Meeting Minutes
Lisa Liskovoi: We're at about 5
minutes past, so I'll get started. so Wendy is away today, so she sends her regrets for not
being here, and we'll be talking about research. So I'm just going to drop the framing
question in the chat.
Broadly speaking, we're looking at. Looking at the current version of
ATAG, what are some things that we want to research and what are some methodologies that we want to
use in order to support that research?
Looking at the current version and future improvements,
what research can we do to support that work? What methodologies or approaches should we try?
and
I'll also drop some of the questions that have come up in our discussions.
that we mentioned
that we might want to address through research. So this is just to get you thinking, but just as a
starting point, if you have burning questions that you want us to talk about, don't feel like
you have to stick to the list at all.
So with that, I'll pass it to the room.
Lisa Liskovoi: Why was ATAG not successful in being adopted by policy
around the world?
Why was ATAG not successful in being adopted into procurement policy?
What
is the relationship between policy and downstream considerations like procurement?
Will browsers
and AI chat apps soon have inherent AI-based capabilities to make everything accessible? (Can they
build such features reliably?)
What are the implications of security and privacy in AI-integrated
authoring tool interfaces or browsing experiences?
If generative UI becomes a first-class feature
in something like a web browser, or through a website that browses on your behalf, does it become an
authoring tool?
Abhinav Kannan:👍
Ned Zimmerman: Yeah, thanks,
Lisa.
Something that I have come across in starting to do some
research around this
topic.
is the
So, I guess to say, particularly with respect to the
use of LLMs and AI in
authoring tools that becomes challenging.
is
I think it would be worthwhile to
research
what existing
standardization is
out there around
or our standards are out
there around natural natural language interfaces.
Because when we're talking about AI
authoring, a lot of that is done using either voice or text prompts in natural language.
A tag 2
doesn't touch on that at all, for obvious reasons.
but
Because that's something that
we're trying to address.
in an upcoming version, I think
understanding what's out
there around natural language.
interfaces would be really helpful to inform
any parts of a
newer ATAG standard that address that.
I did do
a preliminary scan of some standards and found
a few.
ISO standards related to language interfaces, but specifically they are
language
interfaces that don't use natural language processing, so they involve, like,
predefined
commands, which is not applicable to LLMs because they're non-deterministic
and
respond differently to the same prompt, depending on, you know,
a number of factors. So, I
think it's an interesting area that
would be valuable for us to research,
towards this
newer version of ATAG.
Jutta Treviranus: Yeah, I think,
given
that ATAG was very much
quite formulaic in terms of create content that adheres to WCAG.
it
would be good in this round, especially knowing what's happening on the,
the WCAG 3.0 and
other areas to tackle
the notion of,
And the very gnarly and somewhat difficult notion
of
personalization or individualization of interfaces, and so
rather than I mean
Do we want
to open up that debate again, especially given the messiness of what's happened with overlays?
but divorcing
the business practices from the actual opportunities that are offered by
being
responsive to the individual needs of the
the user, or whatever
I think it would be good to
explore that.
I'm not sure exactly how to state the question, but or the research
question.
Andrew Puchle: something that comes to my
mind in this kind of builds on the non-deterministic
case that was mentioned earlier, that it and
it's true, like, every time you tell it something, it'll be slightly different than the
time before it.
what is the delta between its actual outputs and its claims made about it, right?
So there's a lot of, like,
you know grandiose claims about how it's gonna do all these
things, and does them well, and so forth.
But it in practicality, it really doesn't. And so,
you know, what
What is that delta?
current state, and then what can be done from an authoring
tool perspective to ensure that
it's mediated, right? So
You know, if you tell the AI to
make something following accessibility guidelines how consistent is it in doing that, right?
Lisa Liskovoi: Looking at the current version
and future improvements, what research can we do to support that work? What methodologies or
approaches should we try?
Miriam Fukushima: yeah, what I would like
to know would be kind of like some hard data on from actual companies in, like.
Kind of a
questionnaire or something, a survey on how familiar they are with ATAG, if they know if it even
applies to their products, or why. And then maybe if they answer that with yes, then some more
detailed questions about specific areas of ATAG, and.
Maybe a scale from 0 to 10 on how difficult
to implement or to understand. They found that a specific part of ATAG, and then we have some hard
data on which areas need most work.
Ned Zimmerman: Thanks, Lisa. the other
thing that I have been thinking about a bit is the
change in the content authoring landscape
since ATAG 2 was published.
I think what we see now is a lot more of the sort of full-site
editing,
full site builder interfaces along the lines of Squarespace and other tools like
that.
But we're also seeing in the AI realm, we're seeing these
just type a prompt
and we'll build and design and lay out your whole website.
offering experiences that
are
completely different from what I think ATAG2 would have conceptualized.
And so, I
think
I guess understanding the different constraints and requirements around
a full-site
editing experience versus a more
old-fashioned content management system where
you're
editing pages in a CMS that are associated with templates that have been built
elsewhere.
I don't really know what the best research approach for that was. Lisa and I were
chatting about that earlier, but
I think maybe some sort of
even, like, a scan of what
Saif Altalib: ATAG for Vibe Coding and
Agentic AI
Ned Zimmerman: vendors are prominent now in
the CMS
space I know there are there's always the stats about WordPress's usage that
are
that are available, and how how dominant WordPress is by market share in content management
systems, but it would be interesting to see
like, what the top content management systems or
authoring tools are that are in use globally at the present moment.
And how they
where they
fit into that spectrum between
an old-fashioned, I'll say old-fashioned and scare quotes,
CMS and a full-site editor.
along the lines of a Squarespace or something.
Mike Gifford (CivicActions): https://almanac.httparchive.org/en/2025/cms
Ned Zimmerman: Yeah, oh, that's
great.
If that's what I think it is, fantastic.
Mike Gifford (CivicActions): Also, in that
regard https://almanac.httparchive.org/en/2025/accessibility
Jutta Treviranus: Yeah, I think one thing
that I'm not sure if this is a research question, or if it's a,
some framing that needs
to be done before we actually
refresh ATAG.
And it's revisiting that, uh
Division
between what is authoring, what is user agent, what is
web content because, of course, it's
always been somewhat of a messy
area, or
delineation but it's becoming even messier now
in terms of
reformulating the content post
whoever is determined to be the author, etc.
So
working on or looking at,
How we delineate that at this point.
Mary Ann (MJ) Jawili: Yeah, I think maybe
related to what Jutta was just saying,
What I
would like and I don't know if this was a
WCAG thing, but
I would
The example that comes to mind is, like,
Live captions showing up,
or being available in the browser or an operating systems, and
like, author
content authors,
like, saying, well, we don't need to specify the captions anymore, because you could just turn
it on.
We don't need to provide captions, because you can just turn it on on your operating
system or browser, and
I guess I would love to underst-
I don't know, I guess I'd
like for us to ex
explore, like
How much of that would be is acceptable now, or if any of it
is acceptable?
or what do we still expect people to
to at the authoring level. So maybe
related to what Yuta was saying?
and regarding how to do that kind of research,
I don't
know, like
asking people,
if what their experience has been like, and
trying out
AI
solutions and seeing if they're
adequate whatever adequate means.
Yeah.
Lisa Liskovoi: I think this is a thought
that's related to what both you and Yuta have said, and also a little bit of our discussion
before, and that distinction between what is the authoring and what is the user agent? You mentioned
personalization earlier on, and I think maybe one of the things that we can think about
is.
Again, thinking about the two sections of ATAG, like personalization for the author, and then
personalization like what needs to be in place in order to allow for personalization that's of
a certain level of quality for the end user, and because I think there's still a little bit of
a.
maybe misunderstanding about how much has to go into the creation process in order for the
content to be.
adjusted or modified or presented in different ways.
Jutta Treviranus: Yeah, just to comment on
that, I think this goes back to the old debate when WCAG 2.0 was being finalized of, is it
compliance at the page level or at the site level? And it was decided to go to the page
level.
where the site level would have allowed for
all of the criteria to be met by the site
dependent upon and that would have allowed for that individualization of
Charles Hall: From WCAG:
only
accessibility-supported ways of using technologies can be relied upon for conformance
Relied
upon
the content would not conform if that technology is turned off or is not supported
Jutta Treviranus: of content and I
don't know if there's appetite to reopen that debate.
Laura Staniland: I would love to see it be
focused on the site level because if the site isn’t accessible, then nothing else is really
accessible
Charles Hall: I just pasted this in the
Zoom chat from WCAG. But if we're.
If we're looking at where that proverbial line is
between author and agent and content.
Laura Staniland: If the platform can’t do
WCAG 2.1…then it’s not accessible.
Charles Hall: and saying that the user
agent provides the solution so the author doesn't have to. We have to look at existing
definitions.
that helps support us, and WCAG has relied upon.
So, if the agent provides it,
that's fine. But if that's a setting, and you turn it off.
Or in some scenario. where
it's it's not supported for Whatever reason.
premium tier, or authenticated or
something, then the the agent isn't universally providing that. It can't be relied
upon.
Sambhavi Chandrashekar: Page-level has
never made sense to me. Workflows matter, regardless of the path in the site a user follows. So, I
support site level.
Lisa Liskovoi: I'm also thinking about
sort of quality or accuracy requirements for anything that is automated or provided by the platform.
Like, I know a big issue in WCAG is that it requires captions, but there's no requirement
around how accurate those closed captions are.
so I'm thinking about that in other contexts,
like, if there are things that are automated.
Is there a way for us to even do that? Because, you
know, it might be a little bit harder for somebody to to kind of hard decide whether an alternative
text description is accurate enough compared to something like closed captions, which is obviously
easier to compare.
Charles Hall: WCAG never cared about
accuracy
Lisa Liskovoi: Charles, I see your note in
the chat that we've never cared about accuracy. Does that are you saying that in the sense that
we also?
can't or shouldn't be thinking about it in a tag, or Is that just?
Okay.
Charles Hall: No, I I yeah, sorry pretty
much agreeing with you. It's a gap.
Miriam Fukushima: Yeah, I was also thinking
about responsibility. If we say like closed captions are provided by the browser, then the content
that in which we that we open in the browser is not necessarily.
always opened by a browser, so
it could be other programs, other platforms, and just because we made the browser responsible
doesn't mean.
that other platforms are equally responsible and, And so I think to keep the
responsibility as close to the content as possible is probably the only thing we can
do.
Obviously, like, browsers are also authoring tools and whatever other software is developed
to view that content should also then adhere to ATAG. But if like an author of a content is asking
themselves if.
They could rely on other software to fulfill the aid tech for them. Then the
answer should be no.
Mike Gifford (CivicActions): Just to say
that there's often embedded tools like Drupal has to conform with with a tag. We're
trying, at least we're trying to make it conform with a tag. But cke editor is the WYSIWYG
editor that we use. That is a tool within.
Drupal. So there's a nesting of tools that are
involved, and there will continue to be a nesting of tools for authoring environments. So we have to
think about not just about the one system that we're evaluating, but that chain of systems
which can be modified and customized. But still, it's not just one.
It's not just one
tool. It's it's often a bunch of tools.
Laura Staniland: I think it was Mike who
just said something about a bunch of tools. I was thinking how they've pushed AI into all of
these platforms, and maybe that should be included in the, like Requirements is that
The AI
they've pushed in is also accessible.
I was just thinking about what's at the site
level, what's on a platform now that
Lisa Liskovoi: Mike, I know in the
discussion you posted some some findings from a research report around authoring tools. I don't
know if you want to speak about that a little bit.
Mike Gifford (CivicActions): Sure. So this
is the the Http archive does an annual scan review of of of the web and and being involved in in the
accessibility chapter and the sustainability chapter. And so so have have some insights on
that.
But there's also a CMS chapter, which is largely driven by WordPress people, because
they're.
They're a big part of that ecosystem. But it's the closest thing we have
to a snapshot of what CMS are actually being used out there. And I think that there really should be
more interaction between what.
is actually being done on the web, which the HTV archive is
probably the best.
sort of a summary of that, and what standards we're trying to build
towards. And there isn't right now any real connection between the W3C and the HTV archive. I
think mostly because we're all just short of time and energy, and.
And it's a bigger
project than we have the resources to do, but it does make a lot of sense if we care about the web,
that we both look at the standards we're trying to build towards and how effective we are at
doing that. So that's the thought there. But Miriam, you had a comment?
Miriam Fukushima: Me? Yeah, yeah, yeah, it
was just to your earlier comment, Mike, with a bunch of tools. I was just thinking that also that
was one of the problems we identified as problematic for the application of ATAG.
because people
weren't sure if they were responsible, or if someone else in the chain was responsible, or like
how to make it into smaller portions that it's not too overwhelming, and that you get a start,
and that you.
like, identify the problems and isolate, kind of, the workflows and the areas that
need to be worked on.
Mike Gifford (CivicActions): I mean, again,
I can speak to what the Drupal community does with that, and and so much of that is is about about
trying to fix the problem at the source. And so if we discover an issue with CK Editor, we make sure
that it's reported in Cq editor. Sometimes we can fix it ourselves within Drupal, but sometimes
we have to go off and wait for CK Editor.
Mike Gifford (CivicActions): to fix the
issue. Or look for ways to go and work around that, like the in terms of editing tools, there's
a really excellent editing, well, sorry, John Gunderson created a really excellent tool called Ali
First, which was using CCADER 4 to try and.
and create a better interface for accessibility to
prioritize the editing interface to be able to make it easier for people to create accessible
content by default. I've been advocating with the folks at CK Editor 5 for years now to try and
get that built in so that that sort of approach is built into CK Editor 5.
It's not being a
priority for them on the timeline. It's not something that they've been discussed. But
again, there's some things that you can do as your as yourself. There's some things that
you need to go off or have an upstream library maintain, but as long as you're documenting what
those things are and who's responsible.
that I think you've done your due diligence as
a project. So this is an upstream issue. If you've contributed upstream, and you've said,
this is what we've done to address this, and this is how I think it's about trying to make
sure that you're involving the community of software that you're using as part of your
project.
Laura Staniland: What is SeeKey
editor?
Mike Gifford (CivicActions): Does that make
sense?
Lisa Liskovoi: There's a question in
the chat from Laura, I think for clarification, Mike, around what is CK editor?
Ned Zimmerman: CKEditor
https://ckeditor.com/
Mike Gifford (CivicActions): So I've
got in terms of examples. And yeah, thank you for for that, Ned. So this is the example that that
I've I've actually tried to fork it. But what the so I have a cke editor 5 version running
that seems to
work.
https://mgifford.github.io/CKEditor5-A11yFirst/
https://a11yfirst.gitlab.io/
But the
initial CK Editor 4 issue is also, I just added that to the chat as well. And again, like this is
something that I did on the side of the plate, because I think we need to have something like this.
We need to have this discussion about what is possible.
And again, you know, the folks that in
Illinois went off and did an excellent job trying to outline what what a useful WYSIWYG editor could
look like for accessibility. And I've tried to bring that in as an example, and I would love to
see other people.
test and evaluate it, bring it in, like, how do we try and use these ideas and
push them forward?
Miriam Fukushima: Yeah, just to that, that
makes total sense to me. And I was just thinking on how we could incorporate that more into ATAG
than we had in the previous version so that people who actually try to apply ATAG.
Kind of get
the sense, okay, I need to break it down, I need to search for my closest responsibility and start
from there. And on how to make this more understandable, and, Yeah, more transparent on how to go
about it, because it's not clear for everyone.
Mike Gifford (CivicActions): Also
highlighting the Funka We4Authors study:
https://mgifford.github.io/We4Authors/
It's
also, I just wanted to. I put in a link also for the the we for authors cluster, which which was an
effort from Susanna Lauren and others to try and actually engage with Cms.
Saif Altalib: Is 24 September 2015 the
latest version of ATAG guidelines?
Mike Gifford (CivicActions): tool builders
to try and and evaluate what what we could do to support authors better. This was done in
preparation for the European Accessibility Act being released, and it was a came up with some really
interesting stuff that was distinct from ATAG. But again, it's an interesting sort of
like.
Ned Zimmerman: Replying to "Is 24
September 2015 the latest version of ATAG gu...":
I believe so
Mike Gifford (CivicActions): midway point
in terms of what were some of the limitations, what's a different way of thinking about it, but
unfortunately, with issues of Funca, the website was was dropped, and so I've tried to recreate
a copy of it. You can still see some information on the funca site, but it has.
There's been
some changes in how that organization and the leadership on that has run.
Saif Altalib: Thank you! So our goal is to
update these guidelines. Is there a plan of how we are going to do that?
Lisa Liskovoi: Thank you for doing that
work, like, for putting it together. Is there anything you want to highlight from that that you
think we should discuss today from a research standpoint?
Mike Gifford (CivicActions): I guess, like,
try to find the pain points, right? Like, what are the pain points for authors? What are the pain
points for one of the things we just haven't dealt with very well, and and I mean, this is what
we're trying to sort of get to with. I think that that that.
ATAG 2.0 was very much coming
from a more.
Rémi Bétin: Replying to "Is 24
September 2015 the latest version of ATAG gu...":
Yes
Mike Gifford (CivicActions): A theoretical
academic perspective, and with we for authors, it was coming from the site, this authoring tool
builders trying to say, what what do we want to have happen? What are the things that we can do? And
and there's things that just aren't like, how do you express it in a generic way that is
not.
That is a W3C friendly kind of way of conveying it that can be understood by the builders of
these tools. And I think it's, I mean, partly it's coming down to what is who is our
audience? Who do we want to go off into to listen to this? And how do we?
you know, as somebody
that says tried to implement a tag 2.0, now, like a long time ago. But but it's not easy to go
off and to think about how do these success criteria line up with what is it we're doing over a
really complicated interface like.
It's not a it's not an easy thing to go and figure
out how do we support ATAG A or B, but to try and do both is a it's a big challenge.
Miriam Fukushima: Yeah, I really think,
like collecting data would be really good in form of surveys, but how are our capacities for that,
like as as a working group and like.
Saif Altalib: Are we planning to maybe use
an LLM model to create first draft based on our expert insights and requirements...
Miriam Fukushima: Has that been done
before? Like we would really need maybe also not just us, but also Europe and like some just some
wide range data on how it's applied and how the actual.
process of going about it and like
scales, or yes or no questions just to get some some broad.
Numbers.
Abhinav Kannan: Yeah I had the same idea as
Miriam, which is collecting data and trying to validate.
So, what is a good suggested next step
in trying to collect this data? Should we go Should we go talk to should we have people use some
tools, get some feedback?
Do we need to get more feedback, or is there already some feedback out
there? What kind of research? What are some best next steps we can I'm just trying I'm
just throwing it out there. As you know, I'm really new. Yeah.
Ned Zimmerman: Replying to "Are we
planning to maybe use an LLM model to creat...":
This is worth reviewing:
https://www.w3.org/TR/2026/NOTE-llms-standards-20260324/
Saif Altalib:❤️
Lisa Liskovoi: I think as a first pass, we
should probably gather the information that's already out there, so we're not asking the
same questions if we do distribute a survey or something like that. So Github discussion seems to be
a good place to do that. We already have a thread dedicated to research. So.
Jutta Treviranus: … and this data
collection will be made difficult by the changing landscape. What we find out from the past will
likely be disrupted.
Abhinav Kannan:👍
Lisa Liskovoi: Maybe if folks want to
continue populating it with information that we know is out there, whether that's survey data
or other kinds of resources, and then.
from there, I guess we'd have to come up with the
questions that we want to ask, and then figure out where we want to distribute it.
Lisa Liskovoi: I don't know if folks
have ideas about how how best to do that next, or if you've had experience doing this within
other W3C context. Yeah, Charles.
Charles Hall: Yeah, I mentioned this last
time, but one of the research approaches that we held.
In the Silver Task Force, when WCAG 3 was
kicking off.
In 2017 was what's wrong with what we have?
So the the research was all
based around. mostly surveys, but other research methodologies, but around.
what are the current
problems with the current set of documents?
So, if if we're to approach. A tag from what we
think we might be off.
But if we approach it from what research shows that other people
think.
We might be better aligned to producing something more useful.
So I think poking holes
in what exists is a great research strategy. And one of the things that I think we should look at is
things outside of ATAG itself.
So other way resources that talk about ATAG or authoring or what
WCAG says about authoring.
is is what we have accurate, relevant, current.
Lisa Liskovoi: Yeah, Mike.
Mike Gifford (CivicActions): I was gonna
say that there there are the other thing is that there's there's ATAG has been around long
enough, and that the idea of supporting authoring tools has been around long enough. Also, the we
for authors clusters being around so
that.
https://www.drupal.org/project/issues/drupal?text=atag&status=All
There's an
opportunity to look in existing issue queues and existing content management systems on GitHub or on
other places to try to say, can we find references to this that will give us a hint of what are the
things that that content authors are thinking about?
Saif Altalib: One thing to conisder: Since
everything is in a state of transition/change, not sure if this is the right time to gather data...a
lot of uncertainties and ambiguity.
Lisa Liskovoi: One thing that I'm
thinking is that so far we've been looking at this primarily through a kind of gaps approach,
but Mike, to your, kind of case study that you mentioned earlier on. There's also a lot of
stuff that we can learn from things that are working, right? And I'm wondering if we
can.
Mike Gifford (CivicActions): https://www.drupal.org/project/issues/drupal?text=we4authors
Lisa Liskovoi: have a kind of, yes, what
are the gaps and what are the holes, but also in the entire time that ATAG has existed, like, what
are the things that have worked really well that we know that we could potentially replicate our user needs.
Laura Staniland: Are there other working
groups who might have similar research that we could incorporate?
Saif Altalib:👍
Mike Gifford (CivicActions): https://github.com/search?q=we4authors&type=issues
https://github.com/search?q=atag&type=issues
Jutta Treviranus: Yeah, and I think it
might be good to do sort of a futuring exercise as well.
Because this needs to last for a while,
given the slowness of
So, I think one of the issues with all versions of what has come out
of
way is that we didn't adequately foresee or
try to predict or,
how this would
change, and I think this especially affects the data collection, because if we're
basing our
decisions upon past data that will
that will, again, cause us not to have sufficient
Mike Gifford (CivicActions): The web is so
different now than it was 25 years ago!
Charles Hall:💯
Jutta Treviranus: understand, or yeah,
it'll create something that was for that moment, or the past moment, rather than something
that
is for the moment that the
the standard, or the guidelines will need to exist.
Lisa Liskovoi: Yeah, that to me sounds like
a good co-design activity if we want to get into that at some point.
Charles Hall: Replying to "The web is
so different now than it was 25 years a...":
The web is different now than 5 years
ago
Mike Gifford (CivicActions):💯
Lisa Liskovoi: Mike, I had a follow-up
question about something you said around how difficult it is to implement ATAG. Ned and I were
having a conversation about it this morning, and we were wondering if that is more a product of the
requirements themselves, or is it that.
WCAG has this rich, you know, pool of techniques and a
lot more guidance, supporting guidance around how to meet the standard. In other words, should we be
putting effort into thinking about how we structure the requirements themselves in a clearer way? Or
is there a need to.
have more around the requirements of supports authors or creators of
authoring tools in meeting those.
Charles Hall: WCAG difficult to implement
came from difficult to understand
Mike Gifford (CivicActions): I mean, the I
think that that the reason that there's so much additional content around WCAG is that
there's an industry behind it, that there's an industry around reporting. And it's
largely a reporting industry. It's not necessarily an industry of.
Fixing problems as much
as of reporting problems. Again, difficult to necessarily predict, and reporting is the first step.
But I But I think that that part of it comes down to like, what are the the components that people
are using to go off and to to build.
to build the web. And so, like, even trying to think about
it in terms of things like, you know, what? What are the elements of an authoring interface that you
are going to that you're going to be building that are going to be repeated patterns throughout
a system, because, like, the WK does not think about systems. It does not think about.
Saif Altalib: When we say "Authoring
Tools" are we primarly referring to the design and development of on-demand courses or any
content?
Mike Gifford (CivicActions): websites. It
does not think about content management systems. It thinks about pages, and that is a major
limitation when you're thinking about saying, well, how do we.
How do we fix this, this, you
know, the million web pages that run, or sorry, a million web sites that run Drupal that all have
the same error because it's part of this template that screwed up over there? Like there just
wasn't the thought that the web would be.
different, or is it as complicated as structured
as it is, which, again, was hard to predict 25 years ago, but but I think that if we were thinking
about fixing systems, we need to be able to understand those systems and communicate them in a way
that.
The site builders can understand or relate to.
Lisa Liskovoi: I could not agree more,
yeah. And the first step of that is to think about it in systems, which, like you said, is not
there.
Mike Gifford (CivicActions): I don't
know how to get there, though. That's it. Like, other than like, and same with like trying to
get the trying to get the money involved to do this like, if the money was there, then this would be
a relatively easy thing, like.
Charles Hall: Like all the <a>
elements from the default AEM link component that use the alt attribute which is invalid
HTML.
Mike Gifford (CivicActions): Like, but
it's, it's talking about like orders of magnitude more money than, than, say, even the,
the Accessible Standards Canada is going to be able to bring to bear to this. Like it's This is
not a few researchers here or there going off and putting together a report. It's about trying
to say, how do we shift how accessibility is done so that we are fixing, we're shifting left to
where the problems are created, which sometimes is with the tools.
And sometimes it's with
the training, and sometimes it's with you know, but it's not we're not fixing it at
the right place, which is why we're not as accessible as we want to be.
I wish I had a more
sort of like rah, let's fix this sort of solution. But it just like, yeah, this is a big gnarly
problem.
Lisa Liskovoi: Well, acknowledging it as
the first step, right?
Ned Zimmerman: Replying to "When we
say "Authoring Tools" are we primarly refe...":
From the standard: “Any web-based
or non-web-based application(s) that can be used by authors (alone or collaboratively) to create or
modify web content for use by other people (other authors or end users).”
Saif Altalib:👍
Lisa Liskovoi: I know there's been
some discussion happening in the chat that I've been following less, so if anybody wants to
jump in and just speak through it a little bit, that'd be great.
Ned Zimmerman: Replying to "When we
say "Authoring Tools" are we primarly
refe...":
https://www.w3.org/TR/ATAG20/#def-Authoring-Tool
Laura Staniland: Oh, yeah I was just gonna
say, is there research that other groups might have as someone who works in a large institution that
can be siloed?
Sometimes the most effective way is simply to start talking to other people and be
like, hey, have you done this? Because we may be recreating the wheel. There may be other working
groups who have some data that we can incorporate. It may not be all of it, but it may be helpful,
and then we don't have to do that.
Charles Hall: Yeah, the the. Wendy was
going to go and talk to the AG chairs about the research that AG has on WCAG.
to see if any of it
might be useful.
But external, I don't know if anyone's done
Miriam Fukushima: And maybe we could put as
one of the goals to have some additional documentation around ATAG, similar to the success criteria
of WCAG.
Mike Gifford (CivicActions): HECVAT -
https://www.educause.edu/higher-education-community-vendor-assessment-toolkit
Miriam Fukushima: where maybe categories of
systems are actually listed, like Mike said, maybe we could like divide ATAG from an approach of,
okay, do you want to make an UI interface? Do you want to make a, like, text?
editing box, or do
you want to make a chat interface, or, like do you want to make a an authoring tool that processes
documents or something like that, which, like, if you if we have these little systems.
Laura Staniland: I like that idea
Miriam Fukushima: as a starting point for
developers, maybe that is similar to the success criteria of WCAG. Maybe that helps to
like get it easier to
apply.
Mike Gifford (CivicActions): https://docs.google.com/spreadsheets/d/1SlsIivkNfwwJbBEzgGI9UGhVWnJpqiDNd2VjEy097iU/edit?gid=1025021328#gid=1025021328
Mike Gifford (CivicActions): Laura, I
mentioned the the what is it? I don't know how to pronounce it. The heck that which are you
familiar with Hecvat?
Laura Staniland: Sort of
Mike Gifford (CivicActions): So this is a
procurement tool that is being used by a bunch of different universities in the Us. To try and set a
higher bar for a range of different procurement issues, including accessibility. And I had to check
to see whether or not this actually includes.
ATAG, and I did not see anything on ATAG or,
There's there's yeah, there's there's Yeah, there's there's only one
sort of line on authoring, and it's just it might be useful to go off and to pressure those
those universities to try and say, how do we think about authoring tools? Because that's pretty
pretty critical part of the problem that we should be thinking about.
Saif Altalib: Do Authoring Tools include
IDEs and Vibe Coding tools?
Miriam Fukushima:👌
Laura Staniland: Yeah, so, I had somebody
reach out to me because they had this form that they have students fill out in a Google Doc, and
it's 38 pages, and of course a student using AT couldn't had some problem along the way.
And when I went into remediate it, it turns out Google Docs don't really have a built-in
accessibility editor like I could run a test.
And I can re-mediated myself, but, like, you
can't even go in and mark an image as decorative. So, like, at the most basic level, I said,
you know, you shouldn't keep using this and so that's, like, a huge platform, like Google
Docs right there, like.
How doesn't this have basic accessibility editing? But yeah, when we
are looking at things, if I see that the VPAT.
Doesn't meet WCAG 2.1 AA, which is, you know,
what we're working towards. It's like, nope, you're out of the game.
Mike Gifford (CivicActions): Google Docs
has some big issues. MS seems to be better.
Lisa Liskovoi: Replying to "Do
Authoring Tools include IDEs and Vibe Coding to...":
Here is the discussion regarding types
of authoring tools so far: https://github.com/w3c-cg/atag/discussions/12 please feel free to add to
the list.
Jacob Wood: Replying to "Do Authoring
Tools include IDEs and Vibe Coding to...":
I had a similar question. Do code editors and
other interfaces count as authoring tools? (Yes, in my opinion.)
Laura Staniland: But how much of that
It's covered by ATAG, I'm not sure, I I just got back into this group now that you made it
10AM, so but, like I said to the person, if you can't even edit a Google Doc for basic
accessibility, don't even use this platform anymore, so
Charles Hall: Collaboration tools are
mentioned in the WAI research list:
https://www.w3.org/WAI/research/
Abhinav Kannan: Yeah, people I know use
GrackleDocs — a plugin for Google Docs -- to check and fix accessibility issues.
Laura Staniland: I don't know if
that's a relevant response. I think it is.
Yeah, I just don't think you should have to
use a plugin.
Mike Gifford (CivicActions): I’d encourage
folks to ask for OpenACR rather than VPAT PDF files - https://acreditor.section508.gov/
Charles
Hall:💯
Laura Staniland: For a platform. Like, a
large platform as big as Google shouldn't need a plugin for accessibility
Ned Zimmerman: Yeah, just to just to add to
that, I mean, it's interesting, that just makes me think of the fact that in the ATAG2
standard,
The valuation of authoring tools is of the authoring tools themselves without any
plugins.
Saif Altalib: For screen readers, an image
with no alt text and no title is generally treated as decorative and skipped.
Right-click the
image.
Select Alt text.
Ensure both the Title and Description fields are completely
empty.
Ned Zimmerman: So, the existing standard
specifically says, like, if the tool is
Laura Staniland: Replying to "I’d
encourage folks to ask for OpenACR rather than...":
Some don’t even have VPATs. Some wanted
them to be confidential.
Ned Zimmerman: not accessible without a
plugin, or
is accessible, but a plugin breaks it.
you have to exclude the plugin from the
evaluation, so I just
Your point, I think, is very well taken there.
Abhinav Kannan: Yeah, the point about
plugins is absolutely right. You should have Tools like Google should have these accessibility
features built in.
Mike Gifford (CivicActions): Replying to
"I’d encourage folks to ask for OpenACR rather than...":
They are worried about
lawsuits.
Abhinav Kannan: Right. But, uh But yeah,
just saying that people I know are using Grackle docs, and for the last few years, I think Google
has taken some steps to add accessibility features into docs.
I think they're also moving
really, really slowly, and trying to add accessibility features into docs.
Laura Staniland: A basic editor should be
easy to find and use. Some of the images were definitely necessary for understanding what to do in
the form.
Abhinav Kannan: And fixing the pipeline so
that The exported Word document or PDF document out of Google Docs has the accessibility
markup.
But there are a lot of workarounds that people are using there.
So, a lot of people
export in order to create an accessible document out of Docs.
They export a Word document out of
docs so that it has the heading level markup, and then They download it, open it in Word, in
Microsoft Word, and then from Microsoft Word, they export the PDF. So that makes the PDF more
accessible.
But yeah, talking to blind people, and talking to a lot of people who use screen
readers, I know for a fact that PDFs are not very well welcomed.
I
Lisa Liskovoi: I will never forget a screen
reader user who told me there is no such thing as an accessible PDF, even when it is accessible,
it's not accessible.
Abhinav Kannan: Yeah.
Lisa Liskovoi: I know there were a couple
of questions in the chat around whether coding tools and vibe coding tools would be considered
authoring tools, and that is in line with the discussions we've had so far. So yes, we would
consider those authoring tools, is my understanding.
Jacob Wood: As a blind user who has worked
very hard to create accessible PDFs, I agree. There really isn't such a thing.
Abhinav
Kannan, Lisa Liskovoi, Ned Zimmerman:👍
Laura Staniland: Should we be maybe
Inviting the people who are doing the Let's see somebody put the LLM thing in the chat.
Is
that a working group for LLMs? Sorry, maybe I should know this.
Charles Hall: I know that there are many
public / published research papers on WCAG. But not aware of any on ATAG.
Like this
one:
Bridging Digital Wellbeing and Accessibility: An Analysis of the W3C Web Content
Accessibility Guidelines
https://dl.acm.org/doi/10.1145/3706599.3719974
Abhinav
Kannan:👍
Lisa Liskovoi: Sorry, Laura, could you
clarify?
Laura Staniland: Okay, so again, I
hadn't been here for a couple months, so if I should know this, I apologize. Somebody put a W3
a link to a W3C document use of large language models and standards work.
Is this written by us,
or was this written by somebody else?
Ned Zimmerman: I can respond to that,
because I put that link in. It was just in response to a question about should we use an LLM to
draft updated ATAG standard?
And in doing a scan of
relevant W3C materials, I just stumbled
upon this document, which I think is a general
W3C document about
the general W3C approach to
the use of LLM tools in standards work.
So it wasn't created by this group.
I think there
was a broader group involved in authoring that note.
but it just seemed relevant to the question,
so I shared it.
Yeah, I hope that answers your question.
Mike Gifford (CivicActions): I think Hidde
de Vries was was connected with that as well, and Eddie has done some stuff with. He created the the
W3C's authoring tool atag authoring tool interface with with Shady, I think, a while back. So
he's got lots of opinions about about, uh.
use of AI for creation of standards. But I
wondered if maybe one way to think about about the use of AI is we've got the is sort of adding
a third section for what are some guidance for AI? Because I think I don't want to blend it in
with all the other tools.
All the other tools are like looking at the Part A and B, I think are
still that's a useful construction and a useful breakdown. And I think AI is a whole other
third component that affects both A and B, but maybe there's a way of.
Saif Altalib: Google Accessibility
Conformance Reports: https://belonging.google/accessibility-conformance-reports/
Ned Zimmerman: One of my colleagues found
this: https://link.springer.com/article/10.1007/s10209-022-00904-9
Mike Gifford (CivicActions): Of not not
derailing us with the magic of AI, because it's, it can easily overwhelm any discussion.
Lisa Liskovoi: Any thoughts on that? Ai is
a separate component or AI integrated throughout
Saif Altalib: I think separate
section
Mike Gifford (CivicActions): Is that
separate, or Sorry, Miriam, you just put a thumbs up and I wasn't sure if that was for separate
or together.
Lisa Liskovoi: I think Miriam showed
separate, yes?
Jacob Wood: I don't really see a
future in which they're distinct.
Ned Zimmerman: Yeah, well, I I think that I
think the thing is that
like, where my head is at with this now is that
There are some things
that
are impossible to disentangle.
But there are some things that are specific to AI,
And
I think my earlier comment about
natural language interfaces as a research area leading towards a
Type 3 is important.
But it is specifically relevant to the
way that AI is used,
like,
ChatGPT and other LLMs use.
use natural language interfaces, and it's not so
relevant to the rest of the existing ATAG authoring.
guidance, so I can see that there's a
section specific to AI,
Within a new version of ATAG?
And then some general things that are
applicable regardless of whether an LLM or other AI tool is involved or not.
If that makes
Lisa Liskovoi: I'll just make this
quick. I wonder if some of that one will have to live in notes, like we'll have to separate AI,
and then in cases where AI is relevant to a particular clause, we may want to capture that
information and notes.
Charles Hall: I agree that we need to
account for AI as part of an authoring process, as an author, and as an agent.
But I would
not agree to the CG using AI to create guidance.
Ned Zimmerman, Laura Staniland:👍
Miriam Fukushima: Yeah, I would also say
like AI is such a big topic that could a separate part would be useful. And also just to have room
to clarify that it can be anything that is.
already there. just as an AI, for example, it can be
a chat interface. It can be a coder, it can be an editor just based on a different technology, even
though editors in itself already exist.
And I think you can we can then also cross-reference to
the different sections. But I think it just needs room for that clarification. And if it's a
third part, then all the other parts come before, and then you can easily.
back reference and say
same if stuff overlaps, same applies, but this is new, and this is what distinguishes from that. So
I think we need that room.
Saif Altalib: Agreed. There will be
multiple types of Gen AI and agentic tools
Lisa Liskovoi: Thanks so much, everybody.
We are at time. Thanks for the great discussion, and I'll see you in a couple of weeks.
