose Your Own Adventure" but with content behaviourally spawned (w00t big words)

Well, that's an admirable goal. I'm pretty sure that there is significant overlap in our objectives, even though our end points are quite different.

Quote:

why? it's much easier to avoid deadlocks, and you need much less content.

Actually, I think the amount of necessary content is simply a function of the size of the story world, the diegesis. You can have both fixed and dynamic, arbitrary narratives within worlds of the same size. Look at the variability in diegesis in film for comparison: some movies involve dozens of locations in several countries while others basically involve two adjacent apartments and the street nearby (We Don't Live Here Anymore) or two rooms in adjacent apartments, the street below and a hotel room (In the Mood for Love). Night of the Living Dead was all but set in a single house.

Quote:

Someone mentioned a replay. This is a VERY GOOD idea (check it out, i managed to make friends with X and have him betray Z!!)

*bows graciously* [smile]

Quote:

PS: good thread =D

Only because there have been great participants with great insights.

Quote:

Original post by Madster -Is there a reason a primary (anchor NPC) would become a secondary (free NPC) or the other way around? are there restrictions to such change? i can imagine one would only change from primary to secondary but not backwards... since the secondary changes and you don't know beforehand what he'll be like before the end of the chapter, thus precluding him from anchoring the story.

Well, a demotion or promotion could occur as part of a plot twist, and, in fact, those sorts of changes are some of the best moments in cinematic history. Janet Leigh's character in Psycho and Kevin Spacey's Kayzer Soze in The Usual Suspects are just two examples, although the former goes from primary to secondary while the latter goes from primary to "Mythological Proportion." This usually is the director employing misdirection to suggest that the narrative heads in one direction, then pulling a switcheroo and heading off in another.

The other reason for actor shifts from primary to secondary is narrative arcs. In long-running narratives such as television shows with large casts, not every character can play a central role in every episode. Consequently, pertinent to the issues of a particular episode, a certain subset of the cast (sometimes as few as a single character) becomes primary while others become secondary. Since our objectives here including both dramatic gameplay and episodic gameplay, a soap opera or primetime drama type game is not out of the question, where each play session corrals a subset of the characters, establishes a premise and sets the protagonist/Actor (user) off on a quest to resolve all tensions.

I can see it now... General Hospital for PlayStation 3... :P

Quote:

-We've liberally used the idea of symbolic icons and responses that cause psychologycal effects on the NPCs, and that they react based on their traits and allegiances. I can easily imagine a way to represent allegiances (single float for each pair) but how would we represent traits? and how would we calculate the effects of each dialog? how would the NPC decide their replies and how would that affect our own player? how would you measure the internal state of the NPCs? would the player have an internal state?

[Edit: Oops. Skipped this question initially.]

Personality traits are, on a very basic level, just tendencies toward a certain outcome. For ten tries faced with a certain situation/choice, how many times would the character respond in a particular way? In implementation terms, it's a threshold for a bounded random number generation (if we only allow numbers in the range [A, B] and the trait tendency is X%, then every "throw" that falls between (A - B) ± (X / 2) will yield that trait outcome).

Quote:

-Oluseyi, aiming to your idea of a game, since the conclusion is open: how would you decide when the game has been finished? do you score the player?

The current game - and this is an important distinction, because this technology could lead to games that can create a huge number of narratives, obviously varying on themes - is completed when all tensions have been resolved and there are no further tensions to add. A pregnant woman is a source of tension (will she have the baby? will the baby be healthy? etc, etc). Catching your girlfriend kissing the rugged, good looking doctor from Australia is a source of tension (why was she kissing him? will you take her back? does she want you to take her back? is it over? is she going to be seeing him, or was she just trying to get out of your relationship?) If we consider each "tension generator" as a collection of "tension vectors" (each of the questions raised by the event), then tensions have been resolved when all the tension vectors have been erase - when all the questions have been answered. How those questions are answered comprises the "story," which is what the user can then export and share with friends.

Quote:

-Tension control was mentioned. This is akin to scripting the scenes dynamically in order to maximize fun (heh... Your job in this company is to maximize fun!). With a behavioural narrative, how could this be achieved? i can imagine having characters 'stumble' upon the player (interruptions as seen in my thread) when the tension meter goes low (ratings!) but forcing interactions seems... well, forced, considering the world model. Also, how would Tension be measured?

There are elegant ways to introduce tension. Say the play is lagging and dragging the narrative down, we could initiate an event and then select an appropriate (and semi-random) means of communicating awareness of that event to the user. There could be a newspaperboy on the corner shouting "EXTRA! EXTRA! Man found dead wearing Paratrooper uniform!", which would draw the user toward him. If the situation gets dire enough, we could forcibly inject interactions by having "bad guys" come looking for the user because "they heard he was snooping around, looking into things" and their boss "wants to talk to him."

The available injection vectors will have to be specified as part of the diegesis by the writer(s) and designer(s), but the runtime narrative agent will then employ them to help guide the player toward meaningful interaction. Another aspect that deserves mention, however, is that the narrative agent will also need to forestall excess tension, perhaps by relocating a character to postpone interaction with the user ("I came by earlier." "I was in Philly. My Aunt Lois just died." "I'm sorry to hear that." "Whaddayagonnado?" "To business?" "Yeah, whatever.")

Quote:

Original post by OluseyiMy biggest challenges getting into IF have been 1) the steep learning curve (which instructs us that anything that results from these discussions should be pick-up-and-play simple);

I think a speech interface would be amazing though. Everybody knows how to read and talk so you don't have to learn anything except be aware that you have to say things in a simple way. Once the game is running, you'd never have to touch a mouse or a keyboard. Old text adventures use fairly primitive pattern matching to work out what sentences mean, using a real natural language parser could provide a great improvement. In comparison, for a graphical game you'd need a horrible GUI button, a pull-down menu, a keyboard button etc. for each type of action you'd want to do which really is too complex for most people.

As long as you kept the descriptions short and snappy, I think you could quickly draw people in if you accounted for most user actions. Quite a nice part of developing text-adventures is that you could ask your testers to save the script of everything they typed so you could add in new responses/events for actions some users tried but got a generic failure for.

Quote:

and 2) the total lack of graphics. I know it's all about fiction, but that doesn't mean it can't have pretty background images to quickly communicate the environment, or portraits of the characters.

Hmm. There's quite a divide with that issue. I never really liked illustrations much because they tend to go against the mental image you build up, which I find more powerful given good writing. It's easy enough to do though. Interactive music would be a nice feature, so it would change based on current events (e.g. relaxed when nothing is going on, more fast paced when you discover something shocking).

Although the lack of graphics does put people of, it does give the game designer infinitely more freedom. For example, you can set things on fire, blow things up, make somebody angry etc. just by writing the text for it. You don't need to spend millions creating models, textures and animations. As you could spend the time you saved here adding dynamic game content, the game would be vastly more interactive than anything you could make with a rigid GUI interface. For instance, in certain IF games, you can order NPCs to do things for you, tell NPCs things and ask NPCs things. You can't do this in normal graphical adventure games because the interface would be too clunky for it. Of course, there's always the danger that there's so much freedom the user won't know what to do, but it's the job of the game designer to make sure this doesn't happen.

I've pretty much saying I reckon you could achieve your goals with a modernised text-adventure which had beefed up natural language recognition (so you could phrase sentences in quite complex ways), having the vast majority of user actions accounted for by the game and using speech for input (anyone who can talk and know not to put more than one action per sentence could use it). Maybe that's not what you want, but I'd love to see what it would be like. :)

first, quick reply: seanw, what you're dreaming about can be apparently found in www.skotos.net (except for the speech recognition thing, that aint gonna happen in at least 8 years or more) , beware the hosted games, look for those labeled prose

Oluseyi:
To avoid further confusion i'm renaming my NPC classes as Anchor(primary) and Free(secondary).

-About switching NPC classes:
So, in a soapie you'd have one episode about a group of people, i guess you could turn one (or more?) of those into an Anchor NPC for the duration of the mini-arch, freezing his allegiances and traits until the conclusion, and then unfreezing again to let it evolve with the rest of the world. kind of like a mini-quest.
then again for your case you don't really need an anchor at all in the game. I guess you could do without all the hassle altogether. I'd do the freeze-unfreeze thing instead, like say..... Jim hates your guts and wants to kill your brother. Stop him. (er.. or something.. i can think of more dynamic mini-quests than those requiring anchors)

-About traits and feelings
So traits would be represented by a treshold to a certain concept. How would those concepts be represented? i may sound picky, but to make sense of a design i like to nail everything down to the last byte. My idea was more around a pet concept i found some time ago in the net, where 3 meters represent a feeling, and you can attach them to current state, long term state, other people and even objects. I havent thought about how these values would affect each other though.
Have you thought about concept representation? would there be a list of existing concepts in the whole game / episode and each character would have a threshold to it?

-About tension
i see a problem here. Tension is used to resolve the game, yet tension is generated again whenever it's lacking. This leads to an unsolvable game. Am i wrong? how? also, based on the data representation in the previous question, how do you measure tension exactly?

I'm sleepy, so I'll get back to this later, but I just wanted to quickly respond to this:

Quote:

Original post by Madster -About tension i see a problem here. Tension is used to resolve the game, yet tension is generated again whenever it's lacking. This leads to an unsolvable game. Am i wrong? how?

Tension is only generated when lacking as long as the central tension remains unresolved. The central tension is the premise that sets off the entire narrative, and is chosen by the designer/writer.

Quote:

also, based on the data representation in the previous question, how do you measure tension exactly?

You measure it as a collection of numeric composites. The key is to think of tension in completely different terms from a specification perspective. "Tension" is a familiar term for talking about a relatively abstract concept, but it doesn't mean much in rigid mathematical analysis. For our purposes, "tension" would be an amalgam of danger (distance from a threatening entity, intensity of threat posed by entity), desire (reciprocated affection, or the lack thereof) and sundry other quantities that will be specified as components of the tension by the writer/designer.

I'll be back tomorrow. I'm hungry and groggy right now.