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

Wed, 16 Oct 2024 00:31:46 +0200

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

You can't make all your content perfectly accessible to everyone. What makes something more accessible to one person may make it less accessible to another.

But just about everyone with a disability expects you to make all your content perfectly accessible to them individually. If you don't, you're ableist.

And so you have to make all your content perfectly accessible to everyone. If you don't, you're ableist.

#Disability #Inclusion #Inclusivity #A11y #Accessibility

Accessibility

show all 19 comments

Wed, 16 Oct 2024 00:39:50 +0200

Jeffrey D. Stark

JStark@mastodon.social

@jupiter_rowland there's a middle ground to be had

Wed, 16 Oct 2024 00:41:01 +0200

Marilyn Jean

alohapilgrim@hub.hubzilla.de

Flawed generation requires flawed logic. Been there but wisened up.

Wed, 16 Oct 2024 04:05:12 +0200

Bill Statler

billstatler@forum.statler.ws

I think the majority of people are in that "middle ground", but the extremes grab most of the attention -- like with so much of human society these days.

I saw one person who specializes in artistic black-and-white photos -- he was told that black-and-white photos are depressing and need a Content Warning.

Another person makes excellent graphs that present huge amounts of data in an easily understood visual form. He was told to add alt-text. What, he should put hundreds of rows of spreadsheet data into a 1500-character alt text?

And the natural reaction of us folks in the middle ground is to get angry, head for the other extreme, and refuse to put any effort into accessibility.

Wed, 16 Oct 2024 11:16:30 +0200

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

@Jeffrey D. Stark My main issue is with image descriptions because there are very different requirements that actually contradict each other.

It's usually said that images should be described in an alt-text that's clear, concise and, most of all, short. And only in that alt-text. And there are many blind or visually-impaired users who want alt-text that way.

Then there are those who are bored from purely informative image descriptions because they're so bland. They prefer AI descriptions over human descriptions because they'd gladly trade accuracy which they can't verify anyway for whimsical entertainment.

Well, and then there are those who love details, and who need everything described that they don't know what it looks like. Mastodon loves detailed image descriptions, much more than the general Web or other social media. But my original images tend to be chock-full of only stuff that next to nobody has ever seen before, and that these people don't know what it looks like. They're an edge-case that nobody thinks of when wishing for detailed, informative image descriptions.

I regularly spend hours or even days describing one image beyond what its resolution shows. Explaining in an image description that something isn't visible in detail because the resolution is too low or omitting it entirely for this reason feels too much like laziness and weaseling out. The result is a very long description in the alt-text already and an additional, outright massive description in the post.

Both are as informative as they can get, the short description within the limits of what Mastodon, Misskey and their forks don't cut off yet, the long description within the limits that I do impose on myself. But even the short one in the alt-text is too long for not few blind or visually-impaired users. And the long one in the post sometimes exceeds the length of an essay and grows as long as a short story.

I'm caught between describing too little for some users and making descriptions too long for others.

#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #A11y #Accessibility

Wed, 16 Oct 2024 11:40:38 +0200

Charlotte Joanne

Lottie@beige.party

@jupiter_rowland This is a fantastic description. Sometimes I really just do enjoy the whimsical entertainment! Let’s face it if there is genuine important information I will always have to verify it with a human anyway or at least it will have to for the foreseeable future!

Wed, 16 Oct 2024 13:09:31 +0200

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

@Charlotte Joanne That's part of what I mean.

I could sit down for two solid days, morning to evening. Describe an image twice over. 1,400 characters of short description in alt-text. Over 60,000 characters of long description in the post, including all necessary explanations and all transcripts of all text within the borders of the image because they wouldn't fit into the alt-text.

You may think this is completely outlandish, but I've actually done exactly this before. Just search for the hashtag UniversalCampus and scroll all the way down; that's the post I'm talking about. The original is here: https://hub.netzgemeinde.eu/item/f8ac991d-b64b-4290-be69-28feb51ba2a7.

For the third group, I guess it's what they need. Granted, when I mention that there is a preview image of another place on a teleporter, they may still be disappointed that I didn't give a detailed description of this image or even of the place itself, regardless of image resolution.

The first group, those who want short descriptions, will complain that it's too much, and that not providing actually short descriptions of no more than 200 characters in the alt-text is ableist because longer alt-text is too inconvenient.

The second group, those who want whimsical descriptions, may start having the alt-text read to them, stop after a few seconds because it's so bland, fire up their AI which hasn't the foggiest idea what it's even looking at and listen to a description that's half-nonsense, but nicely whimsical. They don't know that it's half-nonsense, and they don't care because what the image shows is not a matter of life and death that requires accuracy. Oh, and of course, they won't even bother with the long description in the post. Not because it's so long, but because it's probably just as boring.

On the other hand, if I catered to the first group, those who want image descriptions to be short, I'd write a very short alt-text of no more than 200 characters that doesn't actually describe anything.

Then I'd have to deal with blind or visually-impaired people inquiring, "Yeah, but what does everything look like?" The question is justified. How are they supposed to know what such a place in a 3-D virtual world looks like and what anything in this place looks like? So whatever elements are in this picture, these people would ask for detailed visual descriptions. Whenever I reveal to them what else is in the picture, they'd want detailed visual descriptions of it. And if they don't understand something, they'd ask for enough detailed explanations until they understand it.

In the end, they'd ask me why I didn't give all this information right away, why I'm so lazy and careless that they'd have to take matters into their own hands and squeeze this information out of me.

Of course, the second group, those who prefer whimsy over information, would ignore this description as well and substitute it with an AI description that's funnier.

If I catered only to this second group and simply used an AI description as it comes out of the AI, I'd end up telling nonsense and outright lying about my own images even though I should know better. But the second group doesn't care for accuracy.

The first group would dislike the description for having too much cruft in it and being too long. The third group would criticise it for not being informative enough, and I would have to explain to them that the description came from an AI, that it's factually half wrong, and I'd have to tell them what the image really shows. Lastly, Mastodon's alt-text police would scold me for the inaccurate rubbish that's my image description. Again, although I could have done better.

The last option would be to cater both to the second and the third group and make it accurate and informative and detailed and at the same time whimsical and entertaining. The workflow would start the same as my actual workflow.

But on top of the two days I need to research for and write the descriptions, I'd need another three days to make both descriptions funny and quirky. The long, detailed description would grow from 60,000 to 80,000 or 90,000 characters due to all the whimsical bits added to the same amount of information. The short description would still be limited in length, and so I'd have to sacrifice information for more whimsy.

In the end, I won't satisfy anyone. For the first group, it'd still be too long. For the second group, chances are that the alt-text is actually too long, because which AI churns out 1,400 characters of image description? If not, it probably isn't the right kind of whimsical, not the kind of whimsical that AIs produce. And all the whimsy couldn't rescue the 90,000-character long description from being too long. And the third group would say that I should have added less cruft and more information to the alt-text, and that all the uninformative cruft has blown the long description out of proportion.

I guess it's obvious that I can't satisfy everyone. That'd require a hyper-detailed description of a highly complex image, including transcripts of all bits of text, within no more than 200 characters which, at the same time, is also funny and quirky. In addition, I'd have to find a way to explain things which the image description reveals and identifies while still making these self-same explanations, including their context, available to people who can't access alt-text.

And as this is completely impossible, perfectly satisfying everyone all the same is just as impossible. In order for one person to be satisfied, another will have to suffer.

#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #A11y #Accessibility

Wed, 16 Oct 2024 14:18:09 +0200

Jeffrey D. Stark

JStark@mastodon.social

@jupiter_rowland well; i think you need to assume the majority rules; and recognise that there are ways to get the larger details and nuances but the majority of people don't want that level of detail

Wed, 16 Oct 2024 19:28:19 +0200

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

@Jeffrey D. Stark I don't know for certain what the majority want in general. And I don't know what they'd want in my very specific case.

I can only try and extrapolate what they might want from other Fediverse users' image descriptions and feedback on other Fediverse users' image descriptions, as little as there is.

The problem with this is that I don't post what everyone else posts. Not real-life photographs, not screenshots from social media etc., but renderings from very obscure 3-D virtual worlds. This means that there is next to nothing in my images that anyone is familiar with and that any blind or visually-impaired user has a rough idea what it looks like.

I've seen real-life photographs, sometimes literally focusing on one specific element in them with the whole background blurred out, that were described in over 800 characters. I've seen them be praised for their alt-texts. On the other hand, I've never seen a real-life photograph in the Fediverse be criticised for too much alt-text.

This, however, doesn't easily translate to virtual world renderings. Real-life photographs are much more familiar and much more likely to mostly contain things that people are more or less familiar with. And yet, people love it when they're described in 800 characters and more, all the way to replying with hashtags such as #AltTextAward or #AltTextAwards or #AltTextHallOfFame.

Logical conclusion: If there's more in the images that people aren't familiar with, I'll have to describe more than in these real-life photographs. And there is more in the images that people aren't familiar with.

Virtual world renderings are a largely unexplored edge-case. Only very few people in the Fediverse post these. I think only two describe them. And I'm the only one who really puts some thought into describing them instead of trying to get away with the bare minimum. This means that what I'm trying to do is a first. Nobody has done it before me. There's no prior experience with it.

Thus, I have to go with my own assumptions and conclusions based on a) observations on Mastodon and b) the differences in familiarity between real life and what I post about.

Three things are clear about my images.

First, if sighted people see it, they don't really know what it is, where it is etc.

Second, if non-sighted people come across the image, there is nothing in the image of which they know what it looks like due to having been told often enough what it looks like because they've never been told what anything in the image looks like. But they may want to know what it looks like. And it's their right to know what it looks like.

Third, this topic is such a small niche and so extremely obscure that if you don't know something, you can't just look it up on Wikipedia. You can't even Google it. Generally, the only source of information that could really help you with my pictures, that's me. I'm definitely the only way "to get the larger details and nuances".

And so there's much more in my images that needs to be described. And there's much more that needs to be explained, one of the reasons why I always describe my virtual world renderings twice.

This starts with the location, the place where an image was taken. There are cases in which it does matter where an image was taken. My virtual world renderings are such cases.

If a real-life location is shown in a photo, sighted people may recognise it because it's so famous. Otherwise and for non-sighted people, simple name-dropping is usually sufficient. There's hardly any place in real life that can't be sufficiently mentioned in under 50 characters.

I can't name-drop. It won't tell anyone anything because nobody would know the name I've dropped. If I want to tell people where an image is from, I'll first have to break it down and then explain it and explain the explanation and so forth. I can't tell anyone, sighted or not, where my images are from in under 1,000 characters. Not if I want them to understand it.

As for visual descriptions, the usual advice is to limit yourself to what's important within the context, describe only the one important element in detail and hint at everything else at most. But I don't always have that one important element. I may have about two dozen important elements. Or, more often, the post is about the whole image, the whole scenery, and everything in it is important just the same.

But even if something in the image is more important than something else, I still have to describe everything. I mean, we're talking about what amounts to glances into a whole new universe for 99.999% of all people out there. Sure, many will shrug it off.

Others, however, may be intrigued, curious even. After all, this is evidence that "the Metaverse" is, surprisingly, alive. It is not suggested in AI-generated pictures. It really exists. And it looks better than all of Zuckerberg's propaganda. These people don't care what matters in the image and what doesn't. They go on an exploration trip all across the whole image and take in all the details.

Blind or visually-impaired people can't do this. But they may want to do it. And they've got the right to do it, just like sighted people. So they should be given just the same opportunity to do it. Remember that I can't assume that they know what anything in the image looks like unless there's a real-life counterpart that looks very much the same.

Whenever there's something in one of my images that doesn't exist in real life, I have to assume that nobody knows anything about it. So not only do I need an extensive visual description, but I often also need an extensive explanation of this one item.

Finally, there's one more thing that takes up a lot of room: text transcripts. The rule is that if there is text within the borders of an image, it must be transcribed. I rarely even see the exception "unless it doesn't matter within the context". And, again, it tends to happen that everything in one of my images matters within the context because the context is the very image itself.

What this rule doesn't cover at all is text that is unreadable in the image as it is shown. There is no exception for this kind of text, nor is it explicitly included in this rule. It isn't handled at all. It has never even been thought of. Hence, I must assume that the rule applies to this kind of text just as well.

Before you say that I can't transcribe text that I can't read: I actually can. I don't transcribe text by looking at it in the image. I transcribe text by looking at it in-world. And all of a sudden, those six pixels in a row that are ever so slightly more greenish than the surrounding white are two lines of text. That blot, four pixels wide, three pixels high, is actually a sign with a 1024x768-pixel texture and text that's perfectly legible. That tree trunk in front of that sign? In-world, I can look behind it.

If I can transcribe all this text, and nothing says I must not do so, I assume I must do so. And so I may end up with several dozen transcripts of more or less text which, including their respective contexts in the image description, take up more characters than fit into a Mastodon alt-text. If this is the case, then the text transcripts must go into the long description in the post rather than the short description in the alt-text.

This is not by user request. This is an accessibility rule that I follow.

Now you may say that I don't have to deliver such an enormous infodump at once on a silver platter, whether people want it or not. You may say that they could always ask if they want something more.

But seriously, this is about accessibility. And if people have to ask and then wait for assistance, it isn't accessible. They could just as well ask for the whole image description, and if they don't, I don't have ot write it. It wouldn't make much of a difference.

#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #A11y #Accessibility

Thu, 17 Oct 2024 23:37:08 +0200

Jeffrey D. Stark

JStark@mastodon.social

@jupiter_rowland I would generally defer to: https://www.w3.org/WAI/tutorials/images/decision-tree/

Fri, 18 Oct 2024 00:31:10 +0200

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

@Jeffrey D. Stark I know that decision tree, and it doesn't really work for my original images because it's limited to what you'd usually find on professional/commercial static websites or blogs with a very small choice of kinds of images.

Let's take this image as an example here. It was the first image I've described in detail. I'm not going to link to that description because it's hopelessly outdated and probably terribly lacking at only a bit over 13,000 characters, and the image does not have a descriptive alt-text yet. It's an old shame, so-to-speak.

Does the image contain text?

Let's say, what the image shows within its borders has 22 bits of text on it, for any definition of text. This could count as yes.

However, of these 22 bits of text, only three are legible in the image as it is, at the resolution at which I've uploaded it: three times a capital M. As I've said, for any definition of text. The other 19 are so small that they're illegible, or they are so small that they can't be identified as text, or they're so tiny that they're invisible at this resolution. The big black sign in the middle with the yellow writing on it has a tree in front of it.

Is it still a yes because, while it isn't readable, there still is text?

Is it a yes for the three capital Ms?

Is it a no because the text does not show itself as text in this image at this resolution? So technically speaking, with the exception of the three capital Ms, there is no text in this image because where there's text in the original, there's just some blurry mush that does not qualify as text in the image as shown?

Or is it a no because I can't transcribe it anyway if I can't read it? Fun fact: I have transcribed all this text. 100% verbatim. And provided translations for everything that isn't English. So "you can't read it anyway" doesn't count because I can read it.

Now comes the kicker: If we pick yes because there is text in the image, the decision tree implies that it is an image of text and nothing else, and that the image-describing process is over after transcribing the text. This wouldn't even work with an image macro.

Does the image contribute meaning to the current page or context?

Does this question even work in this case?

Let's say the post in question is about the world where the image was taken shutting down soon and my avatar in this world disappearing. Because it was when I first posted this image. It's just meant to be a last farewell.

It doesn't add any extra information. This is not a post in a professional commercial or scientific or technological blog. So, does the image contribute meaning, yes or no?

If so:

It's not "a simple graphic or a photograph". It's a digital 3-D rendering, and it's anything but simple.
It's not "a graph or complex piece of information". Complex, yes. But it isn't a graph, and it isn't a piece of information of the kind you'd have on a scientific website.
I don't think it "shows content that is redundant to real text nearby".

Other options aren't available.

Is the image purely decorative or not intended for users?

That's debatable. On professional websites and blogs, this question may make sense. In social media where nobody adds decorative images to posts, it doesn't. In the Fediverse which has way higher accessibility standards than 𝕏 or Facebook or Tumblr or Threads or Bluesky or LinkedIn, it makes even less sense.

Is the image’s use not listed above or it’s unclear what alt text to provide?

Probably.

The top of the article is already a dead give-away: This guide is not meant for social media. Not for the big commercial silos, and even less for the Fediverse where Mastodon re-defines what makes an image description good. Not even two dozen people in the world use social media that support HTML <img> tags in posts.

While professional Web accessibility experts will throw their hands up into the air in utter outrage over 250 characters of alt-text because it's too long, Mastodon users celebrate alt-text that's four times as long.

So this is a case where a whole bunch of edge-cases unhandled by the WAI meet in one place:

A situation in which not the audience comes to the content, but the content comes to the audience without the audience necessarily explicitly seeking out this kind of content.
Social media with possibilities that vastly exceed those of the big commercial silos, especially regarding character count, while at the same time mostly not supporting full HTML.
An audience which has defined its own accessibility "rules". Including blind or visually-impaired people who do want to know what something in an image looks like, even if the WAI alt-text guidelines forbid describing it.
Content that has never been taken into consideration by any alt-text/image description guide out there. Ever.

Trying to force this into the WAI or WCAG guidelines is akin to trying to push a square peg into a round hole that was drilled by people who think all pegs are round.

I mean, as you've just seen, I can't even clearly answer any of the questions in the decision tree.

#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #WebAccessibilityInitiative #WAI #WCAG #A11y #Accessibility

Fri, 18 Oct 2024 00:58:52 +0200

Jeffrey D. Stark

JStark@mastodon.social

@jupiter_rowland you are working too hard at it... honestly... ... as a person who is blind with 3 family members across the vision spectrum... heck... even this is too verbose: [1/2]

Fri, 18 Oct 2024 01:00:09 +0200

Jeffrey D. Stark

JStark@mastodon.social

@jupiter_rowland a virtual environment that resembles a large, open, circular structure with a glass dome ceiling. Inside, there is a mix of natural and industrial elements. A person stands in the foreground, waving with one hand. They are dressed in a dark suit. Behind them, there is a tree with purple leaves, and several tables and chairs are arranged in a casual seating area. On the walls, there are two large screens displaying the letter "M" in a circle, along with other text and graphics.

Fri, 18 Oct 2024 01:00:11 +0200

Jeffrey D. Stark

JStark@mastodon.social

@jupiter_rowland The structure has a rustic, industrial feel with metal beams and a grid-like floor

Fri, 18 Oct 2024 01:01:12 +0200

Jeffrey D. Stark

JStark@mastodon.social

@jupiter_rowland darn 500 char limit on this mastodon instance hehe

Fri, 18 Oct 2024 01:21:24 +0200

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

@Jeffrey D. Stark

a virtual environment that resembles a large, open, circular structure with a glass dome ceiling. Inside, there is a mix of natural and industrial elements.

Yeah, but what do they look like?

A person stands in the foreground, waving with one hand. They are dressed in a dark suit.

Gender known, if so, which gender? Body size and shape? Rough age range? Skin tone or race? Hair length and style? Hair colour? Facial hair? Colour of the suit? Material of the suit? Footwear? Any accessories? Facial expression?

Behind them, there is a tree with purple leaves, and several tables and chairs are arranged in a casual seating area.

Again, what do they look like?

On the walls, there are two large screens displaying the letter "M" in a circle, along with other text and graphics.

Text? What does the text say? Why isn't it transcribed? And the graphics, what do they look like?

As for the extra details of the avatar description and the text transcripts, the questions simply follow several alt-text guidelines out there and what they demand.

There are certain things which, to my understanding, must be mentioned about a person in an image, especially if there's only this one person. And again, there is also the rule that if there is text in an image, it must be transcribed verbatim. Simply mentioning that there's text in an image and then not transcribing it comes over as not only lazy, but outright ableist, probably not only here in the Fediverse.

The "what does it look like" questions, in the meantime, are likely to come from blind or visually-impaired users who genuinely don't know what these things look like, but who'd like to know. After all, sighted users can see what these elements look like, and blind or visually-impaired users must have a chance to experience such an image in all the same ways as sighted users.

And I'm still not sold on the idea that all non-sighted Fediverse users know what image resolution is, and that they accept something being too small in an image to be recognisable as a valid reason to not describe it instead of an excuse to weasel yourself out of describing it.

#Long #LongPost #CWLong #CWLongPost #FediMeta #FediverseMeta #CWFediMeta #CWFediverseMeta #AltText #AltTextMeta #CWAltTextMeta #ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #A11y #Accessibility

Sat, 19 Oct 2024 10:21:01 +0200

Jeffrey D. Stark

JStark@mastodon.social

@jupiter_rowland if want that; I'm not too dumb to not be able to ask a 1 or many of the different image describing systems to find that out...

Sat, 19 Oct 2024 10:29:21 +0200

Jupiter Rowland

jupiter_rowland@hub.netzgemeinde.eu

@Jeffrey D. Stark But you shouldn't have to.

Besides, I'd like to see an AI accurately transcribe two lines of texts that take up six pixels in width and one pixel in height in an image. Or accurately recognise the texture on the clothes of my avatar at this resolution and this lighting.

Again, I can do both.

#ImageDescription #ImageDescriptions #ImageDescriptionMeta #CWImageDescriptionMeta #A11y #Accessibility #AI #AIVsHuman #HumanVsAI