[MUD-Dev] [decentralization] Re: Reputation device (fwd)

[Daniel.Harman at barclayscapital.com]

From: J C Lawrence [mailto:claw at kanga.nu] quoting "jito23"
<jito at eccosys.com>

> Quality here is defined as the likelyhood of buying something of
> lower quality and marked, or of someone lying. What he shows is
> that onces people develop positive reputations, they continue to
> try to keep them while negative reputation tracking encourages
> reseting your nym. Negative reputation SOUNDS good so people try
> not to get it at the beginning, but what really affects our
> behavior long term is positive feedback.
 
> His research is very thorough using a variety of time-lines and
> techniques in a laboratory with real people an a mocked up B2B
> exchange.

That's fascinating, as it strikes me that implementing a positive
system is far easier and less open to abuse than a negative one. Of
course if you want to limit it such that a player can only
positively recommend another once (perhaps per period of time), then
you have a fair amount of data (2n^2 if you want bi-directional
relationships I believe). Of course its likely to be very patchey
data, so run length encoding on a per character basis could work
wonders, and would only need to be decompressed and checked if they
wanted to vote for someone. If the data is really sparse, then per
character lists would probably be fine too.

What I don't fully understand is how the motivation to behave well
is enforced if you can't negatively mark people. Does positive rep
decay over time?

Although we don't have his paper, it does seem that his model
differs slightly from a MUD. Firstly it would seem that the
transactional cost of changing name is lower in his system - after
all, they haven't had to level a character up over however many real
life days. In fact couldn't character time investment be considered
semi-analogous to positive rep? Both represent an investment that
the owner will wish to maintain.

It would be interesting to know how many people who have stopped
using a character due to bad reputation, actually delete that
character as opposed to just letting it lie low for a few months, so
that they can return and proclaim they bought the account on
ebay. If my suspicion that most don't like to delete the actual
character is true, then limiting the number of characters on a
commerical game to one would also reinforce good behaviour as a new
account would have to be purchased in order to keep the infamous
one.

If that seems too limited, then perhaps all characters belonging to
a single player should have a DNA signature linking them as a
family. This could even be tied into an ingame investigation
mechanism. Maybe I'm projecting an amoral standpoint of my own, but
isn't it the fear of the consequences of being identified and
punished that prevents crime, not concern over performing the
dastedly deed.

I also think there is some mileage in having group based reputations
as a self regulating mechanism. If the personal reputations of
players directly impacts other players in the group, then you have
an incentive to behave well, and they have an incentive to exclude
you if you don't. This mechanic could be mapped onto guild
structures. Implementation wise, characters could have a base
personal reputation, which along with all guild members is averaged
into that guilds reputation. This average is then used to create a
reputation modifier that is then stacked onto the rep of every
character in the guild. It would be necessary to store both stacked
repuation modifiers (representing all guild memberships of a
character) and the individuals reputation. Furthermore, rather than
taking a pure average to create the guild reputation modifier, it
might be nice that a particularly bad reputation had a
disproportionately negative impact on the guilds reputation thus
encouraging swift exclusion.

Of course so far this has only addressed a single reputation figure
per character. More advanced would be a faction type system that
recognised that a bad reputation with orcs gives good reputation
with humans for example. My example is bad though as I have
integrated player run reputation with an npc group reputation, which
may not be a good plan.

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