next »

[gimymblert]

@toast
Sorry to burst your bubble but what you are describing is flavor and ornaments. This is totally parallel to strong structure, without structure they come out as shallow (typical video game lore verbosity) but structure alone is dry. Experience writer find the structure first from hints of flavor and elaborate iteratively from it. Good design is always about knowing how to kill the darlings, you are still to enamoured with them to see through imo.

[gimymblert]

http://www.dramatica.com/theory/theory_book/dtb.html

Somewhere on the site there is a pdf version, master this theory before thinking about flavor. There is also a demo software to experience interactively with.

[1982]

Good design is always about knowing how to kill the darlings

Like little Chinese girls. I wonder how the world would be if we didn't kill our darlings.

[Graham-]

@zigzag

No, that's related. You described it pretty well. I like the analogy with the kid with aspergers. I want all of the things that you described to be true about the game.

In fact, the player starts off as an ethereal entity, watching the events of a world. They have no idea how to play. Through the pressing of arbitrary buttons, and watching the consequences, they slowly understand how to affect things, change them, guiding beings in constructive ways or not.

Over time their ability to impact what is happening develops, becomes more complex and deep, and personal, and permanent. They become more "permanent." At some point their decisions materialize as a consistent entity, that is them. Their developing being is a reflection of their decisions. (like the Fable good/evil system, but not 1-dimensional).

Then they can continue building their relationships, abilities, and skill in communicating. The player is a "new arrival" in a strange world, I suppose....




@Gimmy

Hehehe? No I'm not! But that's okay.... We can talk games forever and I'd eventually convince you that its a real story line and everything. For example, I've written the core constructs of every major character in television that inspire me - well.... I've done it for a whole lot. Anyway, whether or not I'm talking about "ornaments" is a reflection of my ability to identify what makes those character's tick.

What you're saying is similar to identifying the relationship I have with my parents and saying, "I bet that's really hollow...." Maybe! It depends! You probably have to meet them first and see me with them.

Your link. Yeah, yeah. I have an education in story theory too! I've read enough books on film structure and plot structure so that I can map out a given work blindly and fluently. A few times now you've taken my "philosophy" to mean "not really real." That's fair. But remember... I'm the math degree. I'd never lose sight of how something is programmed. I would also never waste time in theory that doesn't produce a good game.

"Structure" and character are indivisible parts. You want to talk about structure and character as these separate things. It's like AI guys - again with the AI ... but AI programming relies on the same theory as all this other stuff.... They want a story line for this side-quest and one for that one, and one for that one over there. Then they want one for the side-kick, then one for each enemy, then one that controls all of them in some way to create dramatic structure, at least in a small way. Then they want a pathfinding one, and holy FUCK TON of individual rules that link each of your decisions to the behaviours of each NPC. It is this isolation of systems that produces terrible AI... design... etc.

Structure is meaningless without good character. You could say good character isn't a story without good structure. I'm not denying either of these things. I think you're saying the same thing, that it is all connected (right?). We just happen to be talking about "flavour," because structure is the easy part to program. Humanity is the hard part.

A lot of structure has to do with "things a character does," or "changes within them." For example, Act 1 is the place where we are introduced to the critical elements that make a character interesting and someone who has a goal/weakness that will play critical roles for the rest of the story. Act 2 is development, tension, resolution. Act 3 is results and loose-ends, maybe with another mini-climax or two - James Bond often puts a second, smaller, climax into Act 3.

All that is textbook. You can drill down super deep. That stuff may be as challenging to understand as anything else, but it's still well primed to be programmed. The parts where it becomes challenging are the parts where it interacts with character.

How does one determine the relevant elements of a character to introduce? What if circumstances change slightly, or a "quest line" is being formed dynamically? The relevant "introductory" details change. Maybe the player has seen some of these details already, from a previous "quest." How do we compare what they know to what they need to know, which in itself is abstract?

You need a way to take a character and slice them up according to whatever parameters you have. Maybe the characters are in a particular situation, let's say in some environment and participating in a battle against an enemy (alongside the player). Maybe they each have a "mood" that is a reflection of how the grind has been going over the last hour. Ok, now we've decided is the time for introductory details. We have new constraints. How should the NPC demonstrate the necessary knowledge to suit both his mood and environment? How should he do so in a way that blends with his current behaviour which is also guided by the goal to vanquish the enemy in whatever way he is currently pursuing? Is this even the right time for everything? Maybe different details should come out now because they better suit the circumstances?

Structure is easy. Applying structure to a character - an abstract thing - is hard. So I'm talking about the deconstruction of characters. Yes, they need a structure to play themselves out and all that. I'm not denying the value of structure. I'm just saying... I don't know what I'm saying. Structure and character are both important.

Also, thanks for the link. Your links are always good. This one is 1 under par.

When I work the story I consider structure and character simultaneously. When I talk about AI, I only care about the character, because that's the hard part.

When I develop the game structure and character will evolve together. There's no argument there. Even when I have 1-d personalities running around there will a core game loop, "3 acts" (you know what I mean...) and whatever degree of detail I need.

There's no confusion here. I'm just really, super, accustomed to AI design. So I hone in on the juicy bits first. Applying structure inside an AI is straight-forward. It is important, but straight forward. I don't need to talk about how to do that, unless you want to. Applying character to AI is the new part. That part no one does well. It's the reason why when structure exists in a dynamic story-line game it doesn't matter, because the characters and conflicts are boring.

In other words, structure is a thing that the world knows about. Proc-gen stories fail from an inability to deconstruct the abstract in meaningful ways, not the logical. There will be structure all up ins my game. Don't worry about that.

Good link.


edit:

You could say anything is about knowing when to cut an idea. Goodness comes from producing ideas. Greatness comes from cutting them.
  - Me


@1982

A lot worse. ... oh, you mean real people. Yeah, China.

[gimymblert]

What I'm saying is that you tell a lot of things but say you will get there explain their importance yet fail to demonstrate how.

Also that dramatica theory really look into it, it's not random link toss because you don't know theory. All your questions about character and how to bridge structure, process and flavor are answered.
also one big thing is to treat structure not as big inflexible how to, but a multi rez lend. 3 acts structure goes down to animation (tension and release) and is generally a very nested structure in any work (superposition of various arc, using it only as an overarching structure is loosing its efficiency.

Therefore if you can link function and flavor, more power to you, but you can't do it without a good inventory of all functions. Simple character map simple function in a singular story arc, like who is the hero, the opponent, the mentor. Complex story mixed character function, shift them over time, superpose meaning of story arc in one action, etc... Also strong structure give genre, people like genre which is a proof that they can like structure for itself beyond the character, also same character can be use in totally different setting and genre while remaining consistent, this highlight there might some independence from each other.

[Graham-]

Hunh?

I'm not trying to fight you. I'm just talking about the things that are on my mind. I don't know what you want explained.... ()

A game is a million pieces. I am talking about 1/100 of those. I'm not saying that the others aren't critical. I'm happy to explain some other piece you'd like me to, you just have to ask. Saying, "show me how it really works" is super vague. I'm a nice person, and I get something out of sharing, but I don't understand what it is that you're referring to.

Building a dynamic character is a big process. It is really fucking big. I think maybe one of the issues designers have is underestimating how big of a deal design theory is in general.... I'm not saying that's you. I'm just saying that in general.

Bridging structure and character aren't questions I have. I'm just doing the work. A character needs to be deconstructed into constituent parts. You need to identify each of the pieces, the same way you make a blueprint of a building. You need to see everything, in incredible detail, and how it all fits together. Writers normally don't do this because an intuitive understanding is enough. They can build such a "blueprint" in their minds by considering all the various properties their characters have, such as how they'd react to a situation they have planned in the story, whether they eventually include that situation in the story or not.

With AI that strategy doesn't work. Plugging structure into characters is trivial. Defining a character is what's hard. You can't rely on an intuition of "what a character is" when you're writing a dynamic story, because you have to teach a computer to understand the same things. The computer needs the same understanding of a character as a writer would, so the character has to be deconstructed in a technical way.

The same thing is true for dynamic designs. If you want a machine to construct a level, for example, it would need a designer-level understanding of how all its available level-constructs relate to one another in creating an experience for the player.

Hunh. I wasn't dissing your link. "1 under par" means good. In golf, a birdie is a very good thing. It was a compliment! .

I will read any good source on story structure. Most books on "structure" that I read are written from a writer's perspective, so they're filled with writer-like details. The link you gave me is a little more technical, so it rounds out my library well.

I promise, I will read that thing through 3 times before I release. I have an ever-growing list of structural analysis resources.

Yeah, 3-act structure is a general template. It tessellates down. The general theory is this: introduce viewer to elements, show how elements play out, explore results and recap what happened. You can apply that template to any attempt to communicate to find flaws.

If you want a "how" or something, that's personally interesting, I'll be happy to explain it. I'm just not sure which how's you're looking for. A lot has to be implied in writing. I mean, I never talk about decision trees. A decision tree isn't powerful enough to handle the kinds of ideas I'm talking about. You can use it for very simple characters, but they become unwieldy fast. You need to introduce whole new structures to manage characters that are as rich as those in stories and as dynamic as we'd like them to be in games.

But I haven't talked about that because no one has asked. I haven't talked about threading strategies either, or ways to reduce load times, or ways to animate characters procedurally. There's a lot of patterns in development that you have to follow to ensure that you can develop fluid animations that reflect what you want, and that suit the characters and mechanics that you are developing. Splitting that workload among a team requires an even more articulate process. But I haven't talked about those things... just because. No reason. I mean, I'll probably talk about those kinds of things later... probably when I start doing a lot of animation. But if anyone's curious I'll explain.

There's some hidden "how" that you've hinted at in this discussion and our other one (about design). I know how to take the structure I've presented and turn it into the results we're both talking about, but there are 50 other steps involved, and I don't which ones I've implied that I need to explain. Sometimes very simple changes in definition can make a lot of things clear that weren't before.

I can produce a game that makes people feel as expressive as they do in Mario, or Minecraft, and shift it to other kinds of feelings given what I've said... then expand on it to give the player both more guidance and freedom. It would take lots of work; that's why my game is taking so long.

If I explain 1/10 of the process of something, and you say, "but how does it really" work, all that means to me is that somewhere in the other 9/10ths is something that is unclear. But I don't know where that location is without an example or something. So I just guess and talk around for the exercise, to keep my blood flowing, and I enjoy it. But it's not the best strategy for becoming clear. If you want me to take my theory and apply it to Zelda or something, while maintaining the series' modern level of accessibility, I can do that, if you point to the few areas where it seems to not work. But you've got to point to those areas for me to do that, of course if you're interested. I can't tell. I'm happy either way. We're all here to talk about ourselves anyway .

I know I haven't explained most areas. Writing good characters is a lot of work. Writing a dynamic story is more work than writing a normal story. You have to do all these extra things. A bunch of posts barely explains anything. I need to know what you know, and what you're looking for, to understand what information might be interesting or useful.

also one big thing is to treat structure not as big inflexible how to, but a multi rez lend. 3 acts structure goes down to animation (tension and release) and is generally a very nested structure in any work (superposition of various arc, using it only as an overarching structure is loosing its efficiency.

Yeah, I know. Studying raw structure is better seen as a way to know how to apply it everywhere. That's how theory works: flexibility. We're on the same page on that one.

Why else would I be so crazy about theory :|. It's the argument that prioritizes example that sees theory as hard structure.

Therefore if you can link function and flavor, more power to you, but you can't do it without a good inventory of all functions. Simple character map simple function in a singular story arc, like who is the hero, the opponent, the mentor. Complex story mixed character function, shift them over time, superpose meaning of story arc in one action, etc... Also strong structure give genre, people like genre which is a proof that they can like structure for itself beyond the character, also same character can be use in totally different setting and genre while remaining consistent, this highlight there might some independence from each other.

Mmm. I see what you're saying. This is what we were talking about way before....

Inventory of good functions. Yes. You need that. I do not deny.

I'll give you an example. Say you want to be a sculptor. Say you want to create the most beautiful and evocative sculptures possible. But let's also say you live in a time when sculpturing is a new art form and people don't really do it. There are no good tools. Now let's forget that a lot of sculptors just use a hammer and chisel and a lot of patience and skill.... Assume you want a box of tools, just for the analogy.

Ok. If you want to design good tools you'll need a working knowledge of sculpting itself. You need a box of tools to create the world's best sculptures, because that is our assumption. But you also need to know how to sculpt to know what tools would be the most useful. You need the knowledge of sculpting well to design the tools, the tools to produce well made sculptures.... It's a catch-22 common in programming.

The tools and skill-set have to develop together. As one improves it opens new opportunities to improve the other. I do not deny this.

All I am saying is that designing the tools in total isolation, without considering how they might be usefully applied to an actual product, isn't the best idea. A sculptor designing all the tools before working on actually sculpting is headed for a weak product.

I am reminded of the "step ladder" process that is way too ubiquitous in big-business game development. You probably know it. It's like, "draft outline, scope required resources, produce outline, research implementation requirements, create implementation plan, implement, user-test, polish, real-world test, release, support." Something like that. Nice easy steps, 1 through 8 or whatever. Start with a plan, produce a product, polish and ship. This obviously isn't the best plan for games, which need to be super iterative.

Theory design is the same way. You have to cycle around, developing your tools alongside the projected games that they could be used to build.

A lot of design theory suffers from asking questions that are too small. Design theorists find ways to structurally analyze the aspect of games that play small roles in a game's value to a player. I still have yet to see a single proper deconstruction of Mario. Jon Blow is the only guy I've seen talk about orthogonality... which is insanely important in good software design, and game design, both for creating deep mechanics and choosing initial mechanics (at an early stage in development) that have a very high likelihood of being malleable in the ways that you'd want them to be, so that they can hit the targets you'd like to meet but haven't defined yet, because you aren't that far along.

Orthogonality is also very important to designing characters that are very likely to play well off of one another. If you fill out a structure with characters, then try to manipulate it, you might be headed for trouble. ... though I don't know if that's part of the discussion.

...

I talk about all the "philosophy" stuff because I want tools that are most likely to be useful. If you talk about structure without considering why Mario is engaging, you'll probably produce a structure that will help only a little in reconstructing the value of Mario. If you consider why Mario is Mario while developing your (structural) analysis, whatever structure you produce will be of far more value.

If you create a bunch of tools, then consider "all the bullshit" later, you'll produce sound tools that will help you a good amount. But it will still be up to the developer to do the meaningful work. If you create a bunch of tools while considering "all the bullshit," then you'll produce sounds tools that will do a lot of the meaningful work themselves, freeing the developer to work on even more meaningful ideas.

The catch is that an AI can only understand the structure. All it can use are tools. If there isn't a defined relationship between one tool and its use, then the AI can't use it. An AI is a tool; it can only use tools. If your tools are designed in such a way so that much of the burden of creative creation is on the manipulator, then they will be very useful to a human, not to an AI.

The fact that good AIs don't exist is strong proof that most of our design theory is far removed from what makes a game interesting. Once structure is made clear, AI becomes a natural follow-through. You can use one to judge the other. In other words, if you can explain a thing to a computer, then by definition you understand it at a mathematical level.

I do not deny the value of strong structure.

Yes, there is some independence between "structure" and "contents."

I still think we're mostly just talking about semantics.

What sort of tools are you looking to develop?


---
edit:

I think we'll both easily agree that a story with lots of structure and 0 anything else is totally uninteresting.

However, structure is valuable because a person can use it to create something excellent. The structure doesn't write a story for the writer, but it provides a useful box of tools for the writer to use intelligently to write an excellent story themselves.

I value the pursuit of structure, just for structure's sake. There is a lot of value in structure all by itself.

However. When you want to start generating levels, and things like that - human things - the AI needs to understand how to apply structure, and manipulate the "non-structural" contents. If it doesn't it will generate stories that are just as bland as the writer "filling out" the structural formula.

I care about tools too. In our other discussion I would've been happy to just discuss tools. But you wanted to pursue why I used so much abstract stuff instead. This is the reason. Generative-anything requires it. But if you wanted to just talk about tools, I would've just talked about tools. I need tools just as much as you do. I just approach each one by thinking, "how would this tool be useful?" So I explain everything in a really abstract tongue. The reason I wasn't giving you what you were looking for was because I didn't understand what that was, not because I was defending some strange position. Give me a result, I'll give you a solution, otherwise I just talk about whatever is on my mind.

Also, as an aside, tools that put creative power in the hands of an AI, are also easier to distribute. If I take the structure of a building, I can divide the workload of its constructions among workers, because the plan has all the necessary details. If I do the same thing with a game, I am fucked, because a design implies so much interpretation. If there was a way to divide a game into pieces so you could isolate the dependencies between tasks, you could distribute them among people in such a way that didn't restrict creative freedom. The kind of theory that could do that is the same kind that allows an AI to be creative.

Just saying, because I'm like that.


edit edit:

I think this might be a case of artist and programmer can't communicate, especially over internet. Semantics, semantics.

[JWK5]

[Graham-]

That made me a lot happier than I'd like to admit. I actually laughed, out loud, like a real person. ... I feel a little guilty.

Gimmy, no hard feelings.

[gimymblert]

Yo toast, sorry if I sound like rude, I tried to get my point fast as i'm on borrowed phone and I have no time, so the form of the message might suffer a little.

However there is many point where I would disagree since it look like an ideal vision of creation, the so called romantic take of omg emotion is so volatile! The truth is that making good work is pretty easy, just follow all the discovered rules, that's why you can have it as a work, actually some people are excellent at this while not liking the art/genre they work with. However what's difficult is doing something new OR something that express a particular feeling that touch your sensibilities, the unique snowflake syndrome.

I believe those difficulties is more a byproduct of the difficulties of knowing ourself and the obscure part we don't necessarily want to get into, ie ambivalent feeling, rather than a distinct element of the art. Emotion is consistently moving, what's touching at one moment of our lifes might became disgusting the next, a simple story with rigid character can impact us because it's just the right moment for you (or a generation), regarding that there will never be a right good ay to do thing, I think that's what you called expression that can be a zillion things, it's the illusion born from the dance of feelings.

But all those things are still down to a bunch of word/structure function, all matter in the world are still down to a bunch of atoms. Therefore while chasing this volatile expression is okay, it will always be there as long there is creator, good tools or bad tools, after all platformer can still convey complex emotion (see rohrer's work), the problem is that we don't understand them enough to replicate them, at least good enough than even a computer can create a subset of them. Good tools/structure is not thinking about the math or logic but to find a way to have a container, just like grammar is for all work from shakespearean to troll insult on internet, it's not about good work, it's about covering the whole range of possible expression. If you can do that, then someone will figure out to make taxi driver base on a theory on montage.

When I employ the word expression this is what I mean, the huge range of work, from good to bad, the functional mechanics that bring the big bang of the universe, I want dna (four letter) that brings all the life form from a same base. You can't chase both, because you will limit yourself, I want unlimited expression, not just good (which generally mean What I like).

I also working on procedural stories, and I learned I get better result by breaking it down to component and studying their relationship: for exemple I have character function, stake, progression arc, storytelling, hammerspace, theme. This allowed me to find that I could generalize all plot to only 3 theme components: survival, social, acheivement, basically a summary of maslow ladder because it contain a description of all human need which give in return stories stake components.

For example I could take the story exemple you gave and feed it into my system and explain all the range of possible thing that can happen, where it breaks and what you can use it to create effect.

Let's look for exemple on the premise of romeo and juliet, purely from a structural point. It has 4 components:

Romeo is a montaigu
Juliet is a capulet
montaigu and capulet hate each other
romeo and juliet love each other.

It's obvious where the stake is even if it's implicit, there a paradox from hate/love, the information add up to a contradiction, this logical leap is called a conceptual breakthrough. I know I can present this in any order because you need all piece for the breakthrough to happen, there is no tragedy if one piece is missing, we can relate to this structure without knowing the character because it speak to our humanity (forbidden love, which is a mixed of social (sub theme is affection) and survival (sub theme is threat) theme). Therefore the use of flavor or ornament need only to highlight or understate those element through time to create the emotional roller coaster, talent will only amplify the ride.

Now we have a contradiction, a conflict at the heart of the structure, because the component are known we can see what are the outcome, basically for the contradiction to end there must be a change in the components states, it can be that one character cease to belong a group, that the groups stop hating each other, or that character cease to love each other, or you can introduce a new element that add a new stake on top of the existing. If you look at the story of romeo and juliet, you can see that the author choose the change in character groups, by faking death, juliet is effectively removing herself from the capulet, however it add a new stake since now romeo think she is really dead which bring the twist of her killing her lover indirectly which prompt her real death, thus terminating the contradictions chain. This is all structure, yet the story is compelling, the journey tell us something about humanity, we relate to the ongoing tragedy for itself. Interpretation of the story can bring new flavor or can bring new character without destroying the core appeal of the story, good structure give you timeless stories.

This deconstruction of romeo and juliet is pretty clear and is totally programmable, you can have a multi ending game based on it, with different strategy about resolving the contradictions. It give us also more power if we want to pick the theater piece and turn it into a whole tv season, we know we can pad by bringing stake on top of stake, (for exemple extreme padding would establish what it mean to be romeo in one season by bringing the character element and create conflict that are compatible with his traits, then follow by 4 seasons that only focus on one of the components with the final piece of information (for exemple that the two loving character belongs to rival faction) at the very end of the fourth season, you would have 52 episodes of 54mn just from the premise of the story. Some great series actually do just that and it's fine! Not sure I would consider this a great series but smallville got 10 seasons with this trick. It's pretty independent from qualities and give us a lot of power on focusing on just being good.

[gimymblert]

In short: focus on creating good structure before creating ornament/flavor generation. Once we solve the structure things, we got all leisure to perfect it with flavor. Mixing the two is a bad idea IMO.

[JWK5]

The reality is you can take just about any character (composed of any details) and throw them in just about any situation for any reason and end up with a story, but unless there is a coherent point to it all that the reader/viewer/player can take away from it then it isn't likely to be a very interesting story (it becomes a joke without a punchline).

The goal of any structure is to pace the story and to help ensure that it stays on course and drives that point home. It is possible you can create a good story without knowing much about story structure just as it is possible you can create a color-balanced painting without knowing much about color theory, but in all likelihood you are just intuitively coming to the same conclusions the structures present.

Structure is important in written art for the same reason it is important in comedy acts, music, drawing, painting, martial arts, etc. It's a framework passed down through generations of studying to quickly teach subsequent generations core skills they can further grow, expand, and pass down.

I do agree wholly with Gimmy that you should lay out the structure first (so long as you have already decided the point of the story), though I don't necessarily agree that you should do the details, or ornamentation and flavor as he put it, last. I think most of that will come about as you lay out the structure so they kind of go hand in hand. The structure is your bus driving the story along the path to the point, your conflicts and details are the passengers it picks up along the way. Doing this ensures that you're picking up the right kind of passengers for the place you are going and that everyone's getting there on time.

[gimymblert]

I didn't say it come last, it's more that once you have a vague idea of what to express laying a complete structure first is a the next move. Only when you have a structure you can ITERATE on it with the flavor, you will have new idea or more precise sense of feelings, flavor and structure will influence each other, however you will have clear view of where things doesn't seems to work. The first draft is never the last, only the seed.

[JWK5]

Ah okay, then we're pretty much on the same page there.

[Graham-]

It's going to take me some time to come up with a good reply. I'll do it. I'll probably need to bring up some examples.

[J-Snake]

Most stories made by hand are already bad. Just something to think about.

[Graham-]

Yeah, this is true. But that's for a number of reasons. First most game writers have big heads. They have big games so they think they have good stories. They get confused.

Second, it depends what you mean by story. Mario has a story that's told through its mechanics and the player's imagination. There is narrative when you jump on a bat's head. He swoops down and surprises you, you react unsteadily, but even-out into a corner-clip of his head. You knew you could make it, did so, then launch a shell off a wall to celebrate, making yourself small. Idiot.

In that case Mario isn't affecting the world in abstract ways, the way one would hope a "story decision" would in a proc-gen (story) game. But he could, and the reactions could be expressed purely through mechanics, in the way they already are.

It's kind of silly saying I want a good story _and_ have it generated, but it becomes much easier when I tie everything back down to the mechanics. At the current stage (of design) my game has no text, partially because of the "non-universality" of language/literacy, even in the menus. So everything will theoretically be expressed in means that have both a mechanics and narrative interpretation. The player learns the language through play, then plays to interact in that language.

I hope to keep the game close to the roots of games: interaction.

.

General note. I wrote a blog post, as a response to my various wall-fights with Gimmy. I tried to illustrate "how" some questions can be answered that I hadn't yet. It is really long. But it is very general, non-technical. You probably need some comfort with design theory to get something out of it, and maybe a little understanding of programming. AI would help a ton. I've thought in AI terms for so long I lose sight of "what pieces" of my ideas can be implemented with common knowledge and which require a technique or two. It's just a lack of perspective.

Anyway it is very long. It's also unfinished. But I realized I'll probably never finish it because of other work, and when I return to the topic I'll probably just write something different, or I'll just work on the game. So I figured I may as well share it. If anyone has the gusto to get through it I'd appreciate any feedback about which things feel light on the "how would this actually be implemented" front. That's where I need the most work in my explanations. It would also just help me in general. If you have any questions I'll probably answer them, eventually.

If you do go to it, trust me on one thing, all the techniques are me discussing things that do work with the right application, in the ways I suggest. Err on the side of me not explaining fully rather than me not understanding the problem. My solutions work but my explanations are... what they are.

Anyway, if you _like_ proc-gen I'm sure there's something there worth thinking about... maybe. You know, whatever.

Here it is: http://coldtheory.blogspot.ca/

edit: It's a little disjoint but the content is sound. (that's what I mean by unfinished).

[Graham-]

You know how games like, can put you in the right mind set?

Like, when you go home to visit your Mom, her personality, and the smell of her food brings back your past? Or like, the water-front does this to me. I put on a Pat Metheny album and I remember how cool I am sometimes.... in my imagination. Or I do something else, like put on a Trance playlist and imagine a thousand people dancing outside under the stars... and there I am flashing with my clothes trailing out around me as I weave through the music, building up to something and releasing, again and again.

These kinds of associations are the "basis" (?) of memory. It's the basis of communication... and that's kind of a truism. What I mean is that we relate things to other things. In abstract mediums we relate using common principles. That's what some of the study of "music" and "game design" is about.

... I'll get back to this thought later. (It's related to the game).

[gimymblert]

I have read the blog, game you should play (i mean study, even with faqs):

thousand arms
princess maker 2
tokimeki memorial game boy color
sakura wars
valkyria chronicles
Valkyrie profile: feather of convenant (something like that)
Harvest moon a wonderful life
Gothic 3

What you have done have been already describe and structured, cotillion from evans and short notably, also the GDC about left 4 dead banter system, dating sims provide an (underused) framework that does mostly that (quite similar to facade drama management in general ways).

I can see we strongly diverge in the co author depth of story playing, I want a full debate, you want a flexible ride.

Also when I say we need more structure, you discuss them as a syntax level and I discuss it at the discourse level, we didn't meet eye to eye because I have already done (i'm not alone) all the observation you are doing now, there is no revelation for us (see emshort corpus of work) but I guess you are in the process of organizing the data for yourself.

Here is my early work:
https://dl.dropbox.com/u/24530447/Love%20revo%20gamedesign%20not1.pdf
https://dl.dropbox.com/u/24530447/dnaofgame3.jpg
It cover implicitly anything you told about (beware of early gimmy speak and french diagram)
https://dl.dropbox.com/u/24530447/linear-interactive-storytelling.pdf
here is a thesis that describe some of thing you told about

[gimymblert]

In fact i'm more interested in the part of mario/HL2/Minecraft

[Graham-]

MM... I don't really feel that way.

There is definitely some stuff I see you do, for sure. Right now I feel like I have enough theory to last me for a while. I don't think we diverge that much at all. It's just... how do I put this?

Deconstructions, like finding the grounding principles in a story structure for example, I did that like 3 years ago, then I broke down characters 2 years ago, along with mechanics. Right now I'm not really "exploring" anything. It's more like I'm summarizing (when it comes to AI design).

Like I see all these "simulation" kind of games and they're sort of behind me. It's not for the disrespect; they just are beginning to dabble in AI. So if you think I'm in that area, I mean, we're really not communicating.

The single most important aspect in good AI design is isolating the most compelling parts of a person's decision-making. If you want to have a character who "loves" you, and you want to now betray them, how do you determine what their best course of action is, to maintain reasonable consistency? Particularly, you want to get the most mileage for the data you have.

You have to define what "love" means. A lot of AI designers spend a lot of time on structure, and that's fine, but then they try to apply this structure to some scenario and it doesn't fit. The reason is because every situation is unique. You need to define betrayal as a series of behaviours/thoughts, and love the same way. That's the hard part.

I want a full debate, a flexible ride; I want everything. I don't know where you get this from. I don't make strong debates about the theory you have that I don't see. But you make strong ones about mine. It doesn't bother me. I just don't understand.

AIs need to be so flexible that every definition you give of a concept can be defined as an abstract series of relationships with every other concept, which themselves are all abstractly related. That's how a human brain is built. We associate with other associations. AIs need to be like this.

The idea that "some theory" can be found that produces a data structure that "all" behaviour can fit into is the single biggest misconception in AI design. It literally holds everyone back who thinks that way. Trust me... I left University because the AI schooling was so weak I could learn more on my own.

No one else on the planet can write a functional Poker AI - yet, but soon - because their designs are too restrictive. They don't account for the inherent flexibility of Poker. In chess you can "figure out" the game. But in Poker all that matters is finding patterns in your opponent, and they, being human, can create any pattern at all. Game Theory is the wrong place to start there, and that's where all the AI designers like to start, because it's easier. And that's why the world's best Poker AIs can't play even moderate limits of No-Limit, while Chess AIs can beat the world pros. And on top of that, Poker AI is 10 times more lucrative.

I have never seen an AI design that has the kind of flexibility in it to allow "concepts" to be defined in entirely abstract terms. There is always some "structure." Like, if your AI uses Behaviour Trees you are doing it wrong. There is no "theory of life" that we can find and run our AIs off of. You can find a theory for a particular situation, but not everything. There's a point where you've deconstructed so many scenarios that you realize that you will never be done, ever.

You could say I'm more interested in the process for deconstructing. In a simple sense my game is about putting players into a position where they run deconstructions for me. They feed content into the system, then isolate the relationships those concepts have with others, and I automatically use them as building blocks for the generators.

The reason I keep using analogies is because "theories" are only tools. What you need to pump up your AI is experience. You have to take your theory and just tear ideas apart, over and over again. You never stop. So I don't. Your AI should never run out of fuel.

Why does Chris Hecker's design fail? Why do most proc-gen stories feel forced or uninteresting? I mean, try and plug Facade into Mario so that the player expresses himself through mechanics? Are they close to that? Well, sort-of actually. I'm not dissing them; I'm just like that. They fall apart because their "structure" limits the definition of ideas. Once you pick a "theory of games" that your generators follow to produce ideas, you now have to define every idea using that structure. It will _always_ limit you. The most definitive part of the brain is the fact that any piece is flexible. That's what AIs have to look like. Then we have to feed them reams of data, from horror movies, from our love lives, from our favourite games.

You see me deconstructing all the time because I'm flexing my muscles. It's like you're like, "real writers have a theory for writing, and don't need to explore writing different things all the time." And I'm like, "the best way to improve your writing is to always write. It's when you think you've 'found the formula' that your creativity falters."

I know it always feels like we get into these argument, but I don't mean for it to be that way. The internet is just like that, or I suck at it.


----

Notes.

1. My blog was just on some basics, because I don't know what people know or do not know, or whatever.

2. I think by "flexible ride" you mean structured story. That was the focus on that post. I wrote half of the next one which gets into the, "how to turn that into something totally free-form area." I just started where it was easiest.

As an illustration, currently my game has this formula:
  a. It plays like a movie. It will just go an go forever, with everyone doing everything. The world will grow w/o the player, until...
  b. The player does something, then it starts to adapt to everything he does.
  c. The player can affect anything.
  d. I'm trying to put my generators as "low" into the process as possible, so things like the textures get generated (dynamically) based on several things, and hopefully as much else as I can.

3. Discourse is more complex to talk about. That's why I avoid it. Discourse is a personal discussion. Having a "theory of discourse" I already have. At this point all I need is more experience doing it. That's better left to the real world, applying my theory to find patterns. You think that because I don't discuss theories on this area I don't value it. So I dissect things and you go, "oh, _I've_ already found a formula for this; he's just not caught up." And I'm going, "nope. I've got your formula too, and I know that shit only lasts for a little while, and now I'm dissecting everything I touch, because I can never dissect enough."

You're like, "here's the formula for deconstructing combat systems." And I'm like, "here's 20 combat systems." I have so many formulas I can't express them in a post. It is literally thousands of pages of crap. That's what the game is for. That's where I express those. Here I just keep on deconstructing. When I get to the higher level deconstructions - the ones I'm actually working on - like "love" you get all, "save that for later bitch," and I'm like, "that's where I am man." I have the base formula that suits my purposes. You keep going back to that, and I'm like, "I really value your insights, but I don't need to write about that anymore."

And on that, I really _do_ value your insights. Nearly 4/5 times I read some guy who's like, "here how you generate stuff, or here's the "theory" for game design, or whatever," I'm like, "I _knew_ this shit when I was 19. Get over yourself." But I like your writing because it is to the point. You never quite deconstruct things in the same way, and you certainly don't stretch it so I have to re-read Mr. Mom's theory for the 27th time, and I appreciate that.

Gimmy love, all right.

You have a lot of perspective. Your life is valuable. Your personal insights make you a unique person. But when it comes to AI, when it comes to proc-gen, I just don't understand how the next move is going to come from someone who isn't technical. Right now game design is weak because we haven't learned how to forge art and design together. They have to come together. We don't program in the right way for real AI yet. What we need is a union of creative expression and technical implementation.

I give deconstructions and not their implementation because implementation is what programmers do. I can't draw to save my life, but I can take abstract scenarios and program them because that's what I've been doing for 10 years. I don't need to go over that shit _again_.  Finding the common constructs in a series of abstract concept is what programming _is_. That's programming. So I give my insights into concepts. That's the part I always need to flex.

I think we're a lot closer than you think we are.

Hopefully you take that all in a friendly tone. I have a good deal of respect for you and I don't mean any of that in an aggressive way. I write for the exercise.


----------

Ok, so I skimmed over your final link. Yes, it's kind of funny the similarities. Never before had I had to describe why certain tools would work, like why player expectation is critical etc. It's almost like deja-vu.

Here are the differences, that may not be obvious:
  1. The essence of my post is the tools, not their justification.
  2. My tools are far more powerful.
 
This might seem backwards. That thesis you have produces a similar theory as every other. It tries to define the basic components of people and story using a single structure. That's fine to begin with, but it will fail you once you start up with more complex topics. "Technically" the tools they describe are the same, but they aren't described in their most basic state.

The article admits that it can't create compelling book-quality stories with dynamism. Mine can. Not a joke. You need to take the tools it presents and abstract them further, into fundamental pieces, then see how they can most quickly be applied to existing life-experiences, how their output can be tested and refined efficiently, and so on. That is the next step.

-----------

I'll create a summary for you.

You have this list of deconstructions. You deconstruct the story, and the character, and the conflict, and the climax, and the combat system, and movement of information between players in a competitive game, and so on. And you're like, "we are _this_ close to covering _everything_." And I'm like, "no theory will cover everything." Write a theory that produces a single character that is as interesting as Don Corleone (from the Godfather - easy reference), and procedural? No?

Here you would say, "save that stuff for later," and I say, "the theory that deconstructs a single good character will be very different than the theory that deconstructs another." _That_ is the essence of good deconstruction. This is why AIs always feel so promising but never, ever come close to their aspirations. Because everybody tries to find, "the structure that will work."

Don Corleone is complex, like every other good character. Your theory will deconstruct some of him, but will then leave enormous chunks alone. In fact, it will leave the most compelling chunks alone. The artistic process is a real thing. We go through it because relating to things is hard. There is no theory. That's why it is a personal challenge, because we have to find our own personal way of relating.  Each character construction is unique to that character. Find me the theory for good character design? We've been writing books _forever_. So where is it? Surely if you can find the theory someone found it 300 years ago? No? Why?

This is what makes life meaningful. Expression is about finding the uniquness in ourselves. If there was some theory to produce our experiences then our individuality would be meaningless. To pursue a simple structure to define life is to trivialize it.

So it brings up the obvious question. What is the general pattern for deconstruction? What is the most effective way to define a character in general? I mean the most important pieces? If I had to define Don Corleone in 3 sentences, what should those sentences focus on? If I had 3 metrics that controlled 3 variables in my game, what should those variables be? If I wanted to use the same 3 variables for every character what should they be? What's the common structure _there_?

The best deconstruction for a single character is radically different than for another. Write the, "best summary," for 3 different characters from 3 different books in half a page. How different are they? _Hugely_. What is the common structure to express ideas in which all 3 summaries could live? And thus be manipulated through some obfuscated interface by the player?

It's this diversity in deconstruction, and necessity for flexible architecture, that underlies the proper design for an actual generative environment that produces content of artistic quality comparable to every other medium. If someone hasn't come to realize this I just start going, "you need to deconstruct more, keep going, you'll learn." Whatever structure you get will just break once you use it enough. That's why you can't find it _then_ build it.

It's like game design. Idea, prototype, test. You need to learn from your failures to find the common element in all of your ideas. You need to produce enough structure to realize than any structure only goes so far. Each time you carry a piece with you and learn a little about what the common element is among all of them. And I promise you, the most base element there is flexibility. When it comes to communicating with humans, expression, life-experience, the only core concept is change.

You could say that my system "generates" structure, and allows any new structure to defined dynamically, by relating it to any set of components, then be tuned using any language you wish. You can say, "oh, make him a little more like James Bond, but less cool, and more forth-coming." If I have other concepts in the system that are defined using any of those concepts, then it will understand that definition. So I can deconstruct a character using this system, then another using some other system, and the AI just uses both, and finds which one produces the more compelling being and learns which structure is more situationally useful.

I don't have to pick one system. I just keep throwing every one I've got in there. My characters are defined using as many systems as I can get my hands on. So you could come in and define a character using yours, and someone else could use theirs, and so on. I don't restrict the ways things have to be defined.

I've been doing this a long time too.

But I love your posts. You get me to write. Thank you.

Phew.


---

What part of Mario/HL2/Minecraft?

(thanks for reading btw).