Altering the SIS in CRPG

[M. J. Young]

Paul--I'm not certain I understand your point.

You appear to be distinguishing "computer as referee" being distinct from "computer as player"; if that's your distinction, that's not what I meant. It's generally recognized at the forge that character-players and referee-players are all players.

My point is that the computer is the player, in this case the referee player, and that it is capable even in its current level of technology of interacting with living players through sharing an imagined space.

Callan's point is that the computer cannot share an imagined space, and thus the living players at the console are sharing the imagined space with the game designer, who has made his contribution.

I disagree, in part because that would lead me to conclude that if I were playing the D&D module Keep on the Borderlands, Gary Gygax is participating in our game because he wrote the rules and the module. The standard objection to that is that Gary Gygax contributes nothing to the shared imagined space directly; rather, the players at the table refer to his rules and module and use that to inform their contributions to the shared imagined space. Similarly, George Lucas does not contribute to the shared imagined space of our Star Wars game, but rather created the world which we use to inform our own play. Lucas had no notion of creating a role playing game, and it's doubtful whether he's even familiar with how the various incarnations of his game play. He certainly is not familiar with the many freeform live action roleplay games based on his movies that children have played for decades now; he is not a participant in the game, but merely a resource upon which the participants draw.

In the same way, I would say that the program is the reference used by the computer or console, and the computer/console is the participant in the game, doing what it was taught to do. I cite examples in which such programs are self-modifying (the various chess programs), but suggest that even the most rigid referee-player is still the participant, and not merely the conduit through which the books speak, and thus the same applies to the computer.

So, where are you in this? Is the computer a participant, or merely a conduit for a participant in absentia (as I think we'll agree telephones and messaging programs and forums are), or something else? Is a CRPG a role playing game with a shared imagined space, or is it a role playing game without a shared imagined space, or is it not a role playing game at all? If it is the first of these, with whom is the imagined space shared? If it is the second, what makes it a role playing game? If the third, what is it?

Thanks for your input.

--M. J. Young

[Callan S.]

Hi Christopher,

Ah, before we bust out dictionary definitions can I say something about that. I hate them. Now, if you want to ask me what I mean by exchange and then tell 'Callan, you got it wrong, the right word for that is X', that's cool. The actual word doesn't matter to me, I chose one which is pretty close, even now that I look at dictionary.com.

Let me assure you, there is an exchange. There aren't multiple exchanges, I grant, but the one exchange that does happen takes some time usually. Now ask me what I mean by exchange if you want.

No, I'm not actually mixing them up. In at least one of MJ's Chess examples, the programmer was programming the computer to learn to play. Once the computer has done some of that learning, I think it clearly is playing the game.

You are mixing them, because I only covered current comps. I breifly mentioned an analogy where a really poor but present player might improve his skills so much he eclipses an absent player who plays via e-mail. This will be computers latter.

I don't really want to cover AI, because it deserves a focus all of its own, but someone programming a machine to learn has basically done the same as teach it...except it teaches itself. Teaching and programming are very different.

Anyway, let me be clear, I did say that if the machine can eclipse the input of the game author, then your interacting with it and not the author. This can happen, but isn't now thought it's mildly possible now.

Can you explain (hypothesize?) what a human does that you are willing to attribute as creativity that a computer does not?

"1 + 1 = 2"
This is not creativity.

"There's a cave with a cave painting on it's wall."
This, briefly, is.

And if they are the authors exact words printed by the program, it's clearly him that's contributing.

If you're gaming with your brainwashed GM (for example) but you don't know she was brainwashed and "programmed" for a certain style of running, do you, as a player, think that you're playing with her or her handlers? And then, what's the difference between that "brainwashed" GM and any of the random GMs that have trained themselves to GM D&D a certain way to the point that they won't consider another style as being meaningful or valid? Are they also brainwashed? If so, who are you playing with and if not, how can you ever know? And further, how can you justify the dichotomy that you're drawing between the brainwashed and the creative?

You've missread by what I mean by brainwashed. Now, if I GM by shouting down a tube which connects to the room the players are in and they hear me, I'm GM'ing them right (even if it looks silly)?

So what happens if I replace the tube with some other medium. A medium that will only say what I want.

Were you RP'ing with the tube before? How about this new medium? Are you roleplaying with that?

Finally, you're clearly ignoring the example of systems that are programmed to evolve new algorithms. MJ's Chess system is a clear and extant example. There are lots of alife examples that aren't games. But I think they're possible now, and just around the corner.

Uh, yeah, I am ignoring them. I decided to focus on CRPG's like like fall out or such, just to make it simpler. New emerging tech is getting to be different, but I thought we were mostly talking about what is available now. Well, that's what I focused on, M'kay? :)

[Callan S.]

Heya M J,

Err, weve forgotten my example pretty quickly...the one where one player communcates by IRC or even by e-mail. There is one player actually present and one GM.

Now, couldn't they just ignore the IRC or computer?

If so, couldn't the GM ignore the player who's present?

I'd say they can...it's just that it leads to loosing out on a certain reward from including them. Just because someone sitting there, doesn't mean your forced to include them, like it or not. But there is a reward for doing so.

Now:

I disagree, in part because that would lead me to conclude that if I were playing the D&D module Keep on the Borderlands, Gary Gygax is participating in our game because he wrote the rules and the module.

I want to ignore the rules part, for certain reasons. But the module part...is there a reward for individual(s) at the table to include his work, rather than ignore his work?

If so, how similar is it to including another players 'work', who sat next to you and just gave it?

[simon_hibbs]

I don't really want to cover AI, because it deserves a focus all of its own, but someone programming a machine to learn has basically done the same as teach it...except it teaches itself. Teaching and programming are very different.

They are, and I think it's a distinction that we disagree about. Back in around 1980 or so my father wrote a program in Basic that could learn how to play noughts and crosses, it was based on a magazine article. The same has been done with chess, you set up a learning algorithm and then program in the rules of the game. Note that you don't need to know any chess strategy whatever to do this, or even to have ever played a game.

The program can then be tought to play, perhaps using existing chess computers to play games against it, letting it learn by trial and error. If i then play the program, I fail to see how I can be said to be playing the programmers in any meaningful sense. They don't know the board layout, they don't understand strategy, and would be totaly unable at any given time to say why the program chose a particular move, they wouldn't even necesserily know I was even playing the game. I am not playing with them.

The computer on the other hand does internaly store a representation of the game state (by analogy, the shared imagined space), and I do interact with it. We engage in two-way communication. To me, this is playing a game.

Anyway, let me be clear, I did say that if the machine can eclipse the input of the game author, then your interacting with it and not the author. This can happen, but isn't now thought it's mildly possible now.

It's not just mildly possible, it's been done for decates. In the 1970s a program called Heuristo killed the Trillion Credit Squadron (An SF wargame) tournament games. It was program that generated, tested and selected startegies for designing and building space fleets and wargaming with them. The programmer had no input into what strategies the game came up with, that was an emergent property of the heuristic algorithm and the rules of the game. If you google for 'heuristo' and 'trillion credit squadron' you should find some info on it.

Even deterministic algorithms can produce emergent properties that are not easily deducible from the starting conditions. A computer programmed with a random system for generating environments and resolving character actions within them could quite easily produce situations and outcomes not predicted by the game designers, and in fact this happens all the time.

"There's a cave with a cave painting on it's wall."
This, briefly, is.

And if they are the authors exact words printed by the program, it's clearly him that's contributing.

Random algorithms can generate imagined spaces (a cave, with a wall painting) and then render that as text, or graphicaly or even verbaly. The medium of representation isn't relevent.

You've missread by what I mean by brainwashed. Now, if I GM by shouting down a tube which connects to the room the players are in and they hear me, I'm GM'ing them right (even if it looks silly)?

So what happens if I replace the tube with some other medium. A medium that will only say what I want.

The tube and whatnot are irrelevent because they are merely media for transmitting information, just like the air in a room transmits sound waves. It makes no decisions. Computer programs can make decisions, and can even generate the algorithms and rules on which those decisions are made. This is old technology. However  would say that merely the ability to make a decision, to take certain inputs and generate an output, is enough. Pipes make no choices. Nor do they store an internal representation of the imagined space (and hence 'share' it). Computers do both.

As an aside I do think that scenario writers have input into the imagined space, but they don't participate in the imagined space of a particular game. They don't play the game, any more than the author of Monopoly plays the game with me and my family. They contribute in the same way that an author contributes a novel. Reading a novel isn't playing a game, nor is writing one and nor is writing a scenario.


Simon Hibbs

[Callan S.]

They are, and I think it's a distinction that we disagree about. Back in around 1980 or so my father wrote a program in Basic that could learn how to play noughts and crosses, it was based on a magazine article. The same has been done with chess, you set up a learning algorithm and then program in the rules of the game. Note that you don't need to know any chess strategy whatever to do this, or even to have ever played a game.

Oh, you mean those cute programs! The ones that use possitive feed back.

It really depends, on content of the game created.

For example, lets say I'm roleplaying with you face to face, but during combat I use my laptop to determine NPC moves. And it uses a learning algorythm to tell me.

Who are you roleplaying with? Do the move choices of the program eclipse my input as GM, as I talk up and vividly describe each move (in a CRPG this would be art drawn by humans, text written by them, etc). No, it doesn't. It chose the move...but really, you don't just see 'I kick him' in a book, you see a kick with a lot of extra description with it because 'I kicked him' does tell you much.

Anyway, I'd say your RP'ing with both of us. But really who gives the more significant creative input? Me.

Ah, but I don't have to. Suppose I don't give any elaborate descriptions...I just move the mini's and pretty much do bugger all. Who has the significant creative input? The algorythm.

Now, your right, you can be playing with the algorythm rather than the author. But I think you'll find it hard to find any games on the market where the writers input doesn't far eclipse any learning algorythm involved. It is possible to write one....just like in my example where I did bugger all.

I think we might be agreeing now, but I'd like to pick at a few points:

Random algorithms can generate imagined spaces (a cave, with a wall painting) and then render that as text, or graphicaly or even verbaly. The medium of representation isn't relevent.

I'm sorry, random input is not creative input. By putting RND into 2 + 2, it isn't a creative input. Creative input comes from the expression of what another being has learned and decided to express. In fact, it's something that I can learn from. Your algorythm from above does that a bit and as a learning being I can learn from the moves it expresses, even (interesting, eh?). From a random output, I will learn drivel. It's not the same.

As an aside I do think that scenario writers have input into the imagined space, but they don't participate in the imagined space of a particular game. They don't play the game, any more than the author of Monopoly plays the game with me and my family. They contribute in the same way that an author contributes a novel. Reading a novel isn't playing a game, nor is writing one and nor is writing a scenario.

I think you skipped a lot of what I said to M J.

[M. J. Young]

For example, lets say I'm roleplaying with you face to face, but during combat I use my laptop to determine NPC moves. And it uses a learning algorythm to tell me.

Who are you roleplaying with?

Callan, you make a reasonably good case for the notion that I might be playing with you and a similarly good case for the notion that I'm playing against your computer; but you missed a possibility that I think is very informative for this discussion. What if in this case I'm playing my character, you're the referee, and the computer is playing the characters you're defining as non-player characters? That means there are three of us in this game. It may be that it is necessary for you to communicate the moves the other two of us make to each other; it may be that in this particular description the imaginary space of one of the participants is not so close to that of the rest of us; but I don't think you can rule out the possibility that both you and the computer are playing at this point, particularly as you've clearly defined what part each of you has in play: the computer makes all the decisions regarding NPC actions, and you adjudicate outcomes.

I find it fascinating.

I don't suppose this will be resolved this time around; thanks to all for your input. If someone wants to respond to this suggestion (both the computer and the referee being independent players in Callan's proposed scenario), feel free to do so, but I think we've about run this as far as it will go this time.

--M. J. Young

[Callan S.]

It's no different.

If I were playing with two other real life people and one spoke only a few words here and there and the other gave far more input and I likewise. Basically the bar has been raised to the extent that the few words aren't zero input, but in practical terms they may as well be. If it were a book, he wouldn't be listed as co-author.

If you want interesting, think of two AI's roleplaying with each other and you, while your only able to contribute a bit (someone keeps ringing you up with important calls, say).

I'll wrap up on this. If I were RPing with someone who only whispered a few words every five minutes or so about nothing much, while another person as GM had submitted a rich multimedia presentation, I'd say I'm exchanging with the GM far more. Significantly so as opposed to a few words every five minutes. The only difference is that once the multimedia presentation 'runs out' I can continue to ongoingly exchange with the other player. I know which I'll remember most though.

[Andrew Martin]

One of the big arguments regarding the lack of SIS in CRPG is that you can't really negotiate with the AI that runs the game. That may arguably be a limit of current technology; and yet it may arguably be false.

I have argued elsewhere that a CRPG might be regarded an RPG on the basis of the computer as a player with limited ability sharing an imaginary space.

Remember that each participant in the game session has an Imaginary Space and that they coordinate changes in their own and other's Imaginary Space by communicating information to each other, so the total is called a Shared Imaginary Space. The computer program in the CRPG effectively has a Imaginary Space (or model) and the computer program communicates this through screen and receieves feedback through the keyboard.

If the computer program did not maintain an functional equivalent to an Imaginary Space (or internal model), it would be broken and unable to function in any significant way.

For example, a computer program that plays Noughts and Crosses with it's user has a model of a Noughts and Crosses board, has Noughts and Crosses marks and knows how to mark (and maybe unmark) cells with Noughts or Crosses. The better programs will have negotiation, allowing the taking back of moves and replaying, going backwards and forwards in imaginary game or system time. Similarly for computer RPGs, which maintain a internal model of the setting, characters, situation and so on, and the program can display or communicate it's internal model and it has ways of negotiating with it's users or players.

[simon_hibbs]

Similarly for computer RPGs, which maintain a internal model of the setting, characters, situation and so on, and the program can display or communicate it's internal model and it has ways of negotiating with it's users or players.

That's the way I see it too. That negotiation may be heavily constrained in it's scope, but it's still there.

Simon Hibbs

[Blankshield]

(sorry for not weighing in sooner; I was a tad swamped before and at Gencon)

MJ et al. - I think we're going to have to just disagree and leave it at that.  I think that computers as they are today are fundamentally incapable of being players in an RPG and nothing I've seen through these threads makes me inclined to change my mind.  I will attempt to articulate why I hold that position, though.

A fair amount of weight has been given to the computer's memory as analogous to human memory - in effect claiming that the current game state that a computer holds is an imagined space.  I don't particularly buy that, primarily because the computer cannot change it without human input of some kind.  The game state in the computer's memory is static, and will never change until there is a human to introduce change to it, at an absolute minimum, clicking the "play" button.  A human can (and does) modify their imagined space just about every time they "visit" it.  

A computer cannot "imagine".  It can read stored data, it can write new data.  A properly equiped computer could read The Lord of the Rings, but fundamentally all it would really be doing is making a copy of it.  Lacking inputs, a computer is inert.  Lacking inputs, a human still thinks and imagines (and if sensory deprivation studies are any indication, eventally goes mad).

If you took 1000 computers and loaded any popular CRPG and removed the randomizer - such that the die rolls in NWN were always exactly the same, for example - those 1000 computers would play the game exactly the same way given the same inputs.  I firmly believe (but cannot prove) it would be impossible to find 2 people that would do the same.  With the computer, there is no individual; there is only an algorithm that provides information based on input.

Finally, and this is one of the most telling points for me, the computer is incapable of assigning value in and of itself.  Fed algorithms and data it is capable of providing an excellent illusion of doing so, but ultimately the computer relies completely on external values.

Over these threads, at least a few times, the suggestion that "humans are just really complicated computers" has been made, and that argument could be used to refute the above by saying that people react exactly the same way as computers do, just with much more complex algorithms that we don't understand (yet) governing their choices.  This is probably where the fundamental difference in viewpoint lies, because I don't buy that view at all.  Or perhaps it would be more accurate to say that if that theory is correct, then there is a line (a fuzzy line, probably) that is crossed between a purely reactive set of algorithms and a true intelligence.  Computers have not crossed that line.

Boiling down my argument to it's fundamentals, I would say: One of several charactaristics required for an activity to be called role-playing is social interaction with another intelligence.  Computers are not an intelligence.

Hope that gives a straightforward view of my position on the whole CRPG thing.

James

[M. J. Young]

Over these threads, at least a few times, the suggestion that "humans are just really complicated computers" has been made, and that argument could be used to refute the above by saying that people react exactly the same way as computers do, just with much more complex algorithms that we don't understand (yet) governing their choices.  This is probably where the fundamental difference in viewpoint lies, because I don't buy that view at all.  Or perhaps it would be more accurate to say that if that theory is correct, then there is a line (a fuzzy line, probably) that is crossed between a purely reactive set of algorithms and a true intelligence.  Computers have not crossed that line.

James--I think I attempted to put that idea the other way around. I didn't say that humans are merely highly advanced computers, but that computers are extremely unsophisticated people.

Otherwise, I can see your point. What concerns me more is that if such a line exists, it is admittedly fuzzy, and we won't really know when we've crossed it. My argument in part has been that once we're on the other side of it, we'll be looking back at advances being made now as "where it all began". That causes me to wonder whether we might really have intelligence in computers in the present, just not very sophisticated intelligence.

Thanks for your input.

--M. J. Young

[Andrew Martin]

A fair amount of weight has been given to the computer's memory as analogous to human memory - in effect claiming that the current game state that a computer holds is an imagined space.  I don't particularly buy that, primarily because the computer cannot change it without human input of some kind.  The game state in the computer's memory is static, and will never change until there is a human to introduce change to it, at an absolute minimum, clicking the "play" button.

Consider a computer controlling a vehicle or a airplane. Here's one example from 1998: http://humane.sourceforge.net/unpublished/piper_club_offense.html The computer software piloted a model aircraft from one side of the Atlantic to the other side with no human intervention. Current autopilot software in modern passenger jets can pilot an aircraft from takeoff to touchdown.

A computer cannot "imagine".

But computer software can have an internal model, apply changes to it from sensors like keyboards, air pressure, GPS, and so on, and can read out changes from the model to the real world, through video screen, hydraulic or pneumatic actuators, electric motors and so on. Here's a link to a pile of sites on Autonomous underwater vehicles: http://transit-port.net/Lists/AUVs.Org.html

I firmly believe (but cannot prove) it would be impossible to find 2 people that would do the same.

Consider a well written D&D module, played through with players who play in the illusionist style. Wouldn't the end result be much the same? From personal experience, with playing AD&D and WEG TORG modules as player and GM, I've achieved similar experiences. :)

Finally, and this is one of the most telling points for me, the computer is incapable of assigning value in and of itself.  Fed algorithms and data it is capable of providing an excellent illusion of doing so, but ultimately the computer relies completely on external values.

Can't the same thing be said about a human baby? Does a baby have altruism, can it write poetry? :)

Boiling down my argument to it's fundamentals, I would say: One of several characteristics required for an activity to be called role-playing is social interaction with another intelligence.  Computers are not an intelligence.

What if roleplaying was just social interaction? Then an intelligence that was limited to communication through pictures and text on a computer screen, typed messages using a keyboard, and gestures using a mouse, is roleplaying. After all, isn't that roleplaying by email? :)

[simon_hibbs]

Boiling down my argument to it's fundamentals, I would say: One of several charactaristics required for an activity to be called role-playing is social interaction with another intelligence.  Computers are not an intelligence.

I suppose I look at it form a more metaphysical point of view. Let's imagine a situation similar to the turing test.

Suppose we were observers, but could only observer what one of the participants in a 'roleplaying game' was doing and saying. I think it would be fairly streightforward to say whether that person was playing a roleplaying game or not. If they are controlling a character, and exploring an imaginary space through the medium of the character, then I'd be confident in saying that they are playing a roleplaying game. Conversely, it seems to me that it would be impossible for you to be able to say what the person is doing. You'd only be able to determine whether they were playing an RPG or not if you could see beyond the screens and see who or what they were interacting with.

Similarly if we met someone at a convention and they told a typical RPG gamer story "Then my Paladin decided to charge into the Orcs to save the princess, but at that moment...". I'd have no problem saying that this person was recounting a story from a roleplaying game he'd played. You wuld be completely unable to do so without asking questions about the context of the action - was this a CRPG? Was it a solo game book? You would only be able to say whether this person had ever played what you call an RPG by examining the context in which he had played the game.

Would yo be able to say anything meaningful about what he'd been doing?


Simon Hibbs

[contracycle]

Otherwise, I can see your point. What concerns me more is that if such a line exists, it is admittedly fuzzy, and we won't really know when we've crossed it. My argument in part has been that once we're on the other side of it, we'll be looking back at advances being made now as "where it all began". That causes me to wonder whether we might really have intelligence in computers in the present, just not very sophisticated intelligence.

Yes but... this still seems wrong to me.  I don't think that there is a serious essential, ontological difference between computers and human brains but they are very different, erm, beasts.

I mentioned before that a computer is not alive, does not reproduce, has no history of predation and has no fight/flight reflex.  Can a computer have fun?  I seriously doubt it, becuase I think fun is an expression of fight/flight.  Can a computer have an imaginary space?  Well, yes and no, in that you can build a programme that contains a 3 or 4 dimensional model much like it would (probably) appear in a human brain.  But no inasmuch as our brains, as evolved beings, are heavily related to our sensor suite and interolating the data so received; building "visual" models of the internal world is something it does auytomaticall and inherently.  As I remarked a little while ago, the visualisation of a spoken description is so fast that the distinction between the description and the visualisation is nerarly nil.

The nearest thing we build to this sort of automaticity of IS generation is the AI's flying drones and the UAV's above.  These are our nearest analogs to the functions of living brains coordinating an (insect-like) organism.  I don't think any of these apply to a computer game that produces images for a player, executes scripts or rolls virtual dice.

Such a machine would more properly be an assistant.  It is filling the place that would be taken if you had an army of servants at your beck and call to assist you in visualising thye game or caryring out game functions. If you had a flunky who rolled your dice for you, would that flunky really be playing the game?  If you had a gofer to bring over images from the graphic library, would the gofer be a participant?  I don;t think.

Again, while I don;t think there is an essential difference between computers and brains, I do think there is a sufficient actual, quantitative difference top say the computer is an aid to the game rather than a player.

[simon_hibbs]

If you had a flunky who rolled your dice for you, would that flunky really be playing the game?  If you had a gofer to bring over images from the graphic library, would the gofer be a participant?  I don;t think.

You're not even remotely comparing like with like. Let's look at what (I think) you actualy need to be able to do to play a game:

1. Hold an internale representation of the game stat (in RPGs, the imaginary space).
2. Accept input from other participants.
3. Make decisions based on this input and the stored representation of the game, to further in-game goals.
4. Produce output to other perticipants and update the internal representation.

The dice roling assistant and the gopher from your examples don't even do one of these things, but computers can do all of them.


Simon Hibbs