AI Content Policy - what do you think?

Travel Forums System Talk AI Content Policy - what do you think?

Page
  • 1
  • 2
Last Post
1. Posted by Peter (Admin 7250 posts) 6w Star this if you like it!

Hey everyone,

With the never ending deluge of AI content coming our way (and also in response to this thread), it's probably time to have a clear policy on the site regarding what level of AI we will allow, if any at all.

99% of the time, the AI content coming in is just junk spam that is added to build a profile and eventually (or immediately) drop a link to whatever website it is they are promoting. That content is typically falling foul of our current rules already.

I do see it as a core value of Travellerspoint that we are here to build connections between humans and AI content really does kind of fly in the face of that.

On the flip side, a lot of writing software now also has AI built facilities built into it. To the point where someone might be drafting a blog entry and then AI will be suggesting improvements along the way. That's a far more benign use-case and may even improve people's writing / spelling / etc.

Bearing all this in mind, there are some principles I'd like you to consider and see what you think of them. Please give each of them a rating from 1 to 5.

1 = Strongly disagree
2 = Moderately disagree
3 = Ambivalent
4 = Moderately agree
5 = Strongly agree

Here are the statements:

1. Our forums should be a 100% AI free zone
2. Our blogs should be a 100% AI free zone
3. We can accept AI content in our forums if it is not promotional in any way.
4. We can accept AI content in our blogs if it is not promotional in any way.

Only human answers to these questions please ;) If there's any counter points you want to throw in please do.

2. Posted by AndyF (Moderator 2968 posts) 6w Star this if you like it!

For the forums, I think machine answers are readily available to people's questions elsewhere if they want them; what we are offering is human answers. We're asking our members to give their time, and in return I think there is something of a duty that we try not to waste it.

Where I notice AI, it is generally a lot of words but saying little, and sometimes nonsense. Sometimes quite subtle nonsense such as poor advice.

I think this potential waste of our time is a problem, and something that worsens the forum for regular users.

That addresses the times I notice it, what I don't know is whether it is also being used more successfully in a smooth manner which improves someone's language.

The tool I use to check text is zerogpt. I don't know how accurate it is? It often confirms that the nonsense is AI generated; would it also identify more benign text where an AI tool has improved language? Does this risk outlawing something useful?

I think I need to listen to the discussion before I can give 1-5 answers.

[ Edit: Edited on 1 Jun 2024, 03:54 GMT by AndyF ]

3. Posted by Sander (Moderator 6048 posts) 6w Star this if you like it!

1. 5
2. 4
3. 1
4. 2

I have no doubt there are beneficial uses to "AI", but analogous to our zero-tolerance policy for promotional posts, I think the benefits of a blanket "no AI"-policy are huge. (And yes, that means I'd apply it to the blogs as well, even if I don't feel that strongly about it when looking at the blogs in isolation.) Moderation will remain human, so if we have any doubts, we can always contact posters and make exceptions where warranted.

Additionally, I suspect that in a year or two, it'll become a huge selling point for a site to be AI-free, particularly where it's about giving and getting advice.

There is no 100% accurate way of recognizing AI-generated content, since the AI can be prompted to create output in a different style. However, the default output is very recognizable. If I have any doubt, I tend to use zerogpt (which tends to be on the lenient side, classifying things as human when there's uncertainty - so perfect for our purposes), in combination with hivemoderation (the most accurate from all the tools I've tried, but only for long form content (and images)).

[ Edit: Edited on 1 Jun 2024, 09:55 GMT by Sander ]

4. Posted by leics2 (Travel Guru 6442 posts) 6w Star this if you like it!

5
3
1
3

Am away so can’t write longer comment, sorry.

[ Edit: Edited on 1 Jun 2024, 10:33 GMT by leics2 ]

5. Posted by Peter (Admin 7250 posts) 6w Star this if you like it!

Quoting AndyF

would it also identify more benign text where an AI tool has improved language? Does this risk outlawing something useful?

I guess that is my main concern as well - that we unintentionally flag something that is semi-justifiable.

Quoting AndyF

I think machine answers are readily available to people's questions elsewhere if they want them; what we are offering is human answers

Quoting Sander

I suspect that in a year or two, it'll become a huge selling point for a site to be AI-free, particularly where it's about giving and getting advice.

I agree with both of these statements.

6. Posted by AndyF (Moderator 2968 posts) 6w Star this if you like it!

1. 5
2. 3
3. 1
4. 3

Blogs are more of someone's own space, while in the forums there's the factor of protecting our volunteers' time in a community endeavour. I can't see an angle of public good to maintaining blog quality - though perhaps there is one from the site owner's perspective.

7. Posted by brabo1978 (Budding Member 150 posts) 5w Star this if you like it!

I agree that the forum should remain completely AI-free, as long as there's some kind of guarantee against false positives.
In the wiki-part (the guide) I don't have as much of a problem with AI, especially when it's only used in a supportive role.
On the blogs I don't have an opinion; I haven't checked them so far.

8. Posted by Psamathe (Budding Member 392 posts) 5w Star this if you like it!

Qu 1. Ans=5
Qu 2. Ans=5
Qu 3. Ans=1
Qu 4. Ans=1

AI content is pointless and at best adds nothing but normally just mis-information.

Re: False positives: If a post is deleted because of it being classed as from AI then message/e-mail the poster and if a sensible human reply then restore post otherwise block the user. No genuine posts are so urgent that them not appearing for a delay is going to cause crises.

Ian

9. Posted by Peter (Admin 7250 posts) 5w 2 Star this if you like it!

After pondering all the answers, I think it's a pretty clear win for banning AI fully from the forums.

When you are reporting posts, you now have the option of saying it's because the post was AI. When I'm reviewing the entries reported, it will have a link to to ZeroGPT for those posts so I can confirm. If it's a case where I feel there may be a genuine person behind it, I'll message them as you suggested Ian.

For the blogs, I think we play it by ear for now. Any promotional content is still disallowed which covers most of the AI junk in there.

I'll update the forum rules a bit later as well to ensure this is explicitly mentioned.

Thanks for all your feedback :)

10. Posted by leics2 (Travel Guru 6442 posts) 5w 1 Star this if you like it!

Thanks for that, Peter...and for adding 'AI' and 'copied content' to the report form. I've been using 'duplicate entry to report (copied content more than AI at the mo) and I'm sure the relevant moderator/s will appreciate knowing why I'm reporting! :-)