Hi, I'll give my 2 cents. kiako via agora-discussion [2026-09-01 12:02]: > Hi all, > > Given the 3(?) recent cases in Agora of LLMs being employed by players > (automated and otherwise), the Discord's been discussing if we need a > clear stance on LLMs and "agents" in Agora. To keep the conversation > archived & accessible, I'm moving it to the list. > > I'd appreciate if those with vested interest in Agora could chip in > their opinions.
I have more complex thoughts and feelings about LLMs in general that I won't try to explain, because I'll fail. But I can say that I've always been fascinated with AI (as in, long before the current LLM-based cycle). Seeing a technology that allows us to genuinely experiment with building autonomous cognitive agents (ontological discussions nonwithstanding) is exciting. And I was, until this thread at least, eager to see one genuinely engage with Agora. Differently from some other opinions, I think that these systems are currently good enough to have interesting behaviors – I didn't think so at the start of the year, but things are indeed moving fast, no matter how much I don't want to admit it (maybe I'm wrong and someone could convince me). That said, I recognize the concerns raised in this thread. Having one ostensibly autonomous system play the game could be an interesting experiment to the rest of us, but that of course depends on each person's taste, so all I can say is that I would like to see it tried at least once, as an experiment. As for LLMs being used to generate text for public consumption on behalf of a human player, I think the same as I think for all LLM-generated stuff: it's the person's responsability to make it good. If you are just generating text and sending it to other people, without reading, you are being an antisocial jerk. But, if you are engaging with the text, I can't reasonably say that that is any different to using any other number of tools and techniques for writing. The important point here is authorship: if you generate-and-send, you are not the author. But, if you are the author, all is fine, even if you use an LLM. As for *private* LLM use, as in “generating artifacts that won't see the light of day”, well… that's not my business. Just note that that is actually harder than it seems. Your code can have errors, and those errors can be seen by other people. Do you want you errors, that affect other people, to be honest mistakes, or negligence? As a side note: I don't know what a headmate is, and I feel like I'm out of the discussion on “singlets” and whatever. To my mind, collective player in Agora are persons only in a agoran-legal sense. Come on. People (in the ordinary sense) are those entities that have sufficiently complex/similar to ours subjective experience – so humans, other animals, etc. We should care, in the meta-game, about the feelings of those kinds of people that engage with the game, not the abstract feelings of collectives (note that I include individuals in the basis of a Person that do have feelings) [1]. Anyway, I would conclude by saying that we shouldn't regulate prematurely, if no one is already offended by the happenings. Like, if someone is getting triggered by the already-existing LLM presence, then let's engage with the issue. But, if not, I would like to see things play out. Let's have some mechanism ready to halt things if needed (remember, ultimately, we have the Banned switch). [1]: If I'm being insensitive to someone, I will apologize, and do please let me know. -- juan

