[MUD-Dev] Something in the water

[rayzam]

----- Original Message -----
From: "Sean Kelly" <sean at ffwd.cx>
> From: "J C Lawrence" <claw at 2wire.com>
>> Caliban Tiresias Darklock <caliban at darklock.com> wrote:
>>> On Tue, 24 Jul 2001 10:59:44 -0400, John Hopson
>>> <jhopson at nc.rr.com> wrote:

>> The problem is that implementation is an automatic, or at least
>> computable system requires that the item be subject to
>> determinable metrics (ie it has to be able to be measured), and
>> there lieth the rub.

>>   How can you mechanically determine when RP is occuring?

> Parse what the player is saying, and if it includes "prithee",
> "thine", "forsooth", or "liege" then they're roleplaying ;o)

>> Or, if the problem is not subject to mechanical determination:

>>   What social systems can you implement which are variously
>>   resistant to gaming/overt_manipulation to measure RP?

>> Historically various things have been tried ranging from counting
>> the number and rate of poses and emotes, to having players award
>> each RP points.  The grinding point is that all such systems only
>> work reliably when the player base explicitly colludes with the
>> system

> I can't concieve of a system implementable today that wouldn't.
> What constitutes roleplaying depends not only on what you are
> doing, but the context you're doing it in.  I can't imagine trying
> to code AI to "detect roleplaying."  Also, there is no universal
> defintion for "roleplaying."  Even among self-avowed hardcore
> roleplayers there is disagreement on what constututes roleplaying.
> If policing by real people doesn't work, I can't imagine trying to
> do it in code.

Seems like it's an idea that could take cues from animal behavior
studies and history. If you can judge a rat's 'depression' level by
counting lip smacks, nibbles, nose rubbings, etc; it should be
possible to do something similar for RP. The one caveat is that you
have to reduce the sample set of behaviors to be tracked and
tested. Most RP tends to revolve around free-form emoting. If the
sample set is limited in scope, and the behavioral trains are
limited in length but depend on sequences, an inherently limited
form of automatic tracking could be done.

    Example: 20 emotes including sad, cry, hug, kiss, wink, nudge,
    comfort, chuckle, sniffle

    Boffo: sad, cry
    Billy then uses comfort and hug.
    Billy and Boffo talk.
    Billy says something else.
    Billy winks.
    Boffo chuckles, then sniffles.


Note this doesn't actually track what they say, they could be
talking about the NY Giants beating the Eagles yet again. However, a
behaviorist cannot udnerstand the communication between members of a
species and instead use the various gestures to make
approximates. Thus, Boffo is sad about something. Billy doesn't like
Boffo's distress and tries to soothe Boffo [of course this is
anthropomorphising when it comes to animal behavior, but that's
probably okay when dealing with people :)]. They talk about whatever
is making Boffo sad. Billy tells a joke. Boffo feels better.

By taking a limited set of emotes and arranging them in a large
variety of sequences [single, duo, multiple performers too], it
would be possible to judge these events and instances.

  Problem 1: it can be gamed. If the sequences are known, or if
  enough time passes, players can find out what sequences award
  RP. Possible solution: variable reward ratios [it becomes tougher
  to determine what sequences give awards with this], obfuscation
  [don't show RP as points, just as levels that increase/decrease
  over time and are described in words], or limit repetitious
  rewards by time [similar to various systems implemented for
  macroing in skills-based improvement systems]. Each possible
  solution still has inherent problems.

  Problem 2: Once the sequences are known, even with any of those
  possible solutions, the sequences may be used one after the
  other. Thus, an individual sequence would make sense, but 2
  different ones performed back to back wouldn't make sense in the
  aggregate.

If emotion could be assigned to the individuals in these sequences,
along with other events affecting the character, then emotion
changes over times can be used as an overarching guide to the
behavioral sequences. So going from hot anger to hunger is weighted
very low, whereas going from conflict to pleasure or conflict to
anger would be weighted higher. Track this over time and compare it
to others over time, and you might end up with a comparative RP
usage graph for your player population. Then use the
comparative/relative usages to determine RP.

The limitation is describing and entering in the various sequences.
Saying something, winking, nudging, and chuckling may work with
another player who could be performing a variety of emotive
responses. However, it shouldn't work at all against a plant or a
lamppost. This is why starting with a limited set of emotes is a
requirement. Simplify the problem.

    rayzam
    www.retromud.org

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