Accessibility: Don't Use Fake Bold or Italic in Social Media | Hacker News Reader

Accessibility: Don't Use Fake Bold or Italic in Social Media

adrianroselli.com

64 pointsrapawel66 comments

Surprised at how poorly most of the screen readers seem to handle it. Will avoid the characters knowing this (not that I used them much to begin with), but I do also feel it's something that should be fixed on the screen reader side.

Completely understandable to fail at fake improvised fonts (like ꓄ꃅꀤꌗ) made from characters that just happen to look similar - but here the characters really are meant to represent their ASCII-range counterparts and Unicode defines compatibility equivalence/normalization to deal with it. Here the use is fairly frivolous, but most such character sets do have legitimate use - like fullwidth latin characters for readability when used alongside CJK - so really shouldn't just be skipped.

Isn't this by definition not a legitimate use? They are mathematical symbols, not linguistic graphemes. Readers could dynamically read out the description of every non-letter, though that could be more confusing than helpful in some situations. They could explicitly add these cases as letters, though that would make the reader attempt to read math as words. Simply skipping them seems like the worst option, though. Legitimate or not, it is in practice a real-world usage, so readers should probably account for it, though I can't say I blame them for not doing so. Deciding how to treat every block of Unicode case-by-case would be an impossible task.

> could explicitly add these cases as letters,

Unicode already defines compatibility equivalence[0]. Normalizing seems like the most sensible default for me - then it'd handle ｔｈｉｓ, ᵗʰⁱˢ, and so on.

> though that would make the reader attempt to read math as words.

For symbols originally intended for mathematical use, it probably then makes sense to add a special case to read out like T-H-I-S. But I wouldn't say that's strictly necessary - even in mathematical expressions you can have variables that are words rather than single letters, and either way it's definitely better than just skipping it.

[0]: https://unicode.org/reports/tr15/#Compatibility_Equivalence_...

Call me old fashioned, but maybe if people want to communicate "this" they should just type "this" and stop trying to be so Internet-quirky.

I think that comes back to:

> Here the use is fairly frivolous, but most such character sets do have legitimate use - like fullwidth latin characters for readability when used alongside CJK - so really shouldn't just be skipped.

The use given in the article is quirky - but these are real characters with proper uses and shouldn't just be skipped.

Old fashioned people dead centuries ago used formatting for their letters, no reason this is in any way "Internet-quirky", the real internet quirk stems entirely from silly platform constraints

This is not an easy problem. The relevant Unicode Annex #15 explains:

> The visual appearances of the compatibility equivalent forms typically constitute a subset of the expected range of visual appearances of the character (or sequence of characters) they are equivalent to. However, these variant forms may represent a visual distinction that is significant in some textual contexts, but not in others. As a result, greater care is required to determine when use of a compatibility equivalent is appropriate. If the visual distinction is stylistic, then markup or styling could be used to represent the formatting information. However, some characters with compatibility decompositions are used in mathematical notation to represent a distinction of a semantic nature; replacing the use of distinct character codes by formatting in such contexts may cause problems.

In the context of a mathematical text where the semantic difference is relevant, you wouldn’t want NFK[CD] to be blindly applied.

I agree, however, that screen readers shouldn’t silently omit those characters mid-sentence. At the very least, they should say something like “five unsupported characters omitted”.

To me normalization seems preferable over just skipping the text in all situations including math, and it's trivial to do, so I think makes for a sensible default. Certainly they could then start adding special cases, reading out symbols intended for mathematical use like T-H-I-S (though, mathematical expressions often still have words).

I'm not sure about "five unsupported characters omitted". Would probably need feedback from someone who actually uses a screen reader to be sure, but I feel that's less useful than normalization. Maybe a tone/voice alteration to indicate this? (playing with narrator, it does already change voice for control information)

The solution here is not to tell people not to write this way, it's to tell platforms to do better about giving people formatting options for their text

No one would be doing fake bold or italics if people could actually just put in real bold and italics

Both things can be good ideas at the same time.

It's likely the case that large corporate social media platforms have thought about what formatting options they want to provide and have decided that they will make more money if they keep it as it is. Attempts to get them to change that aren't somehow connected to their profits are unlikely to succeed.

It's possible to convince people to change their behavior by appealing to empathy, sympathy, or a sense of justice.

The current behaviour is: no native bold or italic, but weird Unicode bold and italic if you know the weird trick.

To me this seems unintended, rather than some thoughtful, considered optimum.

I imagine it has to do with not wanting to add anything to the UI, though it wouldn't be hard to support Markdown with no visible UI.

That ship has sailed, unfortunately. On social media where no markup is permitted (or possible), people will use these to stand out — especially those who make money on those platforms (not much of a problem on Mastodon though). We, the people who know how this works, can explain all we want, but it will always sound like a hypothetical edge case, even though this does exclude some people from participating.

The only way to stop this would to be to have the platforms where this is abused prevent its use (which would harm legitimate use too, although that is not nearly as prevalent as the 'I just want bold and italic'-case), have those platforms introduce minimal markup (not necessarily wanted), or have browsers and screen readers support these characters (not trivial, I suspect).

What’s the reasoning behind lack of B/I text on social media? These are mature platforms, I can’t think of a good reason to prefer “math symbols as style” over actual B/I characters?

My guess is that any user input goes through strip_tags() as a precaution.

Abuse. Twitter/X allowed bold text recently for a few days, and a number of monetized accounts immediately started making all their posts ALL BOLD ALL THE TIME to stand out. Feature dropped soon after.

This would explain a ban on some high % of bold text in a post, not on making this % 0

Generally, Mastodon supports at least a subset of Markdown syntax.

Putting text in asterisks for example, will render it in italics.

There's no reason why the various SM platforms can't do this.

Couldn't screen readers apply unicode normalization based some heuristics, like seeing the continuous presence of those special bold/italic characters? To improve accuracy, it can even check if the normalized text resembles to some English words or phrases or not.

When platforms were more open (ahem, Twitter), there used to be a bot you could mention in the replies of offending posts that would post a recording demonstrating how bad/inaccessible the experience was with the offending content in screen readers. Used it a few times to demonstrate on some big tweets, good eye opener for ppl abusing math symbols. iirc one was to demonstrate when Wordle exploded and everyone was posting those little green and yellow emoji squares. Granted, that was not an easy one to solve and never really was as the best option, using an image of the results, was not easily do-able in the pipeline from game -> sharing to social.

What really irks me is social media managers/content professionals posting that way on corporate accounts, sports team accounts etc, when they should know better, or bein accessible/inclusive should be on their radar more as it's been a topic of note for a number of years with the rise of ALT text considerations etc. Shows their inexperience.

This is great for probing scammers, though. Especially in non-English languages as they often use translation tools. And those have the same problem in recognising e.g. “ℍ” as “H”.

At this point I wonder if it’s time we add a kind of markup control scheme to Unicode so we can finally have a standard for rich text that is backwards compatible.

ANSI escape sequences exist and are good enough for basic formatting like bold, italic, underlined etc.

Most screenreaders also respect ANSI escapes! A lot are using a tty interface under the covers.

However, HTML does not.

Though, in my experience, getting Chrome to adopt something new, is easier than getting Dragon to fix a bug.

Accessibility: Don't use color in your paintings because there are color blind people.

Thankfully, the other root comments here have all of the alternative suggestions. The theme of which is Don't blame the author because the [accessibility] tool is broken. So I hope you can forgive me for not enumerating them again.

But I still want to take the alternative stance and assert you should use unicode in a previously novel way. Worry more about how you want to express yourself, and less about being well understood by everyone. Does being fancy here mean that it will be harder for some people to process and understand your message? Yes! But if that's a good thing or a bad thing should be reserved, exclusively for you to decide.

The author chooses to restrict information, the site they used to generate the text. Because you can't be trusted to take their advice. And the chance that the person reading your text might be visually impaired, and the screen reader they use might choke on that text, is more important than what you want, or how you want to express yourself.

I reject this idea.

I wonder how they'd feel about upsidedown text, in an attempt to artistically convey the author's state of mind where they themselves feel upsidedown.

I do agree with the author on the point that you should make any messages you wish to convey, as easy as possible for any one to receive them. But never at the expense of the message, or the art you choose to give your ideas. There are people who will never be able to understand you, and there will be people who choose to misunderstand you. And there will be people who tell you how you must express yourself, and it's seemingly always to make them more comfortable. Sometimes it's ok to ignore them, and sometimes that's even the responsible thing to do.

This will be a shocker I'm sure, but I think the demonstration (and the article in general) is for those who are receptive (or at least are choosing to be receptive) to this issue, and may not be aware.

Not entirely sure why you think the assertion of your own opinion over the supposed assertion of another's is any better by the way.

> Not entirely sure why you think the assertion of your own opinion over the supposed assertion of another's is any better by the way.

I don't think my opinion is better. I'm disagreeing with the take that it's never a good idea to do something fancy. I think the assertion the article makes, which from my read can be reduced to:

> you're wrong if you care about fancy text more than you care about screen readers

Is problematic because, generously it's too simplistic. The author feels empowered to tell others they're wrong. So I want to do the same; because using things in weird ways they were never expected is the hacker ethos. And I think the world would be improved by more people coming up with more fun things, or doing things just because they're cool. We need fewer people who tell others they're wrong for doing something fun, and more people who do things because they're fun.

In other words; because this problem can be fixed by improving screen readers, we shouldn't fix it by reducing the amount of cool shit people can do.

> This will be a shocker I'm sure, but I think the demonstration [...] to this issue, and may not be aware.

I'm not shocked, I simply don't agree. The author was able to write it as a suggestion, and make the argument using "should" but chose to assert it as an affirmative responsibility.

> I don't think my opinion is better.

Not what I was suggesting either. I was talking about your assertion of it. To be more explicit, what you're doing is

> The author feels empowered to tell others they're wrong. So I want to do the same;

a tit-for-tat. If you claim that asserting a preference or an opinion is wrong (which I'd agree with), you were then at fault for doing the exact same thing. And just like the author, you merely felt justified to act that way, and would have been able to make do otherwise, but chose not to.

> And just like the author, you merely felt justified to act that way, and would have been able to make do otherwise, but chose not to.

I'm not arguing for tit-for-tat. I think it's dumb. I took the affirmative stance I did, only to match the tone of the article. Adding the same absurdity to the counter point. I think we're both equally wrong for taking an unnuanced stance. The author and I both took stances that exclude important nuance. I was partially hoping partially expecting the remainder of what I said would come across. That idea being, options are better, warnings are better, absolutes are problematic.

> The author feels empowered to tell others they're wrong. So I want to do the same;

It would have been better for me to say "So I guess I'm required to do the same here". I'd much rather just suggest options, you can tell because I said as much

> But if that's a good thing or a bad thing should be reserved, exclusively for you to decide.

I don't want to make your decisions for you. Some of the authors points are compelling, I don't want anyone to die because they missed the tsunami warning either. But I find their specific arguments here needlessly restrictive. They're not warning, or reminding. They're saying your social media posts are wrong if you don't do it this way, and accusing the author of being irresponsible. Something *do* need to be easily readable, but that important nuance wasn't included in the article, just like I didn't include how tsunami warning without a doubt, need to be as accessible as possible.

You make a good point about being quirky. People using these "font" tools for Unicode should at least know about the problems it creates so they don't think it works smoothly like proper font styling. Personally my brain classifies overuse of Unicode as visual spam and just moves on to other more plain content.

But if that's a good thing or a bad thing should be reserved, exclusively for you to decide.

Target had to pay $6,000,000 because of accessibility problems with its web site.

I sure hope you're not in a position to make decisions about your company's web site.

To be clear, Target had to pay a lot of money to disagree about what U.S. law requires. It would have been cheaper to fix their accessibility bugs and even cheaper to build with accessibility in mind in the first place.

It absolutely would have been cheaper! But cheaper or not, target still should have. There's no ethical reason for a large corporate website to be inaccessible.

> There are people who will never be able to understand you, and there will be people who choose to misunderstand you.

I appreciate in your haste to be mad about something, you helped me prove my point.

I was talking about places where you'd reasonably prefer artistic expression, not large corporate websites. Given different circumstances, I would make different decisions.

You might want to educate yourself about the history of accessibility, the rationale behind many of the design decisions, and generally how accessibility works.

The comment is so ridiculously wrong I don't know where to start. Please, feed the comment to ChatGPT and ask it to explain to you how things work.

As someone who is forced to use accessibility software - please don't just punch me in the face for funsies.

Yes, most accessibility software sucks.

However, translating intent is harder than translating text. It is not always reasonable to transliterate, and knowing when you should is effectively undecideable.

Most accessibility software is more concerned with stability, and the huge number of devices that they need to work with. If your keyboard suddenly decided to be a T9 with autocorrect on everything you interact with, it would likely result in anger.

I am not simply reading your information in a different way. The intent of the information is also conveyed in another way. I don't know if it's a link, unless you tell me. I cannot tell if you're using Greek symbols for a mathematic equation or for typesetting, until you tell me.

Use the standards. It works for both of us, and is less work.

You don't just prevent accessibility software from reading it, you likely stop most software from reading it: spell checkers, search engines, things that read it aloud (which anyone can benefit from), AI. If you want your message to get less attention, this is a good way.

Trying to be cute almost always has negative accessibility impact. Instead of building a long list of thing to avoid or things to do, I've found it easier to just avoid being cute.

Don't underengineer your social media to fail at basic formatting

Isn't a screen reader's job to audibly convey information on the screen as it would be interpreted by a sighted person?

The original intent of the characters as mathematical symbols is obviously irrelevant in the examples. No person reading the posts with their eyes is thinking "oh hmm I wonder what is the mathematical meaning behind this bolded character".

This seems like an example of a technical implementation that misses the point of the value that needs to be delivered to humans using the system.

Sighted people can tell the difference between mathematical symbols used as letters vs used as math. There are ways to solve that problem programmatically, but the point is that it is a problem to be solved, not some kind of refusal to do the "obviously correct" thing.

Laser precise insight. It just so happens that being able to "audibly convey information on the screen as it would be interpreted by a sighted person" is technologically nontrivial, even now in the age of advanced multimodal LLMs. All the while screen readers have existed for decades.

So until that happens, another option is for people to just... not engage in Unicode abuse like this. Or at least be mindful of the fact that if they do, screen readers will take a massive shit, so if that's not their goal, they should reconsider. Which is what the post is all about (as well a general plea to folks to at least mull it over twice if it is their goal).

Not sure why you need to jump to LLM hallucinations as some kind of advance when a much simpler method of looking up a sequence of letters in a dictionary would suffice

"Existed for decades" sure, but some design bugs persist for centuries, that's not enough to be confident a fix is very hard

Because fixing this specific problem with a heuristic is not the same as "audibly convey information on the screen as it would be interpreted by a sighted person".

Have AI write you a browser extension that replaces such characters with their regular versions.

I don't know why ppl downvote you (perhaps it's a lazy non-contributing response), BUT.. I am using ChatGPT to create my own Add-ons for little things/annoyances/improvements on my Firefox. They probably won't help anyone else, but it's fun to see an idea materialize in 5mins or less!

It's because people don't want their silly coding jobs taken away, and they don't want others to be maximally empowered. They are shallow and feel threatened by what's happening.

When the sarcasm is so strong it wraps around and starts sounding completely serious.

It's because that puts the onus for coming up with a solution onto the millions of end users rather than the handful of devs who created the problem in the first place

Why do the browsers not fix this?

It would be like, five minutes to write a text preprocessor in JS to solve this problem,

It's intended behavior. They are mathematical symbols and have a different semantic meaning.

And yet every human that sees them used this way gets it. It would not be that hard for the screen reader software to do similarly.

Difficulty isn't the problem, as stated. The issue is that software doesn't know what "used this way" means - only humans do.

Please. You could get good results with basic heuristics, without even using AI.

They can just add a toggle to enable math mode, since I'm guessing decorative use is many times more common except maybe in school or something.

I'm not sure I've ever seen a document where I'd want in read as anything but normal Italic text, the only time I ever see typeset math is PDF datasheets.

It would be fairly trivial to detect: configure the mapping to the Latin alphabet and run a few words against a dictionary.

Semantic meaning lies in the semantic organ, which easily reads it as formatted not

Making assumptions about the semantic intention of characters is dangerous (it falls into the category of "cute" behavior of software attempting to be helpful). If you do this, the reader will attempt to pronounce mathematical formulas like words.

Admittedly this is a common mis-use of these characters, so I'm not sure what the best compromise is, but there isn't an obvious compromise (hence the answer to "why don't they fix this").

Is it really a common mis-use? I don't think so.

Yes. I see it all the time on YouTube in video titles for example. Common with specific genres of music videos.

Maybe you could train an LLM to recognize formulas (how else are you going to speak a formula coherently?) and treat the rest as normalized text.

Isn’t the “compromise” for platforms add proper text editors to their input fields (lessening the need for using symbols-as-text)? This should be a solved problem.

Because it's not on the browser to fix it, unless it provides screen reading capabilities of its own, which e.g. Google Chrome doesn't. Instead, Google Chrome specifically will make a so-called accessibility tree available through whatever APIs for external code to parse and evaluate.

The relevant parties involved are the websites hosting the content, the people producing the content, and the developers developing the screen readers (and potentially the admins who control the screen reader application that is deployed). None of these are the browser.

Absolutely no one should stop using bold or italics just because screen readers fail to parse. Screen readers need to do better.

Bold and italics add spice to otherwise boring typography.

Screen readers do actual B/I fine. Unless I’m missing something, the case here is using math symbols in place of B/I where the platform doesn’t allow native styling.

This is not about "spicy" typography but Unicode abuse...

Reply on news.ycombinator.com