AppleVis Website September 2025 Updates

I'm not a fan of AI moderating things. I guess time will tell how it does. I do like the ability to delete comments.

Thank you. :) that's all I'm gonna say. Great stuff there.

i am also not a fan of AI moderating the sight cause i know AI can make mistakes some times

I appreciate the ability to delete comments and view forum names on the homepage.

I have the same reservations about AI moderation as other commenters, but I feel that the incorrect removal of posts is less likely to be an issue on AppleVis than some other forums because of the nature of the topics we discuss and the community AppleVis attracts. If time zones and moderators' availability are the impetus for AI moderation, I wonder whether AppleVis could consider diversifying its moderation team to have greater representation outside North America. This may have advantages beyond removing undesirable content from the site.

I have a few points of feedback concerning the main menu.

• I don't particularly like needing to expand the menu and relevant submenus. I feel that this adds unnecessary time when navigating to the item of interest.
• I feel that the "Post" submenu should be completely hidden if a user is not logged in.
• I think that it would make more sense for the "Account" submenu to appear as the first item in the list because being signed is a precondition to being able to post content, which is the reason why I most often access the menu.
• After expanding a submenu, it appears to take about a second for the resulting options to appear. When using JAWS with Brave and Firefox, pressing Down Arrow immediately after expanding a submenu (i.e. before the resulting items appear) causes focus to be placed at the end of the submenu, not at the beginning as users would expect. Can AppleVis do anything to make the items in submenus load more quickly, or can AppleVis do anything to ensure that screen reader focus always be placed on the first item of a submenu after a submenu has been expanded?

I think, as long as false positives can be flagged quickly, maybe by including a button in the email one receives rather than the contact form, to push it up to human moderation, this is a good idea. It's better to accidentally mute harmless content for a short time than let offensive stuff ride through, as I say, it needs to be easy for the poster to put it right without too many hoops to jump through.

Great work on the other stuff.

Next update, no subject lines for replies... Cheers. ;)

I am digging the updates to the website. I do agree with the poster who suggested that the account option be first, rather than the last, but otherwise I like the changes.
Sidenote, don't listen to anything Oliver says, that guy is whack!
True story. 😱

Great job as always, and kudos to the team for continuing to make this wonderful resource known as AppleVis available to the masses. The moderation documentation is important, and it will be interesting to see how it works out over time. Thanks once again and keep up the awesome work.

I'm old enough to remember when whack meant good... Which is how I'm taking this. Yes, I'm very, very whack. Chronically so.

Appreciate the Apple fitness team working so hard to make sure that we have the best experience possible. This is a great website and I hope to stay on here for a long time.

I will certainly offer feedback after trying your changes if I have anything worthwhile to offer. smile

I appreciate everyone's concerns regarding our testing of AI tools to assist with content moderation. I want to offer some clarification about what we are trying to do (and not do).

First, I want to take a moment to address the elephant in the room: AI content moderation and concerns about potential censorship on AppleVis. While AppleVis does not take positions on issues happening outside of our mission and scope, we are also very aware that due to geopolitical factors, censorship is a concern for many in our community.

I want to expressly state that our only purpose in any moderation activities, be it manual review of content or scanning all new posts with an AI tool, is to help ensure that AppleVis is a safe, welcoming, and informative place for people of all ages, cultures, abilities, and lived experiences. While we do place some limits on the content that users may post, these limits are only to the extent necessary to achieve the aforementioned goal.

We are also keenly aware of how AI content moderation tools can "not" work. In September 2024, a post AppleVis shared to a major social media platform about iOS 18 bugs was removed, because the platform's 'technology' determined that we violated their guidelines on trying to obtain followers. How the 'technology' came to that conclusion is beyond me, but I share this anecdote to try and illustrate that we are aware of the pitfalls of AI content moderation and have every intention to avoid them. And if the disadvantages of using AI moderation tools outweigh the benefits, we will discontinue their use. Seriously.

With that in mind, here is what we are trying to accomplish:

It is our hope to try and see if/how AI can identify blatantly harmful content and not allow it to be posted to the site, and let users know in real-time. I liken it to having a smoke detector in your house or going through a metal detector at an airport. We do not want AI flagging posts that require human insight to understand their full meaning and context. We do, however, want to see if there is a way in which We can use AI to identify really obvious and clear violations of our guidelines.

For example, if someone uses the f-bomb or a racial slur in a post (things which are both definite violations of our guidelines), we want the AI tool to identify that and not allow that to be published to the website. If AI can be used to identify and prevent that sort of content from being published, again in those very clear-cut situations, I believe we will be better for it.

Our hope is that the eventual solution will be one where we can input our forum guidelines and prompt the model to evaluate newly-submitted content only against those guidelines. If a post is flagged (and our expectation is that not many posts will be), it will not go through, and the user will be alerted in real-time. As we explore different solutions, one thing we are looking for is something that will share with the author why specifically the post did not go through and present that information in a neutral, nonconfrontational way.

If after evaluation we determine that AI moderation tools are not a good fit for our needs (which is a distinct possibility), we won't use them. We have no interest in regulating speech on AppleVis besides what is necessary to make our community safe, welcoming, and helpful for everyone. If for example the model consistently flags content that is not a violation of our guidelines, this is not something we will compromise on. We will either get the technology to work for our needs, or we won't use it at all.

The one thing I do ask is that if something you post is wrongly flagged, please try to be understanding and allow us to solve the issue constructively with you. We appreciate that discussions about limits on what can and cannot be posted are sensitive topics for many.

Thanks,
Michael