Altering the SIS in CRPG

[simon_hibbs]

The problem with requiring emotional investment and the staking self-eteem in order to be considered a player, is that it excludes inteligences that do not exhibit these behaviours no matter how inteligent they are. There's no obvious (to me) requirement that inteligent, thinking beings must exhibit ths or that emotional response.

In fact many humans, such as those suffering from autism, or various forms of sociopathic or psychotic dissorders, are emotionaly and motivationaly incapable of these kinds of behaviours. Are they incapable of being considered players because of this, regardless of the level of their involvement in a game?


Simon Hibbs

[Doug Ruff]

I don't normally post from work, but I can't avoid getiing sucked into this discussion.

I'm going to add my $0.02, I believe that this is relevant to the discussion as a whole but feel free to challenge (or moderate) me if it ain't so!

1) We're not going to resolve the 'are computers self-aware' question and it's counterparts (I take 'do computers have a soul/true intelligence etc.' to be just variations of the same theme.) Even if it was a very big, very fast, very capable machine, we would never know.

2) For what it's worth, the same applies to other human beings as well.

3) That's why the notion of a 'Turing Test' is so useful - it's a functional definition of computer (and human!) 'intelligence'.

4) One of the problems with applying SIS to CRPGs is that the SIS is by definition imaginary - Imagination is one of those 'soul'-type qualities and is likely to derail any discussion of whether a computer can truly participate in a SIS.

5) So let's throw it out. What we need is a Turing Test for computer participation in a RPG (how to this would be a totally new thread, so I won't elaborate here.)

6) M.J. In your original post you gave a scenario where the (human) player could tell the computer that they wanted their player to do X instead of Y. This isn't a two-way negotiation if the computer just follows your instructions. But what if the computer responds that it thinks that Z is a much better option, so why not do that? Or offers option Z without player prompting? That would give me far more pause for thought.

7) Computers are already creative, insofar as they can produce outputs which are more than the sum of their inputs. This is a fundamental feature of 'neural network' programming, which I believe has its origins in 'Parallel Distributed Processing' theories way back (I think 1970's or 80's, sorry for the vagueness.) It is likely that a well-designed 'computer' of the future would be able to simulate RPG behaviour in the future. But we aren't there yet, even with more 'standard' conversation-based Turing Tests (I believe that there is an annual competition, the main prize is yet unclaimed.)

This isn't intended to be 'Doug's Rant' on the subject, I'm just typing this within a limited timescale (lunch break) and there is a lot of ground to cover.

Gotta go now, I'll read your comments with interest.

Regards,

Doug

[contracycle]

The problem with requiring emotional investment and the staking self-eteem in order to be considered a player, is that it excludes inteligences that do not exhibit these behaviours no matter how inteligent they are. There's no obvious (to me) requirement that inteligent, thinking beings must exhibit ths or that emotional response.

I'm not sure its possible to have what we think of as awareness and intelligence without emotion.  But my argument is not necessarily that this is the case, but rather that if a non-emotive intelligence existed, I predict it would not play games.  It would find them absurd, I think, rather than fun.

In fact many humans, such as those suffering from autism, or various forms of sociopathic or psychotic dissorders, are emotionaly and motivationaly incapable of these kinds of behaviours. Are they incapable of being considered players because of this, regardless of the level of their involvement in a game?

I think its an open question whether or not they can be involved in a game in the first place.  The daughter of a freind of mine is autistic, and she doesn't really treat me as any different to the furniture.  No eye contact, no recognition, nothing.  Its actually really eerie.  I don't believe I could get her into a game in the first place, her ability to relate to other humans is so limited.  But I'm hardly an expert on autism, so...

[Christopher Weeks]

...it seems to me that the computer is incapable of getting the same out of the game as the human player...

Doesn't relying on this mean that another person who gets something different (or "is incapable of getting the same") out of the game isn't really a player, either?

I'm not asserting that a sense of competition or comparison is actually necessary for play, but some of why I think the subjective experience of a game is IMO very very different for us and for (modern) computers.

It does seem like you spent three substantial paragraphs asserting that comparison is the key to play, so I'm trying to understand.  Are you merely citing comparison as one element that some players of some games pursue as a meaningful criterion and that some computers do not?  That seems like such a weak assertion that you must mean something else.  But I'm not getting it.  I do agree that to whatever extent the computer can be said to regard the game at all, it is different than how we do.  But, of course, how you and I regard the game may be substantially different too.

IMO, a computer will only enjoy, and choose, in the senses we mean, when it is self aware, that is, is an AI.

Imagine that we create software that is capable of playing (even in the weak sense) several different games.  No imagine that we set up some rules for the computer's behavior so that it acts as if it prefers some of the games over others.  We could even have it monitor which games and which opponents require more processing (to simulate our own desire for mental stimulation -- which I imagine is a survival trait) and base the acted desire on those statistics.  And then, we program it with the ability to bargain -- to do things that it doesn't "like" as well now, in order to get something it does "like" more later.  Maybe it could even keep track of which humans are likely to keep up their end of the bargain.

Now, I think that each of those layers of "personality" are pretty easy to implement on modern computers with modern knowledge.  We could end up with something at behaved like WOPPR from Wargames, as MJ discussed.  It can express choice, desire, preference, and enjoyment because of the way the system was established.

But we can all see that it's not really self-aware in any general sense.  Would something that sophisticated count for you?  If not, what would?  How autonomous or generally intelligent must it be?  I think that there are certain ennumerable behaviors that the system must exhibit (ignoring for now what those are) to count as a player.  Beyond those behaviors, I don't care what the system's evolutionary past is.

Then again, that's got to be taken as a programmed personality, not a computer's own decision-making capability. Further, we can never reach a point where we could distinguish what a computer does because of its programming and what it chooses to do for its own enjoyment--in one sense, there might not be a difference...

Agreed.  And I think a corollary question is whether a person is any different.  Is my personality programmed or spontaneous?  Am I really making decisions?  Again, there might not be a difference.  Ultimately, I don't think it matters, but some people do.

I think games are heavily related to our history of predation of which computers have none.

If we wholly engineer a creature in our cutting-edge bio lab of the plausibly near future by plugging bits of DNA together and that creature is close enough to us that no one argues that it's intelligent, and it can pretty clearly(?) play games with us, what does that say about history of predation?  Or wll you take things back far enough to suggest that any DNA-based life has a shared history of predation (which is certainly true, in a sense but pretty grasping-at-strawsish when related to complex psychological abilities, I think)?  Haven't we essentially programmed a player?

You've also noted the belief that computers lack sufficient complexity to play, so I realize that you can just cite that as the difference.  I'm just exploring this bit to expose whether we're agreeing and anything new arises from it.

Chris

[JamesSterrett]

Doesn't the "self-aware/can't enjoy" argument, however, boil back down to the "doesn't have a soul" argument?

[contracycle]

The point about comparison was intended to be something along these lines: an awful lot of the things we do are about declaring, exhibiting, or aquiring traits and abilities that make us a good mate for the procreation of the species.  IMO, that level of function is the underlying cause of gaming behaviour.  This activity is fun only because its not actually immediately survival based.

Being in a battle  is frightening, playing paintball is fun, even if most of the things you do are very nearly identical.  It's fun because the inputs and outputs are very similar, but the significance is very very different.  And just as humour can be argued to be the relase of tension, I argue that otherwise dangerous activities practicied in a non-dangerous environment are fun.  The reason they are fun is because the practice is useful and the organism benefits from the practice.  Games, IMO, are autodidactic behaviours.

IMO, game-playing is operational on much the same level as lust, fear and so forth.  These are all devices the organism uses to procreate itself, to trigger certain types of self-behaviour.  Partly because this practice contributes to sexual selection, partly because it contributes to predatory efficacy.

So now lets go back to our notional DNA-assemblage.  If we build a living organism that can procreate, run from predators, hunt its own food, I'd think it can play too, and we will probably see it play in its adolescent and earlier phases.  If it does not have those autonomous functions, it would more or less be a machine.  Or a plant, perhaps, at best.

I think a computer built such that:

It can express choice, desire, preference, and enjoyment because of the way the system was established.

... would rather be said to be imitating those functions.  It is engaging in behaviours that are interpreted by humans as indicating preference, but only becuase it is constructed to push our human buttons.

The brain is a late development in organic creatures.  By contrast, it's the first development in computers.  Our brain is, in my view, a computer whose purpose is to preserve the physical body.  Computers simply don't have those functions at all.  It has no sense of self preservation, no sense or programming to procreation, no adrenaline system, almost none of the equipment our bodies carry for survival purposes and which, IMO, are necessary for the experience of "fun".

[Blankshield]

I'm with Contracycle on this one, (not that my position was in doubt...) and as usual he's managed to express it better than I did.  It does suggest another way of getting across why I don't think computers are players, hopefully without getting too far into the soul/AI/'just a set of algorithms' debate again:

The people who are positing that computers are players are running from a functional definition of "playing an RPG" - which is to say "if something is engaged in the actions that we recognize as RPG behaviour, it is playing an RPG."  If that is how you define an RPG, then I guess you're right.  But I don't define it that way.

Those of us (I think - maybe it's just me) maintaining that computers can't be players are running from a more ontological definition of "playing an RPG".  Which is to say that I think why is a critical and necessary part of 'playing an RPG'.  To rephrase it slightly (and with apologies to Ron if I've mangled one of the core concepts of his model): Why have these computers sat down to play this game together?  What is their CA?  What is the computer getting out of it?  If you ask a hundred computers why they play RPG's will the most common answer be "because it's fun"?  I think the "why of it" is core to playing - until there is motivation and reason, then it is not play, it is just a mechanical process.  

If, on the other hand, you think that computers are capable of motivation and reason, then we're edging back to the soul/AI morass.

James

[Ron Edwards]

Hello,

I think everyone understands one another now.

Can we call this thread closed? I'd very much like to. It's time to ruminate, not to dig in one's heels and continue to repeat one's position.

Best,
Ron

[simon_hibbs]

I think everyone understands one another now.

Can we call this thread closed? I'd very much like to. It's time to ruminate, not to dig in one's heels and continue to repeat one's position.

I think so, I've nothing further to add on this topic for now.

Simon

[M. J. Young]

Gareth's objection may have more force than I first thought. It would appear that a computer could be capable of performing all the individual tasks of play, but that it would be impossible for a computer to have a creative agendum, because it doesn't act from a motivation expressed through its choices.

Of course, creative agendum does not require that we know what the player feels or thinks or wants; it requires that we observe behavior that appears to flow from what the player feels or thinks or wants. So it is not essential for the computer to have a specified motivation; it is only required that it exhibit conduct consistent with that motivation. On the other hand, it would seem that such conduct would always be programmed conduct, and thus not the consequence of what the computer "wants", so no matter how much it "acts" like it has an agendum, it doesn't.

Could it? I know that psychologists use role playing games to learn about their patients. I know that there are players who do the same thing--my wife calls it "psyching" the other participants, and there are some people with whom she will not play for that very reason (because she feels as if their actions are calculated toward manipulating her to reveal personal information about her feelings and attitudes that she does not wish to reveal, or that will be used later to manipulate her in real life). I would say that playing as a means of exploring the personalities of the other players is a creative agendum (let's not in this thread debate which, despite how interesting a question that is--we don't need two imponderables running around the same thread). Could a computer be programmed in such a manner that it makes its decisions based on an agendum of stimulating and analyzing player response, and so better understanding humans in general or individuals specifically? If so, would that be the line in the sand, the moment when computers cross from emulating gameplay to playing a game?

I am again reminded of Whopper (which again is a fictional representation). At times it appears to have a gamist agendum, the desire to prove that it can win. It's difficult to imagine how such an agendum could have been programmed into a computer. It seems rather that they would have programmed it with simulationist tendencies, to determine every possible win/loss strategy or outcome so as to determine the optimal strategy for any situation. But then, if the computer is using a self-learning routine (as was discussed for tic tac toe (a.k.a. naughts and crosses) and other games), doesn't it already have a built-in agendum to learn optimal play? Has such a computer crossed the line, because it has an agendum?

As for Gareth's arguments that gameplay is survival related, does it have to be? Could it be related to other needs or desires, such as understanding the player's place in the world? Although it hasn't been done, in terms of AI development wouldn't there be some value in teaching a computer to play a role playing game as a player, so that it would be able to explore how to be a human through feedback from someone who understands humans?

I suppose these may be unanswerable. In deference to Ron's desire to put the discussion to bed, I'll leave them unanswered for the moment. It does appear that we're pretty solidly divided between those who believe computers are already playing in a limited fashion, those who believe that computers are not playing, but will develop that capability as they advance, and those who believe that computers will never have the basic need to "play" and so can never be "players" in the only sense that matters.

And whether CRPGs really are role playing games or merely emulations of role playing games is now even less clear than it was before this thread began.

--M. J. Young

[contracycle]

I'll happily concede that my argument is not evidence based and is outside my expertise, so it is indeed quite possible that play is based on something other than survival behaviour.  But I do agree that we understand each others positions now and there is little more to be discussed.

[M. J. Young]

I'm tagging a closing footnote to this, because I just posted yesterday that I'm not sending any private messages until I find some spare time to clear out my sent box, and I just received a message regarding this thread, from Paul of Inner Circle. His position is that the computer is not a player, but merely a conduit for the game designer; however, he thinks the same of a referee who runs a module, and never considers the referee a "player" but merely an arbiter.

That appears as a distinct position in the discussion, which I otherwise consider closed.

Thanks to all for their time.

--M. J. Young