[MUD-Dev] RE: Understanding Simulation

[hart.s at attbi.com]

[Kwon Ekstrom]

> don't discount that over the course of millions of years that
> there have been an uncountable number of interactions between an
> uncountable number of individual organisms from microbes to the
> dinosaurs.  Not to mention an unknown number of environmental
> factors.

If you're pointing out that we don't know precisely what creates the
stabilities present in the real world, you're right. We don't know
what creates the instabilities, either.

The original argument had two thrusts - one, that vast instability
isn't inherent to ecologies, and two, that some degree of
instability is potentially tolerable, particularly when the
instability has low functional impact, e.g. when the function of an
individual NPC to players persists beyond its death, in other,
functionally equivalent NPCs.

This latter point is really important - it says that numerical
instabilities in the model don't have to be threats to the
functioning of the game. Note that I said "don't have to be -" they
could be, there is just no overarching reason why they have to
be. You might know such a reason, but I don't think I've heard it.

> There are simply too many factors to take into account for a true
> simulation, in order to handle a reasonable amount of factors
> certain inconsistencies arise.  These inconsistencies must be
> dealt with by external factors either constant, random, or
> calculated

Absolutely. If one wants to handle extinctions in a graceful way,
she/he'll probably have to figure out how to do it rather than just
throwing a grab bag of ideas at it. This task in particular isn't so
hard to think about - you respawn the species or add others,
explicitly or as part of some other mechanism (if you simulated
speciations and migrations you might get it "for free," probably in
the same sense that you get a free book-bag from PBS for pledging
more than the book-bag is worth).

I don't really believe in "true" simulation, you just simulate one
set of things or another. Bringing in more details isn't necessarily
helpful for all problems, although it can be. The world is not ideal
for all purposes, so even if it were possible to simulate it
perfectly it might not be desirable.

> A "simulation" is not inherently instable, rather it's quite easy
> to sustain a closed simulation. However a MU* environment is not
> closed.  Players are quite ravenous consumers and tend to produce
> quite a bit less than they consume.  Even when given appropriate
> tools for production.

Players can't drive rabbits extinct if you don't give them swords,
or if you give them swords but there are too many rabbits to kill,
or if the rabbits reproduce more quickly than they can be killed...
Even if the players have swords and could theoretically kill the
rabbits, but it would just take a lot of time and not have much
payoff (not even making you angry or getting them recognition), they
probably wouldn't do it.

Players are only part of the system insofar as you let them be.  You
may have trouble predicting their choices, but you also have trouble
predicting the output of MUD-ubiquitous "die rolls" as well. In both
cases the author decides implicitly, by what he writes, what effect
the unpredictable bits might possibly have.

On the other hand, it's hard to predict what people and die rolls
will do, so I suppose it might make sense to treat them as
interventions from elsewhere. But unless you are referring to
people's ability to exploit buffer overflows or something, they
aren't going to simply eat your MUD alive unless you let them.

I've heard talk of oversupply linked to crafting systems, so not
everyone agrees that players will not produce anything. If nothing
else, they do produce monster corpses. And I don't know if you need
to be reminded that there are whole long- lived games predicated on
players making objects and friendships.

> Most negative feedback loops don't work on players since they can
> simply "move on" either to another part of the realm or to another
> server.

I remember this argument and liked it.

It works because we aren't comfortable jerking around players to
make them leave so the population of rabbits will recover. It's a
good point.

But I don't see any reason why using players as predators in this
way is essential. You could even use their preferences
interactively. You could let players be the "tie- breakers," as a
population deciding whether the goblins or trolls won, or whether
the rabbit population fell below that critical level that makes the
ecosystem of the frozen north collapse. You could make the tools for
affecting the populations more effectively specific to people who
you are more comfortable giving decisions to. In a more boring line,
you could make the players functionally unimportant to the system,
e.g. if there were 3 billion rabbits growing on a strict logistic
growth curve calculated per-second, and players could only kill one
rabbit per second. Unless you're getting a very large number of
players that would be a model which is effectively immune to
whatever players could throw at it. One might question why not just
bother with spawns (especially since 3 billion rabbits is a lot of
overhead for most representations of a rabbit).

This is all just to express the opinion that players don't offer any
insurmountable difficulties in coexisting with population models.

> Yes we can make our own rules to an extent except unlike nature we
> have a single external factor which poses extreme threat to the
> stability of any system.  A factor which is mostly unpredictable
> and molds itself to any rules we may create...  Players

Not just to an extent (unless you count our inability to do things
which are logically or technologically impossible).

I don't see the hard distinction between players and human beings,
but it's not so important. Neither poses an _inherent_ threat to the
stability of systems. If the argument is that they are
unpredictable, note that there are systems which encompass
unpredictability and remain relatively predictable overall (the
position of an electron is pretty unpredictable, but the things I
say are pretty predictable.  Or; the die roll is unpredictable, but
the way DIKU combat works is very predictable).
 
If the argument is that they are rampant consumers of NPCs, then
there are ways to slow that down or curb it entirely. It's something
that happens in the game.

> The question is not whether you can make a simulation that is
> stable, it's whether you can make a simulation that is stable with
> a completely external factor as ravanous and unpredictable as a
> hoard of players.

Players' consumption of NPCs is dependent on the xp gain they get
from consuming them (etc.) but also on the tools you give them to
consume NPCs. Your players only select from the choices they have in
the game - they don't get to impose any impossible decision. If you
hate rabbits dying, you do have the option to remove all of the
swords from the game. Obviously you can't demand both that rabbits
never die and that players have full capacity and incentive to kill
rabbits, but you can make it work so that they can kill rabbits
constantly but the rabbits never go extinct, there are several ways.

> The fact that humans co-exist with species in the real world is
> not a valid argument because we aren't an external factor in the
> real world.

Because real world consequences aren't doable in MUDs, because
anyone can log out of a game, but you can't log out of the real
world.

OK, the problem is that some of the disputed behavior exists both in
the real world and in the game. In the real world, human beings
destroy animals at a fairly high rate, yet animals aren't all gone
(yet). In addition, there are many more of us in the real world than
in any MUD ever seen, we pollute, and so on. Yet there are still
animals. If I understand you correctly, the fact that players kill
critters - even just for fun - is taken to be evidence that you
can't have anything like an ecology model without animals all dying
to the players. But of course, the world is something vaguely like
an ecology model, and despite the fact that there are 6 billion
people, most of whom who have consumed animals at one time or
another, some multiple times a day, and despite the fact that there
are constant extinctions and so on, there are still more than enough
animals in the real world if judged against the standard of what
would be sufficient in a MUD.

The real ecosystem might possibly collapse in another hundred years
for all I know. But the fact that it's lasted a few thousand despite
people and a few hundred despite pretty egregious pollution and guns
and so on, well, that DOES suggest something about what is POSSIBLE
in a mud ecology model. It doesn't suggest that we know how it is
possible or how to do it, just that it is possible.

It certainly isn't an argument that people in MUDs act just like
people everywhere else. Making the argument fly only requires
relevant similarities. If you want to think that animals exist
because of some real-world restraint that is eliminated by the
ability to log out, go right ahead (but I don't buy it).

> Any decisions we make here we have to live with, we can't just log
> off and ignore them.

Sure, but that applies to aspects of player behavior that would
otherwise be controlled by the presence of consequences and the
absence of a logout to avoid them. Many other aspects are the same
either way. People are predictable in some respects and productive
in some respects. People's misbehavior in MUDs is all, in the big
picture, relatively minor stuff and not incomparable to people's
misbehavior in general. Some amount of bank- robbing works out OK
(not great) in the real world, and so on. Some of it gets prevented
by punishments, some doesn't.

However, I don't need to bank on that. It's practically impossible
to keep players from cussing in-game, although there are things you
can do (and similar techniques can be applied to keeping order in
the ecology, insofar as those techniques are useful.) Different for
ecologies - players simply can't do anything to the model that you
didn't at least implicitly allow, because you have full control of
the model, its computation and its interface to players.

Designing solutions to problems under multiple constraints is
difficult- e.g., if you wanted to have a game just like Everquest,
but with a 'real' ecology. In that case you will probably have to
decide what is more important than what, and compromise.

I'm not asserting that it is always possible to make an ecology
model work for any purpose in a game, regardless of what other
design decisions have been made. Just that there is a very, very
broad range of possible models and combinations of models with
player interaction.

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