Hey, Mendel, :)

Interesting thoughts.

Question: does this allow for/recognize the same player locking into different views depending on the game being played, or even, depending on who that player is playing with and what the other players' views are?

Cheers,
J.

Hi, :)

I wait on in anticipation, then. Interesting stuff.

Cheers,
J.

Thomas,

Private language arguements don't really apply. The "unknowability" is a softer one. You cannot be certain what is being attended, but you can begin to build an understanding of a view by observing what content does and does not cause a response. Likewise, you have trouble being sure what is outside your attention, simply because you are not paying attention to it.

Views are filters, that remove everything which is not attended by a player. Controls is where we start talking about modeling thoughts, and even then, the specific thoughts need not be shared, even the means by which the view is implemented in the player's mind. The private aspects are bypassed in order to focus on observable and testable properties of play. We don't need to know why a player reponds predominantly to emotional content, only that this trend can be observed from watching him or her play.

As far as congruence, the neat thing is that you want views to align only partially, which prevents stagnation and encourages patterns. To link back to a Big Model perspective, you need a little coherence and a little incoherence in order to get fun.

- Mendel

RPGs should not aspire to be art or fiction, they should aspire to transcend such things.

Mendel,

I think I'm having trouble parsing out how this is anything specific to roleplaying. Is it supposed to be? I mean, it seems like you're edging around linguistic uncertainty theories and stuff, which are totally cool. SInce those theories are about how we interact, and roleplaying involves interaction, they're definitely applicable.

But I get the impression that you're talking about something different, or at least more specific, and I don't think I'm getting it.

Thomas

Jess,

You seem to be spot on with the idea of views. Especially with your spotlight analogy. One clarification though is that the "surface" of the game is more than just the story, and views can extend to include such content. Indeed, that is what happens when you extend your view to identify and include the views of other players.

Additionally, moving beyond the static, views are also what drives the strategies of players as they seek patterns. One good example of such a strategy is the Nash Equilibrium-style situation where you and Chris recognize and feed each other's views. Play controls, which is my term for these strategies, can become even more complex if you consider that views can also change as a result of these strategies.

As far as Koster's book, I've read his earlier presentation by the same name. He relates some of the same ideas I'm working with, but he doesn't appear to be taking them very far. On the subject of game design, I ought to post a summary of what I've called induction, the process by which peripheral content, such as game texts, manifests as patterns during play.


Thomas,

Views and PCon3 in general are not specific to RPGs. Rather, they are what you might call native theory to RPGs. Views are relevant in any situation where multiple people influence common content and there is no strong external criteria for outcomes. RPGs are especially good examples of these situations, and views were intented to explain them predominantly. But this does not mean that views cannot be applied elsewhere.

- Mendel

RPGs should not aspire to be art or fiction, they should aspire to transcend such things.

Thomas,

The last time I introduced views I had an entire scheme set up in terms of atomic views and ways to combine views. It made a regular language, with serial views, composed views, and even a kleene star operation. If recollection serves that was one of my last posts in the Forge RPG Theory forum.

Since then, I've reconsidered if the structure of that language is a sacred cow that could do with some slaughtering. I'm still debating that, but there seems to be clear types of patterns, and it is worthwhile to speak of classes of these.

For example, you could consider views of random walks, where the step-by-step continuity is the pattern viewed. Likewise, you could consider the repeated test pattern, where the view focuses on differing inputs and outputs of a cyclic procedure. The later could apply to social or geographical exploration, while the later to learning any sort of procedure.

But in general, I don't believe that a classification system is as useful as a proper language of views. In either case you can specify general groupings of views, but in the later you can do so without sacrificing greater levels of detail. In short, classification systems end up being nested, forcing the creation of more and more terms, while a language presents a logic to build terms at any depth.

- Mendel

RPGs should not aspire to be art or fiction, they should aspire to transcend such things.

Mendel, I'm glad Jess is getting something out of this because I'm finding it horribly opaque. I don't know how much of that is the subject and how much of that is your writing style. In fact, I read your initial post exactly like Thomas. It seemed like you were just saying "people have different preferences/reactions and these preferences/reactions are sometimes in alignment and sometimes not." I'm glad people coaxed a further explanation out of you, but I still feel like I'm missing something.

What you're saying sounds a little like the reading I've been doing in communications theory, especially small group organizational communications and the formation of communities of practice, but it seems much more focused on the personal/player level which, I think, has been the traditional focus of most traditional roleplaying advice, including GNS. And, honestly, I feel like this approach to roleplaying communications is misguided. We end up trying to satisfy/understand the individual needs/preferences at the expense of creating a truly fuctional and supportive community.

Like Thomas, I'm trying to figure out how this "views" stuff is differerent than any other meta-perspective on inter-player relations in roleplaying and what new tools it possibly offers us. Right now, I'm not really seeing anything. Is there a way you can link to this PCon3 stuff (or explain what that is, at least) so we can get a better sense of where this is coming from? Because right now it just looks like a different way of saying the same thing, which is fine, but I'm having trouble seeing what we learn from saying things this way.

Jonathon,

As far as views are concerned, they are nearly as simple as you describe. The only addition is the complexity property, where player enjoyment comes precisely from intermediate complexity within a view. What may not be apparent is that when I say that I'm actually talking about a quantifiable property which we can estimate by recording play. If you'd like I can give a short summary of complexity methods and why this is the case.

To have fun in play, you need to meet that complexity criteria, and you can apply views solely with that intuition. Views are looking for a particular type of pattern, the worst dysfunctions in play occur when multiple views all percieve their pattern beginning, but at some point the pattern diverges and the enthralled player is suddenly starring at static, for example because one person wants a cycle of hope and despair, while the other wants a downward spiral of heightening doom.

You are right that views only deal with what a single player wants. But that is mostly because they are the part of PCon3 which deals with what a single player wants. Play content, play controls, and play constraints form the rest of the theory, which attempts to model RPGs as a dynamic system fed by the players, and hopefully achieving a stable set of constraints over time.

My original Forge post on this can be found here. It may be of use.

- Mendel

RPGs should not aspire to be art or fiction, they should aspire to transcend such things.

Thomas,

Fair enough. My professional background is in mathematics, physics, and computer science, focusing in complexity and dynamical systems. I have some professional exposure in biology and biochemistry, there was a time I purified hemoglobin from human blood. Beyond that I have a hobby level interest in modern philosophy, history, and cognitive science. I'm currently attempting to build a similiar level knowledge in ritual and cultural studies (based on Chris Lehrich's suggesting reading from a few years ago).

What I've been talking about here is a theory based on my perspective of RPGs, intended to be experimental in nature. It shouldn't be too surprising that dynamical systems and complexity play a significant role.

- Mendel

RPGs should not aspire to be art or fiction, they should aspire to transcend such things.

Thomas,

Very well, would you prefer I started explaining the background here or split off to another thread? Either way, I suspect the first step would be Kolmogorov complexity, leading into the broader topic of complexity and patterns.

- Mendel

RPGs should not aspire to be art or fiction, they should aspire to transcend such things.

Mendel -

If you want my two cents, getting more into the theory is not going to help people understand what you're saying. Your writing style is very opaque and abstract, although I like your ideas. Maybe, while you're explaining the theory (here or elsewhere), you could also give people some concrete examples instead of just talking about general principles.

--Jess

Jess and Fred,

Sssshhhhh! If you keep this up, Mendel might not post background stuff. Then I'll be forced to go figure out Kolmogorov complexity on my own. I mean, it sounds interesting even outside of this "views" context thing, so don't spoil this for me!

*wink*

Thomas

Alright. I'll try a different approach and mix all three (background, PCon3-style theory, and concrete examples). I'll be posting what I hope are bite sized pieces. Here goes.

Strings

The simplest concept for understanding complexity is the idea of strings. A string is a sequence of symbols taken from what is called an alphabet. So for example:

abbababbaa
baababba
aa
(nothing)

are all strings. What's more all of these strings use the same alphabet containing both a and b. Not every symbol needs to show up, and the string can be as short or long as you like, what matters is that only symbols from the alphabet show up, and they are in an order.

Why is this important? Because we can model the active content in play as a string. Each glance, or suggestion, or roll of the dice, gets added onto the string one after another. The problem is for all practical purposes, the alphabet is unlimited. Each glance is different from the last and a character action may never have been uttered before, and may never be uttered again.

In the most basic sense, a view is something that takes a string, and turns it into a string from a particular limited alphabet.

Take a particular play example, like this one. Each sentence is in a way a symbol in the play content, but a player doesn't care about every last detail, so we can replace the symbol of:

JMendes says "I stand up ready to face them, but there is a trail of blood..."

with "change of fortune" or "gory detail" or "dramatic moment", and indeed we would expect different players to replace that symbol in different ways. But the advantage of the "gory detail" perspective is that more such details will be making an appearance. In fact, it may be Nathan's intent here to nix any endings which have too many gory details, focusing instead on clever or classy details, in which case a view that reduces to "gory details" works fine for him.

Computer Science Aside

You can think of a view as a function mapping a symbol stream of an infinite alphabet to a symbol stream of a finite alphabet. You can think of it as two parts, first a map that takes the infinite alphabet into a finite intermediate one, and a computable function which takes input stream under that mapping and produces the final output stream.

For example, we have an infinite alphabet {a1, a2, ...., b1, b2, ...} which is mapped to {a,b} by any ax into a, any bx into b. And the computable function is 0 if two consecutive symbols match, 1 otherwise. Transforming a2a7a89b1a8b7b54 intermediately into aaababb and then to 001110.

End Aside

Now people don't always reduce alphabets like this, sometimes you just sit back and try to absorb everything going on around you. But if you want to respond, then you need to do so. The reason is simple, without reducing unique events or stimuli into a broader symbol, we can't recognize patterns.

But what is a pattern? Stay tuned for the next post.

- Mendel

RPGs should not aspire to be art or fiction, they should aspire to transcend such things.

Kolmogorov Complexity

It's not as scary as it sounds, really.

Remember strings? Well consider a string that looks like this:

"repeat once" "repeat once" "add a" "add b" "end repeat" "add c" "end repeat"

Now this is both a string and a set of instructions on how to produce a string. This is an example of how strings can also be programs. In this case the string above tells you how to produce the following string.

ababcababc

One of the things about programs, is that they have a certain alphabet of instructions, which translate into actions, this program uses the following alphabet:

{"repeat", "end repeat", "add a", "add b", "add c"}

And using this alphabet we could make any string of a's, b's, and c's. But what if you wanted to write instructions for strings with d's or "repeat once"'s? There is a standard trick, using binary, that lets you reduce any alphabet into 1's and 0's. This is called coding, but all that's really important is that you assign a string of 1's and 0's in place of each symbol:

So we could replace "repeat once" with 00, "end repeat" with 01, "add a" with 10, "add b" with 110, and "add c" with 111. If we do that the program we had above looks like this:

0000101100111101

And we can do the same with the output too.

Now we can get back to Kolmogorov complexity. If we have a specific set of binary instructions, we can define Kolmogorov complexity for that set of instructions. If we take an input string, the Kolmogorov complexity of that string is the smallest string of intructions which will produce the input.

For example, using the instructions we have up above, the Kolmogorov complexity for the string abababababababababababababababab (ab repeated 16 times) is just the length of 000000001011001010101 or 21.

One of the problems with looking at a single instruction set is that some strings are hard to write instructions for. If we had repeated ab 15 times, then the Kolmogorov complexity would have been much larger, since we couldn't use the doubling effect of nested "repeat once"'s. We can imagine a different instruction set, perhaps one with "add ab 15 times" as an instruction which would do far better.

Fortunately there is a solution, it is possible to have an instruction set that can execute other instruction sets. This is the same principle by which your computer executes programs using programs. This universal instruction set gives us a general Kolmogorov complexity, which is just the minimum sized program to produce the input string.

Kolmogorov complexity is also called description length complexity, because these programs are the smallest complete description of the strings they produce. Casually you can estimate Kolmogorov complexity by describing instructions on how to completely produce a string. A short description means a small complexity, like a long sequence of "I attack" "I roll" "I miss". If you can't describe a sequence of symbols or events shorter than the sequence itself, then you can treat that sequence as random. Imagine if a player created dialogue by randomly picking words from the dictionary, or more reasonably from his word list in Og. You could only reproduce what he said by recording each word.

Computer Science Aside

I am glossing over the Turing machines in favor of the instruction sets. Basically, Kolmogorov complexity is the minimum input required to produce the string as output for a given universal Turing machine. By simulating universal Turning machines you can show that the difference between any two universal machines Kolmogorov complexity function is bounded by a constant, which lets us abstract out the specific universal Turing machine.

Also, Kolmogorov complexity is uncomputable. There's a simple diagonalization arguement which goes something like this. If it were computable, then you could use it as an instrucion in a program, which means you could write a program which produces the "lexicographically first string of Kolmogorov complexity greater than the length of this program", which produces a contradiction by producing a smaller description than the Kolmogorov complexity.

End Aside

When using Kolmogorov complexity you should always be aware that you cannot usually determine the complexity precisely. You can produce smaller descriptions, but you can't be sure that you've found the smallest one. It's always possible that there is something you've missed. In a way, Kolmogorov complexity tells you how much unique information there is in string.

Many applications of Kolmogorov complexity use a basic principle of information, namely that it behaves opposite to entropy. Just as entropy always increases, information always decreases. So if I have a some input string, and I have a process, which I can also describe as a string (using instructions like we did earlier), then the complexity of the output is limited by the complexity of the input and the complexity of the process. Basically you can't produce any new information that didn't come from one of them.

This is one of the reasons randomizing is so useful in RPGs, the output of a die roll is a high complexity string, which means even if we lose information in the process of playing the game, we still have enough information left to keep things interesting. On the other hand, in a very interactive game, the process of play and the input of the players often provides ample information, and the randomness would risk adding too much. Hence randomizers are often avoided or ignored in those play situations.

But but it is the patterns that are "interesting" in that information. With the idea of descriptions here, I can start loosely defining patterns. Patterns appear when Kolmogorov complexity is reduced, as the pattern lets us describe the data more succinctly. But when we play, we're not just looking for any patterns, we want patterns with some complexity, which means we want to avoid both extremes of Kolmogorov complexity, the low complexity repeatitions and the high complexity randomness.

Notes:

I apologize, there wasn't much in the way of concrete examples for this theory bite. However I promise to remedy that next time.

Also, Fred, views are lossy, more from the alphabet compaction, than the computable function, although both can be lossy.

- Mendel

RPGs should not aspire to be art or fiction, they should aspire to transcend such things.

It strikes me as intuitive, Thomas.

Uncomplicated strings are boring.

Overcomplicated strings take too much work to process.

Yeah, but you have to translate for us wishy-washy humanities majors who like soft-serve social sciences. With fudge on top.

Mendel,

The last one is the most fun of the three, because it balances randomness with pattern. You have some reliability that after your first success, you'll get a second one for free. At the same time, you will quickly learn when failure is possible and when it is not.

Is this a hypothesized, or studied answer? I'm seeing that your views model suggests that intermediate complexity is ideal for human pattern-sorting, but I find myself unconvinced that this is actually the case.

Here's a quick recap of what I'm getting from you: we can use techniques from Kolmogorov comlexity study to analyze patter complexity in game interaction, this is interesting. For some reason, you hypothesize that humans prefer an intermediately complext patterns over extremely complex or extremely simple patterns.

At the moment I'm not sure I buy your intermediate complexity hypothesis, but I'm willing to grant it to you for the purposes of discussion because I sense that it's not really the focus of the views theory anyway. So feel free to move on, for now at least, and we can come back to it later.

Thomas

Mendel,

Oh, wait a minute... "I'm talking about pattern learning, particularly as the source of fun". As the only source of fun, the primary source of fun, one of many sources of fun? That's going to be important, because I'll probably fight you on the first, and might on the second, the third I agree with.

Once we figure out where we can agree on the above, I think I can agree that some concept like intermediate comlpexity is required in order to evaluate the fitness of patterns for this sort of fun. However, intermediate complexity may be too complex for anything but evaluative work. Some patterns are going to be too simple for me and too complex for my friend and just right for you. Further, with time patterns grow more familiar and their apparent complexity is reduce. This would seem to indicate that a pattern that is "intermediate" right now might be sipmle in two or three months of use.

Is that ll in line with what you're thinking?

Thomas

Mendel,

Okay, see, that was causing me a lot of confusion. Now that we've established it as an axiom, I think I'm on board with you. So...

In the interest of full disclosure: I have severe reservations with your proposal that all internal fun (by which I assume you mean all fun, period) is the result of pattern learning. Still, I'm willing to grant you the axiom in the interests of seeing where this is going.

Part of the problem is that it strikes me that the question of whether fun is due entirely to pattern learning is quite possibly an emperical one. Since it's non-intuitive and emperical, I tend to demand that emperical proof. Still, as I said, I'll grant you your axiom: say on, Mendel, say on!

Thomas

Mendel,

I think I'm having some trouble with your use of the word "views". Are you using it ambiguously, or am I misunderstanding. You speak of views as things that are, as an identifier of the interpretted patterns of play, and you speak of views as things to be attained, as things that players try to accomplish.

Are these two aspects of the same thing, two different things, or one thing that i can't seem to grasp all at once?

Thomas

Mendel:

But from your argument, wouldn't you need actual data about what the players are thinking, not just what is happening in the shared fiction? I guess BitV has some of that stuff in the OOC chatter, but you might want to consider finding places that you think are interesting and then doing in-depth interviews about them. I'm happy to talk. :)

--Jess