Jupiter Rowland

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

Duothematic channel. Primary topic is virtual worlds/OpenSim, secondary topic is the Fediverse beyond Mastodon. Follow me from Mastodon or anywhere else in the Fediverse.

Ort:: The Hypergrid

Geschlecht:: Männlich

Homepage:: https://hub.netzgemeinde.eu/profile/jupiter_rowland

Sun, 26 Nov 2023 14:41:37 +0100

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

Got nice pictures, but inclusivity makes it too tedious for me to post them and for you to enjoy them; CW: long post (795 characters)

Artikel ansehen

Fediverse OpenSim

show all 8 comments

Sun, 26 Nov 2023 20:16:55 +0100

cybrkyd

cybrkyd@s.pfm.cx

Mon, 27 Nov 2023 00:48:54 +0100

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

@cybrkyd If this actually worked, the obvious downside of it would be that I'd have to tag pictures #NSFW that aren't actually NSFW.

The other downside is that even a test picture requires a full-blown image description with all bells and whistles. Last time I've described images, a night's rest was between starting and completing the descriptions for two images.

Mon, 27 Nov 2023 04:24:18 +0100

Scott M. Stolz

scott@biztechtonics.net

I am thinking that people who need a specialized view either need to use specialized software or need an addon that helps them pre-process incoming content.

For example, AI could read incoming images and hide the ones with eye content for those who are sensitive to that. No labelling by the content creator required.

And for those who have trouble seeing, it could present the content in a way that is easier for them to digest.

Mastodon is designed for people who can see, so I am not concerned about how Mastodon renders things. It is not designed for blind people.

But if a purpose-built app for blind people could not understand my posts, then I would be concerned and would see if I could find a way to rectify that.

Mon, 27 Nov 2023 04:25:49 +0100

Scott M. Stolz

scott@biztechtonics.net

Do people who are blind or have trouble seeing actually want a 10,000 character description? I'm pretty sure that would fall into the TLDR category in most contexts.

Mon, 27 Nov 2023 11:48:41 +0100

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

Just proverbially crossing my arms and saying, "We've never done anything for accessibility on Hubzilla, that's not the Hubzilla way, that's not what Hubzilla's culture is, that's adopting Mastodon mannerisms, I'm gonna do nothing and wait for technology to do everything on the readers' side," can't be a solution. Unless you have Pubcrawl off. Probably not even then because someone with Pubcrawl on and Mastodon connections might share one of your posts.

In case you haven't noticed, AI for image descriptions is a hot debate on Mastodon. Blind or visually-impaired users who do post images don't have any other choice but to use AI-generated image descriptions. They sometimes also try to resort to AI to describe images in other people's posts with no alt-text. And many sighted users are waiting for the convenience of a one-tap AI image description so they don't have to write one themselves.

Don't forget that while almost everyone on Hubzilla is on Linux PCs/laptops, almost everyone on Mastodon is on a phone with no hardware keyboard. Typing on a screen is tedious, and it's hard to describe an image while it's obscured by your touch keyboard that takes up 2/3 of the screen or more.

But AI is not reliable. At the very least, it can become vague and use a lot of "might be", "could be", "seems to be", all of which is a huge no-no in image descriptions. The more complex an image is, and the more niche and obscure its contents are, the bigger is the chance that AI will simply blurt out complete non-sense. Humans will remain better at describing images than machines.

This goes doubly so for my images which are extremely niche. This means two things: One, people don't know anything about what these images show, so they need extensive descriptions and explanations. Two, AI can't deliver these because AI is completely clueless about my pictures.

AI-issued, reader-side content warnings don't even exist yet. And if they did, they'd be just as unreliable. Such an AI wouldn't be nearly as sensitive as a high-functioning autistic person who is easily triggered by eye contact, also because for autistic people, "eye contact" goes much further than a neurologically typical person might expect. It isn't only for facial portraits of people who look directly into the camera.

It's for whenever a face is in a picture. Or part of a face. No matter where the face is oriented. No matter if only part of the face is visible. No matter if it's an actual human being or not. Even animals or aliens or robots can cause an eye contact trigger. If there's one eye in the picture, it can be an eye contact trigger. Some animals technically always require an eye contact content warning because it's next to impossible to show them without showing at least one eye (rodents, horses, most birds...).

Also, it doesn't matter how big the face is. I know from an actually autistic Fediverse user, @♾️ Yuki (스노 雪亮) ? ?, who is in contact with other autistic people, that a face that's smaller than one pixel can cause such a trigger. A highly-functional autistic person can deduce fairly easily from the context that those three or four stacked pixels in an otherwise very busy image are a person who is not completely oriented away from the camera. And there's the trigger. I wish I was making this up, but I'm not.

To answer your last question: What they want is not unlikely to be a detailed and sufficiently informative image description. However, in my case, "detailed and sufficiently informative" inevitably means humongous.

Don't forget that I don't post real-life cat photographs. I don't post real-life photographs at all unless they're parts of memes templates. My own pictures are all from virtual worlds.

Sure, many people don't care, especially since "The Metaverse" is widely considered a dead horse. And you may say that virtual worlds are a very visual topic, so blind or visually-impaired people wouldn't care in general, so describing virtual world images is pearls before swine because they've got no target audience whatsoever.

But some people do care. And no, it can't be fully ruled out that these include blind or visually-impaired people. Some people do become curious when they come across a picture from a virtual world.

"Curious" means these people want to know everything. But right now, they know nothing.

They want to know where the picture was made. Especially, but not limited to which system. "Is that Second Life?!" Or more likely: "Is that The Metaverse?" (where "The Metaverse" refers to Zuck's invention because nobody knows that Zuck's invention is named "Horizons" and not "The Metaverse") Trouble: It's neither. It's something that even experts don't know, experts who know everything from Horizons to Decentraland to Rec Room to VRchat to The Sandbox etc. And laypeople know it even less.

So I inevitably have to take a deeeeeeeeeeeeeeep breath and start explaining. Which system ("OpenSimulator"). What that system is ("OpenSimulator, OpenSim in short, is a free, open-source re-implementation of the technology of Second Life yada yada..."). Which grid that runs on that system ("Dorenas World"). What a grid is. Which sim on that grid ("Nihilon"). What a sim is. Which specific location on that sim ("The Rockhouse").

That's over 1,000 characters. And that's only covering the location. You can easily name-drop "Empire State Building" or "Chrysler Building" or "White House" or "Eiffel Tower" or "Taj Mahal" or "Sydney Opera House" with no explanation, and just about everyone knows what you mean. Try name-dropping "Sendalonde Community Library" or "OSCC stage" with no explanation and see how many people know right off the bat what you mean. But if they're curious, they want to know.

Then comes the actual description. Most real-life photographs posted in the Fediverse focus on something, and they get away with only describing that something because that something is all that matters within the context of the image. People don't need the rest described. Those with longer image descriptions usually display a whole scene without focusing on anything specific.

It's different with virtual-world pictures when people get curious. For when they get curious, again, they want to know it all while currently knowing nothing. Even if one of my pictures focuses on something specific, they still want to suck up all information they can get about the surroundings like a sponge. Same if it doesn't.

Even sighted people will go, "What's this? What's that?" This goes much more for Second Life and worlds based on OpenSim than for Horizons because Horizons, geared towards VR headsets, is very cartoonish, simplistic and lacking details and decorations whereas Second Life and OpenSim, both geared towards 2-D screens and desktop GPUs, can get more bustling with details than the best MMORPGs out there.

And in stark contrast to what sighted people may expect, blind or visually-impaired people won't be like, "Can't see jack shit, don't care." They will want an extensive description of what is in this "new" and unknown world because they can't see it.

When describing an image with such obscure content to blind or visually-impaired people, you can't get away with name-dropping what's there either. You can't get away with saying there's stuff. Blind and visually-impaired people will want to know what that stuff looks like. If they're really curious about a picture, and this is the case I'm writing about here, they won't care whether or not that stuff matters within the context of the picture. If you don't tell them what that stuff looks like, you're ableist.

All this goes even more for stuff that doesn't exist in real life. OSW beacons. Teleporters. Dance machines. In fact, buildings because few virtual buildings are exact replicas of real-life buildings, and even then, blind people don't know what the real-life originals look like either. And, of course, everything specifically purpose-made such as billboards, sales boxes and such stuff. Even sighted people wouldn't know what that stuff looks like if you didn't show them a picture.

If it doesn't exist in real life, neither blind nor visually-impaired nor sighted people know what it is in the first place nor what it does. But, again, they might be very curious about it. A teleporter, well, it probably teleports you somewhere, but where and how and... A dance machine, well, it sounds like it has something to do with dancing, but what? And the term "OSW beacon" is absolutely no indication whatsoever for what that is and what it does.

So I can't get away with name-dropping. I can't get away with name-dropping and describing the look either. I have to name-drop, then explain what it is, then describe what it looks like, then describe and explain what it does and how it works.

I have to do that with everything in the picture because people will be like, "What's this, what's that?" And blind and visually-impaired people will be like, "What does this look like, what does that look like?"

Avatars add another layer of complexity in the description. It might be necessary for the general understanding of such avatars to mention that they aren't monolithic like in Vircadia, but modular and very extensively customisable and configurable. Mind you, on top of the general description of who that is, what that avatar looks like and what that avatar does. But then again, this doesn't matter much because I avoid showing avatars since each avatar is an eye contact trigger risk.

And then comes text. If there's text within the borders of a picture, so they say, it has to be transcribed fully and verbatim. I always do that. But: My pictures are likely to have many bits of text strewn about. Again, if there's text, it has to be transcribed.

Yes, those bits of text are often so small that they aren't legible in the image itself. But for one, there is no rule that says that illegible text must not be transcribed. In fact, most "How to write alt-text" guides don't deal with illegible text at all because whoever wrote them is unaware that it could possibly occur.

Besides, just because it's illegible in the picture, doesn't mean I can't read and transcribe it. Because I almost always can. Not from the picture. But I don't only have the picture. I can log into OpenSim, go to the place where I've taken the picture, go to where that bit of text is, zoom in on that bit of text, and bam, I get that text at a hundred times the resolution it has in the picture! And it's perfectly legible. In fact, I usually write my image descriptions immediately after taking a picture without leaving that place and then going back.

Just saying that it's illegible is lazy. Unnecessarily so. And it's unfair towards curious people who are likely to want to know what's written there.

When all this comes together, over 1,000 characters for the location, everything having to be described because nobody is familiar with anything, a lot having to be explained, lots of text transcripts etc., this results in truly massive image descriptions. 10,000 characters are actually short for my standards. My longest image description is 40,000 characters, not counting the preamble that covers both images in the post.

By the way, I'm working on an article that explains this in-depth.

Mon, 27 Nov 2023 13:19:34 +0100 zuletzt bearbeitet: Mon, 27 Nov 2023 13:22:12 +0100

Scott M. Stolz

scott@biztechtonics.net

When I mentioned AI, I did not mean that it should write descriptions. It is pretty good at detecting objects in pictures though and can hide images that might contain certain items. It knows what a face is, what an eye is, what nudity is, what meat is, etc. And an AI can be trained to detect specific things.

It won't be perfect, but for someone triggered by eye contract, it could look for faces & eyes, and hide pictures with eyes in them. It might not detect an eye 1 pixel in size but most humans wouldn't even notice that. And if someone is so sensitive that they are triggered by a 1 pixel eye on a 4 pixel face, then ALL images should be shut off for them, because any image with a living being in it could have an eye in it. There might be an animal hiding in the bushes of that picture that you thought had no animals in it.

You still need humans to write meaningful descriptions, but an AI can certainly hide images if it detects an eye in the image.

As far as long descriptions go, it depends on the context and your audience. Since you like writing long detailed descriptions and posts, people who like that are more likely to follow you. And people, including blind people, who want short and sweet will just unfollow you or scroll past, since you write a book describing your photos.

It's not right or wrong. People have different preferences. So, ultimately it comes down to what you want to create, and what your audience wants.

And as far as what platform people choose, it's like a blind person picking up a printed newspaper. Not the best option. They should get a braille version of the newspaper, an online version that could be read to them by screen-reading software, or have one of those handheld scanners that can read the text to them. Because if they can't see, that printed newspaper by itself is not going to be readable by them without assistance.

While we should not ignore than people will use less-than-ideal platforms, we should still work on developing technologies that help people with disabilities.

Mon, 27 Nov 2023 16:44:00 +0100

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

@Scott M. Stolz

It might not detect an eye 1 pixel in size but most humans wouldn't even notice that. And if someone is so sensitive that they are triggered by a 1 pixel eye on a 4 pixel face, then ALL images should be shut off for them, because any image with a living being in it could have an eye in it.

As I've said, I've got written confirmation that some autistic people are triggered by eye contact from a vertical stack of 4 pixels that shows a person. As in, the whole face is less than a quarter of a pixel. It can still trigger them.

There might be an animal hiding in the bushes of that picture that you thought had no animals in it.

As long as that animal is sufficiently covered up that it's invisible, it doesn't require a content warning.

As far as long descriptions go, it depends on the context and your audience. Since you like writing long detailed descriptions and posts, people who like that are more likely to follow you. And people, including blind people, who want short and sweet will just unfollow you or scroll past, since you write a book describing your photos.

It's not right or wrong. People have different preferences. So, ultimately it comes down to what you want to create, and what your audience wants.

My audience is kind of blurry. It actually is everyone out there who comes across one of my posts, be it as a follower, be it as someone who finds one of my posts in their federated timeline. After all, I think I got most of my new followers over the last half year through federated timelines.

Mon, 27 Nov 2023 19:36:45 +0100 zuletzt bearbeitet: Mon, 27 Nov 2023 20:13:09 +0100

Scott M. Stolz

scott@biztechtonics.net

@Jupiter Rowland

As long as that animal is sufficiently covered up that it's invisible, it doesn't require a content warning.

My point is that an animal in the bushes may not be noticed by the person sharing the picture, and even if they wanted to issue a content warning, they wouldn't because they didn't even notice the animal staring at the camera between the leaves in the photo.

@Jupiter Rowland

My audience is kind of blurry. It actually is everyone out there who comes across one of my posts, be it as a follower, be it as someone who finds one of my posts in their federated timeline. After all, I think I got most of my new followers over the last half year through federated timelines.

You can't please everyone. I know someone who gets triggered when you include a trigger warning. I suspect they are autistic, but they would never admit it. They sure do cuss a lot and their blood pressure raises when they see a trigger warning. I don't understand why they get so upset at trigger warnings, but they do. I just told them not to use Mastodon and not follow people who used trigger warnings. It helped, but sometimes a trigger warning slips through.

But instead of indulging them, we are helping to rehabilitate them so they don't associate negative feelings and thoughts to trigger warnings. Because that is what is happening in the brain: an association between something and an emotion, and the emotion is triggered by whatever item they have the association with.

In one person, eye contact is associated with fear. In another person, trigger warnings are associated with anger. In another person, a particular song is associated with suicide and death. They get depressed when they hear it.

It's the same process. And the more you repeat this pattern, the stronger it gets, until you get to a point where it is almost impossible to break. Breaking these patterns is not an overnight process, but unless you want the person to dependent on others for the rest of their lives, it's better for them to break these unhealthy thought patterns, preferably under the supervision of a professional. And for an autistic person, the process seems to be harder, since they are faster to latch onto thought patterns and are more resistant to change. But in a way, that is the same with some very stubborn older people. Their brain might not be as flexible as others, making it harder for them to cope with real life. So a professional who is experienced with autism or stubborn older patients or other similar brain issues would be ideal.

So, we do want to be accommodating and don't want to upset people, but at the same time, we should not encourage a path that is unhealthy for them. It's a hard situation to deal with personally, but it is even harder when trying to deal with people you don't know who all react differently.