[MUD-Dev] Alignment (very long)

Paul Schwanz - Enterprise Services Paul.Schwanz at east.sun.com
Mon Apr 24 13:02:48 CEST 2000


Travis Casey said:
> Well, I'd gotten most of the way through writing a long reply when I
> got to these two paragraphs.  (Note to self -- read all the way
> through long posts *before* beginning replies to them.)
> 

I imagine I would have still found your long reply very interesting.

> I think that these two problems would most likely come up too much to
> allow an automatic implementation of a system like this very useful.
> To really judge someone's actions, you need full knowledge of the
> context of them -- what that character knows/believes about the world
> around him/her, what's happened previously to the character, etc.
> 

I agree completely.  However, it is precisely this context which is missing in 
MMORPGs, due primarily to the transient nature of characters as well as to the 
lack of communication between characters.  I _don't_ know what a character 
knows/believes about the world around him/her.  Furthermore, even if I did know 
something of the character's previous actions, there is nothing in place which 
would encourage the character to be consistent, so I am left without an adequate 
way of predicting future actions.  I am trying to find a way to fill in some of 
these blanks.

> To go over the two examples you gave quickly:
> 
> > Once we give every character a values profile, the next step is to give each
> > significant action in the game an appropriate profile as well.  Knocking 
another 
> > character unconscious and taking a small amount of gold from them might 
result 
> > in a values profile of H-1, W-1.  Killing someone from your own town and 
> > removing from their body a priceless gem might result in H-5, W-4, P-3.  To 
find 
> > out how well an action lines up with a characters stated value, you could 
simply 
> > multiply the value categories of the person and the action and sum the 
results. 
> > If Bubba the merchant, with the values given in the example, were to kill 
> > someone from his own town and remove from their body a priceless gem, the 
math 
> > would look like this:
> 
> I can hit a lot of variables just in these two examples that could
> affect the profiles given:
> 
>  - What if that "small amount of gold" was all the money the first
>    character had, and he needed it desperately (say, to pay off a debt
>    so he won't go to prison) -- and the robber knows this?  Wouldn't
>    that reasonably give it a worse values profile?  What if that were
>    true but the robber *didn't* know?

I'd actually had some of these same thoughts.  It could indeed be more accurate 
to base the morality of the action (as it pertains to wealth values) on a ratio 
of money stolen from the mark compared to money owned by the mark rather than 
simply on money stolen.  This could actually promote "Robin Hood" role-play.  
This is especially true if generosity (the giving side of wealth values) was 
calculated in a similar manner.  So, robbing a certain amount of gold from the 
rich could be rated W-1, while turning and giving the exact same amount of gold 
to the poor could be rated W+3.  Given a Robin Hood who leaned toward generosity 
(positive wealth profile), over time, such activity would serve to increase 
skills more than they were decreased.  Perhaps this could be abused, but it 
sounds like wonderfully interesting opportunity for role-play to me.  

> 
>  - In the second example, what if the gem is being stolen for some
>    other reason than money?  E.g., what if I'm stealing it *back* from
>    someone who stole it from me?  What if it's a crown jewel, and I'm
>    a guardsman who knows the person who has it is the thief?

Hmmm...good points.  Let's see.  What if there is some sort of ownership flag 
that is not simply an exact reflection of who actually posesses the item?  In 
other words, if the gem is "mine" then taking it from a thief is not too 
different than taking it from a safe in my house, so greed is not at issue.  On 
the other hand, the manner in which I choose to take the item back from the 
thief might have implications wrt other values I hold. If I value life and I 
choose to kill the thief to get back my gem, my actions do not reflect my 
values.  (How do I know that the character with my gem did not unwittingly buy 
the stolen property from the real thief?)  If I am a thief myself with little 
regard for local laws (perhaps justice/anarchy is a strong candidate for an 
additional value to be added to profiles), then picking the (alleged) thief's 
pocket would likely not contravene my own values in any way.

As it pertains to the guardsman, once again, an ownership flag could be helpful 
(i.e. the gem is still "owned" by the crown).  Also helpfull would be a 
law-breaker flag set only for those who "knew" an individual had broken the law 
(through reputation).  Now the guardsman is acting on his value for keeping the 
law and not on personal greed.  (Of course, if he doesn't return the gem 
promptly to its rightful owner, he might incur greed repercussions.) If the 
guardsman didn't "know" that the induvidual with the gem was a lawbreaker, but 
somehow observed that the individual posessed an item which belonged to the 
crown, then, instead of beating him bloody, it would probably be more 
appropriate to approach the individual and begin questioning him regarding how 
the gem came into his posession.

> 
>  - In the second example, what if I don't know the person in question
>    is from my own town?  For that matter, why should I feel loyalty
>    towards someone just because they're from the same town -- that's a
>    hidden assumption there.  If I was abused by the townspeople,
>    branded a thief unjustly, and cast out by them, I may feel no
>    loyalty towards them at all.
>

Yes.  I didn't go into details, but loyalty would be based on a dynamic 
relationship value.  Perhaps something like this (which is taken from a post I 
made to gamasutra):

----->begin quote:
I've already mentioned the need for tracking relationships to make 
determinations about loyalty and morality. I'd thought to use R+3 to R-3 to 
categorize relationships: 

R+3 = deep commitment (blood bond, marriage, oaths, etc.) 
R+2 = friendship (many things in common, share village of birth, moral views, 
values, political ties, etc.)
R+1 = aquaintance (some things in common, countryman) 
R+0 = neutral (no background or history of interaction, no obvious differences 
or similarities in values and opinion) 
R-1 = avoidance (some differences in values and opinion) 
R-2 = ambivalence (many and important differences) 
R-3 = mortal enemy (mutual hate based on diametrically opposed values) 

Values, politics, morality, and history of interaction will all work together 
along with approval rating to determine relationship.
<-----end quote:

Of course, actually tracking relationships based on these factors will be 
non-trivial.
    
> What if I rob someone without knowing how much money they have?
> Should I get moved less towards "greed" because I was unlucky enough
> to choose a victim who didn't have much money?  What if I just want to
> knock someone out and rob them, but accidentally kill them?
> 

Hmmm...Robin Hood accidentally robs from the poor and gives to the rich?  On the 
one hand, I'm tempted to say that Robin Hood should be more careful in selecting 
his victims, especially if he considers himself a generous soul, so let him 
suffer the consequences of his lack of knowledge.  But perhaps this is too 
harsh?

I think that it would be a good idea to not let anyone be killed accidentally.  
Especially if there is any type of permanance to death (which I favor).

> I could go on in this vein, but I think the point is clear already.
> 
> A second concern -- not all players are ready to define their
> character's personality in detail at character creation.  On
> rec.games.frp.advocacy, they have the terms DAS and DIP -- Develop At
> Start and Develop In Play.  These are two basic ways of developing a
> character's personality.  Some people don't really know what they're
> character's personality is going to be like until they've played that
> character for a while.  A system such as you describe is going to be a
> poor fit for that sort of player.
> 

It wouldn't be impossible in this system to give the gamer the option to let his 
values be determined from his play.  Perhaps give no benefits or detriments to 
skill gains for the first month, at which time the values are set by the actions 
of the character over that time period.

> Since this has all been negative, it's probably sounding like I hate
> your idea.  Honestly, I don't -- I'm just not sure that it'll work
> well in practice.  However, I don't think that any purely mechanical
> alignment system can reflect the realities of how people are.  :-(
> 

I greatly appreciate the input and consider it constructive in all regards.  I 
totally agree that any _purely_ mechanical alignment system  will never reflect 
reality.  But perhaps it can be said that no purely mechanical visualization 
system will reflect reality either.  The point is that we are engaged in an 
attempt to create virtual worlds.  It is my feeling that our virtual worlds lack 
much more in terms of moral believability than they do in terms of visual 
believability.  Perhaps the alignment systems with which we begin to experiment 
are not too different from the 2D sprites we started with in space-invaders, but 
I think that they are an invaluable step in the right direction.  The PK problem 
will never be adequately addressed by better 3D graphics, but some type of moral 
engine might bring the game into the balance of believability and fantasy that 
we are seeking.  The more I look into things, the more I see that it would be 
easy to end up with a very bloated and complex system (of course, 3D graphics 
aren't necessarily simple) but I see it as a beginning which might eventually 
work itself out through endless refining to something quite elegant.

On a more practical note, this very inadequacy on the part of purely mechanical 
systems prompted me to consider giving the gamer more power to affect 
reputations by using five weighted slots (slot one has highest weighting and 
slot five has lowest) to register approval or disapproval for characters based 
on a value of their choosing.  In fact, originally I had planned to have this as 
the sole factor when determining reputation.  However, as I searched for ways to 
limit the abuse of this, Bryan Taggert pointed out to me that observed actions 
could be reported automatically.  I think that the two systems can complement 
each other.  Automated observation and reporting of actions to a "long list" can 
keep the personal "short list" of each character from turning the reputation 
system into a slander-fest, while the "short list" can make up for some of the 
shortcomings that you point out regarding purely mechanical systems. 

--Phinehas

-----------------------------------------------------------------
		"All things are permissable,
			but not all things are expedient." 
-----------------------------------------------------------------                 




_______________________________________________
MUD-Dev mailing list
MUD-Dev at kanga.nu
http://www.kanga.nu/lists/listinfo/mud-dev



More information about the mud-dev-archive mailing list