RE: Torch letter - Given the occasion... / Brandbrief - Aus gegebenem Anlaß…
Thank you for your post on this subject.
AI presents considerable difficulties for a platform like Steem.
The AI detection tools are very far from perfect. If Moderators do use them then probabilities of at least 90% should be required for any AI generated conclusion should be reached.
The use of translation tools and grammar correction tools (like Grammarly) further interfere with AI detection.
'Natural detection' is more likely to be reliable - but this only really works for people checking texts written in their own native language, including comparison with an author's previous texts. Subject matter and context are also very important.
We hope Community Admins will give guidance to their Moderators.
As a first layer of arbitration in disputed cases we would suggest that Steem Representatives are called upon to give an opinion.
If there are particular communities where such 'AI Accusations' are occurring frequently it will be useful if we are tagged so that we might take note for future support offerings.
As a footnote we are keen that more people from Europe, North America and Australia apply to be Community Curators as those areas of the world are currently under-represented in the curation teams.
A thousand thanks for these clear words! Of course we can mention you if such conflicts accumulate in one place or around one person.
I also heard the subtle hint that applications for Community Curators from Europe are welcome ;-)) I am still grateful for the opportunity I got with Team Freestyle last year! Under the very liberal conditions of this team, I was able to get on well with the task and the responsibility. With the current profile - unfortunately not. I also fear that the amount of work I very much enjoy doing for Dream Steem may prevent me from applying again.
I had imagined in the spring that I would be a good fit as a Steem Representative, who after all does not have to be a team worker - and applied without meeting the minimum requirement for my own SP. Since this threshold has been raised again in the current round, I don't see any possibility of contributing in this way.
However, I am always good for cautionary words and a few unconventional thoughts. I promise ;-))
1500 STEEM in 13 days...
That's a challenge ;-))
I agree with your suggestion that we should have more Community Curators from Europe, North America, and Australia. I am an Indian and in India right now but settled in Australia for a long time. I will be back in Australia by the next month and have my plan to apply for this responsibility in the coming month. Thank you!
First off, I agree that current AI detection tools are far from perfect. Setting a high threshold like 90% probability seems wise before accusing someone of using AI. False accusations could really damage the community and people's reputations.
Your suggestion about "natural detection" being more reliable is well taken. As humans, we tend to have a good intuitive sense when content doesn't "sound" quite right or natural. Comparing writing styles (especially the comments) over time for a given author seems like a smart way to identify odd deviations that may indicate AI use.
However, as you pointed out, that too depends on the subject matter and context. For instance, the writing style of a fictional story is totally different from a casual post. Moreover, moderators checking the post should also be aware of “language diction” to avoid any misconception on their part.
Additionally, a transparent appeals process for accused creators could help fix any mistaken AI allegations. Handling disputes calmly and giving benefit of the doubt initially seems prudent. It’s possible that if such a system existed, I could’ve avoided the unpleasant situation I had to go through. Although I tried resolving my case on Discord first, it didn’t work out...
Overall, I appreciate you raising these concerns and suggestions in a constructive way. You clearly want to see this platform thrive without damaging content creators' reputations unfairly. 🙂
Congratulations, your comment has been successfully curated by @irawandedy at 10%
AI can actually be used positively as a reference for writers to develop their ideas in writing, but using it brutally actually endangers the continuity of their creativity in writing.
As an active user, since the beginning I have actively tested various types of AI detectors, but this cannot always be used as a final reference, even in personal experience I have experienced surprises when using English grammar repair services, the results of the 6 detectors I tested show that 99% of the writing is a product of AI.
I was surprised and had discussions with many platform users, both in Indonesia and even other users who understand English grammar well. It can be concluded that the detector will detect text generated by AI if the grammar writing for each sentence is very good, based on our discussion, it would be better for the writers of this platform to use their own language, without using a translator tool or Grammarly service so that unwanted things can be avoided.
As a verifier in the community, I continue to build interaction and polls regarding this in groups and individual discord. It is not easy to convict that an article detected by AI is an AI product. I did the steps that took a lot of time and effort by:
Engagement Challenges are also prone to be found by AI Products, but the percentage is very small but it takes the admin and moderator's hard work as verifiers to be more thorough in checking each article to be verified.
In certain cases, diary games can also be written using AI, the author outsmarts this by setting points of activity from morning to evening. This is a very difficult job to do, because it involves the user's daily activities, but back to my previous discussion that background and his writing style is an important indicator to suspect him and focus on searching and proving.
Weaknesses for the global community when they do not represent moderator representatives who master a particular language, of course it will take quite a lot of time for just 1 post. Should the representation of a moderator who understands a variety of languages be considered in combating this problem.
I will add other special problems that are crucial, for example in writing a game diary. The game diary is principally a diary for the writer within 1 day, meaning that in addition to the activities, the author also includes photos of the activities as a complement to the story. In fact, not all diaries use pictures on the same day, even pictures on several days are made into 1 diary.
In my opinion, this is ethical, and everyone should pay attention.
You have provided very thought-provoking material that should serve as a reference for everyone responsible for this AI checking process. What I feel is that at the end of the day, the writing style of an author is the only way to prove whether he used AI, Grammarly, or wrote it himself without using any tool. How would you take broken language suddenly turning into something that sends seasoned writers into shock?
You have raised a very important point. Although it is difficult, it is still considerable. 👍
Greetings Mr..
AI content detection tools do their respective jobs, even though they are not perfect, in fact, the most popular anti-plagiarism tools also have their flaws.
What you should do in such cases is to investigate the suspicious content both with AI detection tools and also using the anti-plagiarism tools.
In my experience, AI detection tools always give me clues to fractions of texts that have been taken from other sources.
I am at your disposal for any suggestions.