If anyone is disqualified, I think it would be reasonable to expect a detailed explanation on a line-by-line basis of what prompted the decision. There may need to be a back-and-forth with the author. It would also probably be best to base the decision on feedback from a range of knowledgeable people, which would need to be collected and studied. I would suggest organizers then be prepared to accept some degree of public scrutiny for the decision as well.
I agree with this sentiment.
Well, the āvocal minorityā at the moment appears to be just me. I donāt think there has historically been a consensus, though. There wouldnāt have been hundreds of messages if there had been. In any case, Iām not really suggesting that the policy be reversed. At this point, it would probably be foolhardy for competition organizers to act in contravention to the clear wishes of the community. Iām merely suggesting we prepare for the fallout. If discussions of the general principle have been heated, it could be much worse when itās about specific games. People need to be ready. Expectations need to be set and procedures put in place.
Iām also not sure I understand the point about āedge casesā. There are bound to be difficult decisions sooner or later. Itās worth trying to determine how they will be handled as specifically as possible. Also, give that anti-AI sentiment is quite stringent in some quarters, I suspect there are people who will demand a stricter application of the rules than others are comfortable with. That will need to be navigated as well.
I like to think I might have offered some original insight.
Weāve had IFComp disqualifications before. Itās generally announced as a one-line post saying what happened and what rule was the basis of the decision. Turning it into a public pick-apart-the-evidence-fest is a rookie mistake, and IFComp is not run by rookies.
I would describe that as running the competition like a private fiefdom, which is particularly worrying since youāre significantly expanding the grounds for disqualification to include something that is highly ambiguous. As far as I recall, previous disqualifications have been for clear-cut things like using copyrighted material or having already published the game elsewhere.
This is a big ask, especially when, as far as Iāve seen, every piece of IF that is known to have used AI art/text has been glaringly obvious, and any non-glaringly-obvious cases would likely be impossible to definitively identify no matter how many interviews are done or how much public scrutiny is directed at them.
More like an oligarchy. Thereās plenty of people on the committee. But I take your point. Thing is, Iām comfortable with that. I like the work theyāve done so far. And even if I didnāt, so what? They arenāt a public institution, they arenāt funded by taxpayer dollars, and they have no real obligation to the people who choose to use the service they offer at no charge to us.
I guess I donāt understand. IFComp has always been a venue for humans to judge other humansā work and was in operation before AI was widely available. The only people who would argue against that rule are people who think they will do better than other historically skilled writers by using AI. Which is, to be frank, cheating.
Is the fear that a human-authored work is going to be mistaken for AI? I donāt think that will happen.
Is the fear that someone wants to kind of use AI but slip it in as an āedge caseā and have it judged along with the other entries that didnāt use AI? Thatās obviously against this rule. But if you can fool people, youāve won. Is that the flag people are wanting to plant? Fool everyone? Pass the Turing test?
We have the same rule here on the forum. And yes, you can call if a fiefdom or an oligarchy if you want, but everyone agrees to the terms of service. AI generated material canāt be posted with any kind of intent to fool people into thinking they are human.
I think the point of the IFComp rule is so they have a rule that is transparent and unambiguous and can disqualify an AI entry and point to a rule instead of fostering any confusion.
TL;DR: Do the work. If you canāt write, collaborate with a writer. If you want to create AI games, do it and donāt enter them in a competition where the point is for people to judge how good a human is at creating a game.
I would have to defer to someone with more experience on that matter. I know the Choice of Games folks have dedicated significant energy to detecting AI use so they may have some useful insight.
Itās still incumbent on IFTF to administer the competition in a fair and reasonable manner. If they donāt, the community can jettison the comp and start a new one. I donāt think some transparency is too much to ask.
This is how IFComp, Spring Thing, and numerous smaller comps have always been run. The balance is that people arenāt obligated to enter or to vote and will walk away if they feel itās being run poorly, as a decent chunk of people already expressed they would if AI use werenāt restricted. The fact that IFComp has a good reputation and a long history isnāt enough to keep people in line if theyāre unhappy. If they administer the new rules unfairly people will certainly make their displeasure known, especially since next yearās post-comp survey will undoubtedly ask about it.
Yes, absolutely. This is one of the several reasons I feel confident that the committee wonāt begin a reign of terror and start arbitrarily disqualifying entries in ways that leave the rest of us scratching our heads.
Are those questions meant for me? If so, then I donāt really know what to say except ānoā and that I think one of us must have misunderstood the other.
Quite right. If. Personally, Iād rather the organizers err on the side of aggression in enforcement, and I have faith that the process will be appropriate even if the exact details of every case arenāt aired for everyone to see.
Moreover though I donāt think itās worth spending so much time worrying about how a few ambiguous cases are going to be handled. We should (imo) not spend 90% of the time figuring out how to handle 1% (or less!) of entries. I submit that any game utilizing an AI parser or generative text in real time is going to be transparently obvious. Any game utilizing prewritten AI text will be quite obvious unless exceptional effort is taken to edit that text to disguise it. Art is a bit more complicated, butā¦ somehow I donāt think weāre going to be seeing stuff from top-tier models.
Without wishing to rehash previous debates at length (unless youād like me to), I think there are plenty of other reasons: difficulty of enforcement, potential for conflict, general dislike of banning things.
When AI was allowed and I chose not to use it, I canāt say I ever worried about the people who did having an unfair advantage.
Thatās the main concern, once again pointing to the incident on the Choice of Games forum described above.
Itās less than ideal but not the main problem as far as Iām concerned.
Well, sure, the rule is transparent and unambiguous but the actual implementation of the rule is far from it.
I would contend that giving people the benefit of the doubt is generally preferable in most situations.
I would prefer to get it right even if it takes extra effort.
Furthermore, I, at least, intend to stand up for any authors who appear to have been unjustly punished, so there will be a āpick apartā of the evidence, either here if the forum allows it or elsewhere if need be. (Of course, hoping that no such case occurs).
For a point of comparison, weāve been running this forum with a āno LLMsā rule for a few years now, and havenāt run into any significant problems with it. There havenāt been any witch hunts against people with limited English skills, nor has the public demanded evidence or called us feudal lords when someone got banned.
The vast majority of people here want to participate in good faith, so just posting a āno LLMsā rule cuts the amount of LLM-generated content almost to zero. The people who want to troll/spam/otherwise break the rules will generally want to troll/spam/break the rules anyway, whether or not we forbid LLMs, and so removing them is a net positive.
For comparison, a sampling of the last few people weāve banned: āI hope this message finds you well. As a dedicated loan broker, I am working with an organization who has interest in investing in your country,ā āDo you have any opportunity in allowing content with a dofollow link leading to sports betting/casino sites from your website or blog?ā āNotification of 8 pending messages. Some messages are restrained from delivering to intfiction.org. Due to low bandwidth we notify you to take prompt actionsā āExquisite Luxury Watches - Starting at $250ā
I donāt remember banning an actual human user in at least a year nowāthe people who legitimately want to use this forum generally also want to follow the rules. I donāt think litigating these bans is a worthwhile use of any of our time.
Everyoneās heard the warnings. Like it or not, weāre just going to wing it.
And hey, maybe itāll be easier for IFComp and Spring Thing than it was for us. (I doubt it.)
Surely you canāt mean that literally nobody will mistake human-authored work for AI. That happens constantly on larger forums/socials where strangers interact regularly.
Maybe you just mean that the IFComp committee will infallibly separate the wheat from the slop?
Speaking for myself, Iād pay good money for an infallible slop detector. But I wouldnāt trust anyone who claimed to have one, and neither should you.
Slop fighting is much harder than spam fighting, because spammers are trying to sell watches, to score backlinks, to scam people.
When people submit slop games to competitions, they have nearly the same goal as human-written authors: to entertain the player and win the competition.
Yeah, I like this update to the ruling. Iām interested in what this yearās Spring Thing and IFComp have to offer now and see what we can learn from organizing our own smaller competitions and communities.
Instead of speculating, why donāt we see how the new rule goes this year, give the organizers some grace for dealing with such a controversial and polarizing topic/solution, and, at the end of the comp, fill out the feedback form?
I like the rule change. Maybe some AI-generated material gets entered but if it is obvious it can be removed or get bad votes; otherwise, well it is better than without a rule.
Personally I would like the rule extended so that it is requires that a game can be played offline. I love that I can play a 30 year old game that donāt balk because an internet service no longer exists. But thats me.
Iāve worked in customer service for years. One of the strategies is to have a āmagic ruleā in posted policy. That basically is āwe can refuse service at any time for any reason.ā That doesnāt mean that it needs to be enforced draconically and doesnāt apply to every case, but provides a rule that can be pointed at for an edge case that doesnāt fit into the other rules.
I believe for IFComp this just means āwe can say noā to a complicated edge case if they believe it violates the rules. It also prevents people from wasting time building an IFComp release using AI since the rule is clear.
Iām also a big proponent of archivability, but there are a lot of platforms out there that donāt support it, and IFComp has historically wanted to include those platforms. Itās unfortunate, but cāest la vie.
