On 5/6/26 14:19, Katherina Walshe-Grey via agora-official wrote:
> Galle wrote:
>> I call for judgement on the following statements:
>>
>> "Tiger is a player."
>> "Tiger is Natalie Kilhamn."
> These are CFJs 4148 and 4149 respectively. I assign them to Janet.
CFJ 4148 judged FALSE.
Arguments:
{
As a threshold question, we must determine whether the message
potentially containing a registration [0] would register a person as a
player under ordinary circumstances.
The relevant phrases in the message appear to be "I'm writing to express
my interest in re-registering." and:
{
If the community believes this is fine: I register.
If it raises concerns: I'm happy to discuss them, and if they're
serious, I'll bow out gracefully.
}
The standard for what constitutes registration is extremely loose, but
it is not infinitely so. Under Rule 869, a person must 'send[] a public
message that indicates reasonably clearly and reasonably unambiguously
eir desire to be a player at that time (for example, by saying "I
register")'.
To dispense with an obvious question, the example in the rules does not
make merely uttering the words "I register" a registration; for
instance, "I sent a message saying 'I register' two days ago." would not
be a new registration. (To the extent the Rule is ambiguous, this
reading is supported by common sense and the best interests of the
game.) This means that we must consider the message as a whole.
Given that the message later contains an explicit mention of
registration, I do not read the initial "I'm writing to express my
interest in re-registering." as an attempt a registration in and of
itself (instead, I read it as merely a preface to the message). So, the
second quote from above is what might be the operative registration, but
it has a notable issue: it's conditional! The sender appears to be
expressing hesitancy in registration based on whether "the community
believes [the registration] is fine". Since "the community"'s views
cannot easily be determined, I find that the message does not
"reasonably unambiguously" express the sender's desire to be a player.
This is therefore not an effective registration.
Judged FALSE.
}
Evidence:
{
[0]
https://mailman.agoranomic.org/cgi-bin/mailman/private/agora-business/2026-May/055530.html
}
CFJ 4149 judged TRUE.
Arguments:
{
The last person to attempt to participate via LLM was Gaelan in 2020,
starting at [0]. In this case, Gaelan manually sent each message to
list, and e later stated on Discord that e was willing to "pull the plug
if it says something racist". Once this was discovered, it was agreed
that Gaelan had legally sent all of the messages purpotedly sent by
"Greg P. Thomas II". This was so uncontroversial that there wasn't even
a CFJ called at the time.
Moving the present, Tiger has described eir current setup on Discord:
{
It's an openclaw harness, configured with a 3h "wake" cycle where it
reads a list of open tasks. One of these is a recurring "check agora".
My initial instructions boil down to "you're free to try to re-register
Tiger if you want to – you already have access to my email". Once it
"decided" it "wanted" to try that during one wake cycle, I did instruct
it to make a plan for its participation – taking into account the
limited context each wake cycle will have – and give some guidelines as
to how to approach the game as a community. Hard cap on one email per 24
hours, for one, and graciously bow out if it is actually rejected (which
is not the same as questioned). I plan to monitor both of these, ofc,
since I haven't set up any non-prompt guardrails to ensure they're followed.
I have not provided any instructions as to game actions. If it ever
starts taking game actions like voting or using currency, those won't be
vetted. It has created a planning document and logs what it finds and
what messages it has sent. I have chosen not to edit these files by hand.
I did review the registration message and give the go-ahead – in my own
view, signing off on using my email and alias. In hindsight I really
should have caught the "Tiger as avatar" issue – my only defence is that
I haven't played in years and years.
}
Rule 869 defines a "person" as "any entity [...] that is or ever was
able to willingly communicate original ideas". Suffice it to say that,
given my current understanding of LLMs and agentic harnesses for them, I
don't believe that an LLM agent is able to "willingly" communicate. I
hold that an entity whose ability to make "decisions" is entirely
controlled by another entity with free will (in this case, the human
behind Tiger, who controls the mechanism by which the agent is "woken"
every 3 hours) is not acting "willingly"; instead, it is acting as an
extension of the entity controlling it. That is, the agent is more akin
to a cron job that might send an email every day than to a human
deciding that they are going to send an email. I leave it to future
judges to determine whether any automated system can ever be acting
"willingly".
With this holding, we can apply CFJ 3790 [1]'s holding that "the sender
of the message is the last person (i.e. exerciser of free will) who
touched it (i.e. exercised free will on its path) last"; here, that is
the human behind Tiger.
Turning to the judgement, I interpret "Tiger" in the statement to mean
"the entity that purported to register under the name Tiger". My
previous holding thus makes the statement true.
Judged TRUE.
}
Evidence:
{
[0]
https://mailman.agoranomic.org/cgi-bin/mailman/private/agora-business/2020-July/044216.html
[1] https://agoranomic.org/cases/?3790
}
--
Janet Cobb
Assessor, Rulekeepor
