Meeting Minutes
Wendy Reid: https://lists.w3.org/Archives/Public/public-new-work/2026Jun/0007.html
Wendy Reid: It was announced today that a charter for a proposed a tag working group is in what W3C calls charter refinement.
Basically, the period before the official vote on the charter, where people can give feedback.
The hopes behind charter refinement is if anyone has strong opposition to anything, you can address it during refinement, that versus.
Putting it up for a vote and someone voting against it. So that started now, and I believe it's due to end on July 3rd. Am I right?
Rémi Bétin: Yes, that's the estimated date. Extension may happen, but that's the estimated date in the review notice.
Wendy Reid: Yes. So,
We'll see what happens, I guess.
I figured I'd share the news, it's pretty exciting.
The other thing I did get many people's responses to my question about meeting at.
TPAC.
I'm going to put a request in for us to meet at TPAC.
partly because if everything goes really well, the working group might have started by then. And so that might be the kickoff, partly a kickoff meeting, and we would do like a joint CG working group
session and to bring everyone together. If it doesn't work out by then, then we'll just have a CG meeting and that's totally fine too.
For
those who asked.
about, oh, if I'm not a member of W3C, can I go to TPAC?
I had to refresh my memory, but I ended up looking up the policy from previous years, and essentially the policy is that if a community group is meeting at TPAC, then you are more than welcome to attend TPAC as a full participant in TPAC. So members of community groups that are meeting at TPAC.
essentially get the treatment of any other participant at TPAC, which means not only can you attend the CG meeting that you're there for, but you can also attend other sessions if you would like.
The only and it's not a rule, but it's more of a courtesy thing is if you want to go to other meetings, I highly recommend reaching out to the chairs of those groups whose meetings you'd like to attend.
Partly because they can tell you, oh, this, here's a good time to attend, here are the topics, here's the agenda, because a lot of these groups meet for days, and you may not, you may try and come at a time that's not all that interesting.
Or to you. The other thing, too, is just like so that they know there's people coming. I always appreciate it just so I can make them feel welcome. But it's also just fine to sit in the back of the room and observe. It's a very welcoming environment.
people are used to running into new people there.
But you don't have to be a member of W3C. As long as you're part of a community group that's invited. There's also, room for guests as well, it's not unusual, especially, when you're in a city for people from that city in the web community to occasionally come to essentially visit.
but as a… if we end up getting a meeting slot, then we're all allowed to join as full participants.
So, don't, don't let that deter you.
Charles Hall: You answered a question about TPAC that I was going to ask about the working group, which is what happens to non-members when the community group becomes a working group?
Even invited experts like myself, I'm not a member, I'm an invited expert, so I have to get invited into the working group.
Wendy Reid: Yeah, so…
At least my intent is that we would actually keep both groups. We would have a CG and a working group.
Andrew Puchle: What is involved in being a member?
Wendy Reid: so we wouldn't close one and start up in the favor of starting the other.
this is helpful because while it is completely welcome for people to apply as invited experts.
There's, some circumstances where people shouldn't apply as invited experts. Very rare. It's like it's usually, if you work for an incredibly large company, it's okay, maybe you should be a member. but…
Otherwise, there's going to be cases where maybe it's either you don't feel like you want to be an invited expert.
Totally fine, or if we just have people that are much more casually engaging with the group, which is, again, fine, I want to keep the CG open for that. also just to have, more community engagement, more incubation, especially when we're not sure about an idea, or we want to do more research. the CG can be for that.
Sambhavi Chandrashekar: @Andrew Puchle the company must pay an annual fee.
Sambhavi Chandrashekar: For a company like D2L with an Annual Recurring Revenue (ARR) of $220 million, the fee is USD 25000.
Wendy Reid: And the working group can be really focused on just standardization and the boring stuff.
Mary Ann (MJ) Jawili: I can’t apply to be an invited expert. Company too large!
Charles Hall: So if the CG stays open when the working group starts up again, that means you'll need chairs for each.
Wendy Reid: Yes, you read my mind.
Rémi Bétin: Yes, just a slight detail on what you mentioned about.
Once you are a participant in TPAC, there are kind of two different kind of statuses. for example, if you…
a member… a participant in a working group, like you're, a regular participant, but the meeting may be open to what we call observers. Just to add this name, we call them observers, so there are certain rules.
related to being an observer, but the overall thing is just the same. Some meetings are open for anyone to observe. It's important to realize that you may be a participant or an observer.
But that doesn't change what Wendy said originally.
Wendy Reid: That's why it's
good to check with,
chairs as well, just because there might be, parts of their meeting that they close off to the public, because they're maybe discussing something really sensitive. But otherwise, I as a chair and other chairs that I've spoken to, they're pretty open, but it's always good to check because then they know too that someone's coming and that they can, you know, contextualize or, you know, make people feel welcome.
Andrew Puchle: Thanks!
My company is large, so shucks
Sambhavi Chandrashekar: @Mary Ann (MJ) Jawili Invited experts come on their own merit. Their company does not have to be a member.
Shivaji Kumar: Yeah, so I was wondering, so this is exclusively in person meeting or there is an online component as well?
Wendy Reid: All of the meetings are supported with remote. So there's Zoom links for every single meeting that is held, cameras in all the rooms, mics in all the rooms. So you can join as a remote participant or as an in-person participant.
Shivaji Kumar: Great. Thank you.
Wendy Reid: All right.
Okay,
That means we can get into the meat of today's meeting, which is to talk about. So our work continuing our process of working through the different.
working through ATAG principle by principle to discuss where different ones may need augmentation with AI, or like they might change, not just AI, I should say.
AI or like changes in technology over the last 10 years since the last version of ATAG. So obviously AI is top of our mind, but it's there could be other things that come up because of changes in technology.
Charles Hall: Replying to "I can’t apply to be an invited expert. Company too...":
There is a caveat to that rule.
If you can prove your employer does not have an interest in the outcome of the W3C.
When I first became an invited expert, I worked for an agency. So its clients would have an interest. But not the agency itself.
Wendy Reid: So I'm going to share a link.
Wendy Reid: https://www.w3.org/TR/ATAG/#principle_a3
Wendy Reid: so I think, I was reading over the minutes from last meeting, and I think we got to.
Principal through A3.
which is…
Editing views are operable. And so, starting with 8.831 provide keyboard access to authoring features.
My window's up in a better way for this.
This is one where…
in my read of it, nothing jumped out as needing to change.
the keyboard operability. I think we can all agree that all basically all interfaces should be keyboard operable.
Jutta Treviranus: Yeah, I think one of the discussions we had last week was, and correct me if I'm wrong, is the being able to determine what we cover here that is not covered within other standards.
So what would be the accessibility requirements that should be covered in ATAG? And the rationale for what was covered in ATAG 2.0 is its actions or processes.
that are highly specific to authoring environments, so that you would not encounter them as a consumer of.
it… of content, or as… as part of the user interface, although, of course, the… the question of… I mean, given that there isn't any UAG anymore, or further UAG activity, then that becomes a little bit fuzzy as well. Do we cover the…
The browser user agent components, how far do we stray within, WCAG and the accessibility requirements within that?
Wendy Reid: Yeah, I think something to note too is if you haven't looked at the WCAG 3 draft.
There's definitely some things in there that straight stray towards UAAG and even.
I won't say ATAG, there's some things that I think might overlap a little bit, but nothing yet that's jumped out as me is Oh, no, they're doing their complete overlap.
But this one also struck me as something where, while this is…
written for us is, for the authoring tool user interface, because WCAG 3 is much, you know, is beyond web technologies, or it's trying to be applicable beyond web technologies.
This also might be a principle we don't need anymore because it's reflected in WCAG already.
Or that we mod… we update to say, look at… look at this other document that that's going to give you much better, uh…
Clarity.
But I definitely don't think there's anything in here that has changed in the sense of like new technologies have come in and we need to think about new ways to do keyboard operability
Shivaji Kumar: Yeah, to me, I need to think a bit more about this, but.
To me, it appears that.
It's one of the
fundamentals of accessibility.
no matter where it appears, whether in WCAG or in ATAG.
With that mindset, I think it does merit to stay here.
because not everyone reads WCAG closely, and WCAG is, you know, aimed at a general public, and everyone.
But then someone who is in the, authoring business, and a company in the authoring business.
They would like to, refer to something that is specific to their.
functions and use cases. And therefore, to my mind, it should stay here.
Charles Hall: It occurs to me that we might have to do a little bit of exploring before answering a binary question.
question on this. I do agree that WCAG should cover anything keyboard related.
And there's probably nothing unique about an authoring environment, but some authoring tools, particularly ones that include rich text editing, provide built-in.
virtual keyboards for symbols and emoji and other characters not expressed by the regular keyboard. So that virtual keyboard would have to be accessed by a real physical keyboard.
And I don't know if that distinction would need to be made.
Because the authoring tool in that case may control that virtual keyboard versus the user agent or the operating system.
Wendy Reid: That's a good point.
Though, I guess, that falls back into WCAG land of that's the interface and how it's built should
meet the criteria.
Charles Hall: It should, as long as that virtual keyboard is identified as a user interface control.
then…
WCAG would cover it.
Miriam Fukushima: I was wondering if that falls under A316, present keyboard commands that.
if there are such virtual keyboards that.
It is basically a keyboard command that where the user needs to know or needs to know how to find out that this exists and needs to know what it does, and hence.
It needs to be operable.
Wendy Reid: I think that makes sense.
I guess a question, just to go back to Shivaji's point, because I think I'm just scanning ahead to the other parts in A3.
This probably becomes comes up every time.
Is…
are we interested in…
When things overlap with WCAG, either saying, no, you should look at WCAG or saying.
Here's the A tag specific read on like we agree with these criteria. And in fact, we want to adhere to the same criteria. But here is the.
a tagified.
read of that criteria.
Miriam Fukushima: I think for most, just to simplify ATAG, we should refer to WCAG, but, as well as with, AI and how that complicates the user interface. And, I was also just thinking about maybe.
formulating that the history of changes and every keyboard command regarding to that which comes later as well.
this… these are things that are not necessarily common in WEP, or, maybe overlooked easily when coming to ATAG,
So those things, I feel should have a sentence. And then we can still divide if it's, if it needs to be in ATAG specifically, like some things we already discussed, or if it's enough if, if there are, examples, guidelines, best practices around ATAG that explain that. I think all three.
options are valid.
But the default should be to refer to WCAG just to make it simpler and to not have so many parallel standards floating around.
Mary Ann (MJ) Jawili: Replying to "My company is large, so shucks":
It may be worth applying for IE status anyway. I was discouraged from applying because they thought my company specifically should become a member, and they saw me as leverage, if that makes sense.
Mike Gifford (CivicActions): Redundancy is often a problem with standards.
Wendy Reid, Andrew Puchle:💯
Mike Masey: This about raising the concept also improving the context in which how WCAG is used within ATAG, so saying about virtual keyboards.
Trying to provide that information in a way that is kind of saying things like virtual keyboard, but in an agnostic way in case there's other technology, essentially non-physical keyboards beyond that.
And then providing readers of a tag like information on how best to.
or which WCAG rule adheres to.
implementing, sort of, things like virtual keyboards in a way that's visible and helpful. So, sort of, saying if you're going to include a keyboard.
with stuff beyond the normal width, make sure you're thinking about or including an accessible way to access that, and then providing the relevant WCAG information. because then that, I guess.
Could that more easily grow into taking into account other AI or future things where it's…
You know, if you're going to have a content authoring tool.
That is going to integrate, I'd say, I don't say non-standard, but non-default functionality.
Here is a set of WCAG or other guidelines to follow.
So whether that's adding something to… yeah, whether it's adding something to the A316…
Or different rule or rewording. I don't know.
Wendy Reid: Something your comment just sparked for me was.
Generative UI is floated around quite a bit.
What do we need to say anything? Or is this exactly where these kind of criteria for us make the most sense in explaining, like.
If you try to generate.
a new UI for an authoring tool to suit your needs.
what does it need to be able to do?
regardless of what the original interface does.
you know, what needs to be present, and then I guess, if I'm the creator of an authoring tool, and I'm conscious of the fact that some users may alter my UI. What do I need to do in my implementation of it to ensure that that's even possible?
Or doesn't break horrifically.
Sambhavi Chandrashekar: Does the entire bouquet come back there?
Wendy Reid: There's a
lot of just referring to WCAG.
okay, that's A3.
One,
A, 3, 2, I think, again, is very similar.
This is A.3.2 is provide authors with enough time. So this is the timing related.
requirements, auto-save, timing adjustable.
static input components and content content edits save extended.
Are there any considerations here?
Miriam Fukushima: Yeah. I was wondering since also regarding AI, we have more and more interfaces that also have statuses like waiting on something or, partially generated or half results.
that all these statuses also somehow need to be accessible…
interpretable. And yeah, so maybe we should add that somewhere.
Sambhavi Chandrashekar: Replying to "Redundancy is often a problem with standards.":
As well, lack of harmony across similar standards.
Charles Hall: This is a tangent, but what's lacking is.
Between…
Using a keyboard and having enough time is…
A cognitive memory test?
Right, so 324 is saying content edits are saved like an auto-saving feature.
but…
Accessing those in a way…
Individual saves.
without having to…
Memorize them.
in order to find them is is not covered here. And and an example is like pressing the up arrow in.
IRC.
Gives you a saved history of everything that you've typed in that session.
So it doesn't really apply to enough time, but it's a…
It's a…
a memory or cognitive function test issue to remember individual saves.
Abhinav Kannan: Yeah, this is my question. So based off of what Charles just said, which is, I'm I'm I'm wondering.
based on my very limited experience. I'm wondering what is where in in in what kind of authoring scenarios would time.
Would time be a concern? Session time limits?
Can we… can we discuss maybe a couple of…
Examples where.
This guideline might be most.
Most important, in common authoring tools that we're all using today, so that we get.
a bit more of a.
Sambhavi Chandrashekar: Abhinav, when you're answering as a student, when you're answering a quiz or…
You're using an authoring tool.
The LMS's HTML editor.
Wendy Reid: This also is maybe an area that has evolved over time. Like now versus 10 years ago, because I can remember a time when like your session would time out in something like Google Docs and what you had done had not been saved. Or if you were using an online version of.
like Adobe or something, or WordPress. And if you didn't save the what you were working on.
and you close the tab, or at times like you left it open overnight. Your content would not be there the next day. That happens less and less just because.
Andrew Puchle: With respect to keyboard usage, and the comment about cognitive load, rather than approaching it from a universal design perspective, does it make sense to require or advise keyboard key mapping/customization features?
Abhinav Kannan: Yes.
Auto save has become ubiquitous. Yeah.
Miriam Fukushima: Yeah, but, also, basically any tool that requires you to log in and logs you out for safety reasons should say before it logs you out that, stuff is not saved yet or should be saved or.
will be lost.
Or let's you extend your login.
Wendy Reid: Workday for sure
Shivaji Kumar: Okay.
Okay, I can share one more example, you know, I was…
recently filling out an application form on Massachusetts government website.
It's a very long form. And, uh… and… and each time, it would log me off. if it… if it, stayed open for,
15 minutes without any activity, without giving any time out to me.
And I had to log in again, refill those forms each time each it didn't save any of the content that I had previously filled out. A form that should have taken me maybe 45 minutes or so.
Took me four tries to fill it out. So, you know, just imagine 45 minutes multiplied by four.
Mike Gifford (CivicActions): Both the cloud and for that matter local storage. Browsers have so much more capacity to store content for both online & offline usage.
Andrew Puchle: Replying to "With respect to keyboard usage, and the comment ab...":
In my own research with ND professionals, including engineers and content creators, this was ubiquitous.
Sambhavi Chandrashekar: So, Shivaji, that's a massive failure of WCAG 2.2.1 session timeout.
So I was just wanting to ask how much of it is not covered in 2.2.1?
Mike Masey: I imagine notifications from the authoring tools would also count as time based information. Ensuring these are accessible is super important, especially if there are errors etc
Mike Gifford (CivicActions): https://web.dev/learn/pwa/offline-data
Charles Hall: Same issues with the state of MI unemployment forms. But I don’t consider it authoring.
Wendy Reid:👍
Sambhavi Chandrashekar: It doesn't specify, different…
scenarios, it just says you're working on a system and it times out. It should save, it should allow you to come back where you were.
Although…
Things we take care under a keg.
Shivaji Kumar: Right, right.
We should, but it doesn't work out that way for some reason.
Sambhavi Chandrashekar: No, that's because that company, whoever has created that form, violated point to point one.
Mike Gifford (CivicActions): Why wouldn’t it be authoring?
Sambhavi Chandrashekar: They didn't design it well.
Shivaji Kumar: Right. I think it's a government, and so they go slow, and probably whoever designed it designed it probably 8, 10 years ago, and haven't done any updates.
Sambhavi Chandrashekar: And it did not involve people with disabilities.
Shivaji Kumar: Exactly, yeah.
Abhinav Kannan: Yeah, I'm trying to understand when online forms are considered authoring for ATAG, and when they would not!
Wendy Reid: to get back to, I think…
The spirit of the guideline, but also what I believe Charles was raising with his tangent that is also, I think, something we need to consider. So, a major change also in WCAG 3 that I think we should consider and reflect on is the significant increase of considerations for cognitive accessibility. And so we probably also need to consider, do any of these also need.
Or do we need to add additional things around the cognitive aspects of using authoring tools or interacting with authoring tools?
something that struck me for enough time is we're using enough time, but we could also just kind of broaden this to think about.
time and sessions, because something that struck me with this for, the AI LLM use case is…
A lot of these AI tools have the concept of sessions, concept of memory.
And you have to manage it. You sometimes need, like.
The tools have already have embedded functions for oh, this is your session. You can go back to that session, or you can start a new session.
There's the memory that's built like you can, you can consciously add things to the memory, remove things from the memory.
but sometimes it's not clear when things are being added or when things are not being added, or what gets… if you…
exceed the context window, what's been lost?
from
the memory.
Those are all probably things we have to consider and like identifying sessions and cool like like kind of also like auto saves like knowing like which what did each of my auto saves actually correspond to? Or how do I know that if I arrow up an irc or in terminal, that it's going to show me my previous.
And how far back does that go? Does it go back just the last five actions, or does it go back?
Every action I've ever taken.
some knowledge about that.
It may also just be like not part of enough time, but a new section that we have to consider around session length, or
session identification.
Charles Hall: Replying to "Why wouldn’t it be authoring?":
If I visit a site to fill out a form, it is web content. I did not create the form.
Mike Masey: Slightly on, it feels like it's a similar sort of tangent with it.
Is there anything in there around?
Context, around providing context around what is being authored.
I know it's not quite time based, but when thinking about sort of the cognitive, cognitive side of things, when editing content via whatever authoring tool.
I'm not sure if, unless I've not seen it in ATAG, anything around ensuring that the context around what's being.
authored is clear.
I don't know if that combines a little bit with time-related stuff. Not quite the same, but…
And whether or not it's needed, I don't know.
I guess it's part of perceivable.
Wendy Reid: Hmm.
I was wondering this, too.
Mary Ann (MJ) Jawili: Hey, everyone. yeah, just to kind of,
respond or add to Mike's comments, I think that would be really helpful. I'm thinking about, when…
You're, in an authoring tool, and you can specify the,
Wow.
sorry, my cats. when you can specify, link colors, or, like a…
it would be really helpful to understand, okay, I'm changing this, but what… what are the effects here?
Like.
To provide get more context around like the. The authoring configurations that you're setting. How is that going to be applied or what might that look like in all these across all these different widgets or across all these different pages on your site if that makes sense.
Wendy Reid: Now you have to show us the cats 👀
Abhinav Kannan:😂
Mike Gifford (CivicActions): Replying to "Why wouldn’t it be authoring?":
But you did author the content that got entered into the form. Even if that is just your name and address.
Miriam Fukushima: Yeah, I think in the time-based section, we should add that, first of all.
any, warnings or,
Things like that around the time-based issue should be accessible.
then that, other… that there are other time-based, um.
Things that need to be added, okay, waiting status, or memory, or whatever, anything that delays,
Your ability to get to work further.
and,
Add another point.
I'll come back later, yeah. But those two things I think we should add.
Yeah, I remember that if something changed based on a previous input, for example, control Z for reverting stuff or control Y, or whatever you have it set to.
that sometimes stuff changes also off screen. So you can't see where you changed it. For example, if you have a, you do a step and change 20 places and only three of those are on screen, that there should be, it should be visible where.
Everything changed, and
That you can just scroll through, or, skip through the changes.
Wendy Reid: Yeah, I was Mike's comment, I think was making me think this is maybe something that's difficult to attach to any of the existing sections.
In A3, but is…
I guess, in the family of like you need to be able. So like the the main difference presented now between traditional.
both, you know, traditional authoring tools and, the new wave of AI-enabled, LLM enabled authoring tools.
Is that previously, when you entered content or created content.
It was generally kind of a one-way relationship in the sense that you put things in and things happened. But.
There wasn't there wasn't like it wasn't a dialogue. Whereas now it is creation is a dialogue between the user and the interface, and you're getting feedback in both directions. So you're putting content in and content is coming back at you. That is not necessarily.
the content that is being produced. It's…
The, what it's thinking about, what it's referencing, you know, what it's looking for when it goes to, make you whatever it's making.
and you need to be able to interact with that.
as much as the end result.
Or the content as it's being generated.
So you almost have… essentially have two editing interfaces, because you've got the conversation and you have the canvas or the the where the content is being produced.
And both need to be as interactive.
Miriam.
Miriam Fukushima: Yeah, we had that a bit last time and put part of it to editing views are perceivable and.
accessible?
In A1 and A2?
so the question is, if we put it in there, in how…
Editing interfaces change.
Then the question is, what of it?
Is the time based part for a three that we should.
Charles Hall: To just respond to what you had said, Wendy.
Sounds to me like we should explore.
a second definition of content where there is draft content.
created by the tool and not the human. So the draft content…
Sambhavi Chandrashekar: When I joined D2L, I started creating ATAG reports but stopped from next year as no one asked for it.
Adding this one as a sample of how amateur practitioners might interpret the standards we create:
Brightspace_Learning_Environment_ATAG_2 0_Checklist_December_2018
Abhinav Kannan:💯
Charles Hall: Would then also have…
considerations across all of these.
guidelines.
For example, I would need to be able to understand it if it used language beyond my comprehension.
So if the content didn't come from the human, the content itself is subject to all of these.
Wendy Reid: I was just thinking, Miriam, for the timing aspect, an interesting thing to consider is, so…
I'm talking to…
Cloud code. It's doing things, and it's actively updating something on the side.
But the things that it's doing…
tend to update so fast that it can actually be kind of hard to process what's going on.
Or, intervene even though it does say like they have a slack command. I think it's like flash Btw or something.
I always wonder about, like.
what the timing aspect of that is, in that if I intervene, do I have to intervene in time, or does it not matter? I can intervene after it's done, and it'll… usually you can intervene after it's done, but, is it more effective or less effective to intervene in time?
are there ways to slow it down or speed it up? Do we have to actually say that that's an aspect? It's like you need to be able to set it into a preference mode where I can.
interact with it one action at a time, not…
is that even? Is that a thing we have to consider?
Miriam Fukushima: Yeah, and that's also what I was thinking, the…
there, I think it would be, beneficial if we divide it into, okay, if that… let's take the example of the AI. The AI is generating content, upon a request that I did, and it gives me reference materials and lists links where it looked up stuff.
And,
also a button to okay. Should I also search the web or something? And then the links and the button.
what I would say falls under A1 and A2 in itself.
whereas the, maybe the screen reader, ARIA, polite announcement that the button is there, or that, whatever else it asks me,
would be time-based under A3.
Miriam Fukushima: And I think we should add that.
Evelyn Wightman: What you were saying with being able to.
see its ongoing process and interrupt it. I think in terms of…
The content and like being able to go back and correct after.
that capability is there.
So the interruption is not essential.
But in terms of resource consumption.
It doing stuff has…
money cost associated with it. So if you're not able to interrupt it, it's more expensive to use, which is…
inequitable, so I like the idea of being able to slow it down to something that you can process.
Wendy Reid: I'm just scanning ahead.
Okay, so interestingly, so what we don't.
have really…
in ATAG 2.
is, uh…
Mike Gifford (CivicActions): Replying to "When I joined D2L, I started creating ATAG reports...":
Thanks for sharing those. Getting folks to ask about ATAG is an important first step. It is still widely unknown.
Sambhavi Chandrashekar:🙏
Wendy Reid: error prevention, which is obviously in WCAG. But we almost have to have a slightly enhanced or I guess specific to us version of error prevention, which is and it's not even error prevention. It's like.
Uh…
I'm trying to think of how to word this error prevention in the sense of don't, don't let me accidentally spend a hundred dollars in tokens.
creating this document.
because I…
let a process happen, or you didn't tell me that a process was happening.
and let me intervene before you started the process.
Mike Gifford (CivicActions): Replying to "When I joined D2L, I started creating ATAG reports...":
Even in the accessibility world.
Charles Hall:🫠
Miriam Fukushima: Maybe a more general term would be…
authoring workflow, and then have that like as an or transparency of the authoring workflow. So that we have like an umbrella term where where cost or.
irreversible steps have to be announced beforehand and to have to be able to be stopped. And then also have a transparent history of what you've done.
be it AI, be it the thoughts the AI has, be it even in a static conventional user interface, that it's just transparent also for cognitive.
disabilities that you just have the yeah, the possibility to see what you've done and to revert it, but especially if it's really time consuming, or if it's irreversible, or if it costs are involved subscriptions or whatnot.
that can be prevented beforehand.
Wendy Reid: Mm-hmm.
Yeah, I'm wondering partly that might be a whole new section of its own, or it's.
There's maybe fit with like we have A.3.6, which is manage preference settings, and… and…
Things
like that, though I do like the idea of some kind of, like.
error prevention disaster prevention.
section.
Miriam Fukushima: Yeah. I mean, we can call it workflow transparency and have both in there because I think, preferences, keyboard preferences or.
Mary Ann (MJ) Jawili: Did we talk about the need for version control?
Evelyn Wightman, Abhinav Kannan:👀
Mike Gifford (CivicActions):👍
Miriam Fukushima: key binds, whether it be mouse or keyboard, or whatever other interface is different from actual error prevention, because error prevention is not only the keybinds or the accessibility of those warnings or whatever, but also, the contrast of the warnings, and maybe ARIA to,
make it accessible for screen reader users or whatnot. It's not…
Just preferences. And I think preferences are optional. I like my…
I don't know my preference to… I like to use all my 20 mouse buttons, but,
Error prevention is, I think, non-negotiable.
Wendy Reid: Replying to "Did we talk about the need for version control?":
No but good point
Mary Ann (MJ) Jawili: Okay. I'm just saying this because I put it on the chat and it doesn't look like we talked about it, but I think version control would be really useful. I'm seeing it in more and more authoring tools these days.
so, one, I think there should be version control, two, I think it needs, dates, and if there's multiple people authoring it, we need to have, the names of the users authoring it.
and at the very least, you should be able to revert to different versions, or… okay, at the very least, yeah, I mean, well, whatever, we can talk about what that could look but I think this would be really helpful.
and it's been helpful for other work… in tools where I've…
Where it has existed I've I've definitely used it.
Wendy Reid: Yeah, version control is fine.
Sambhavi Chandrashekar: So history is different from version control, right? We were talking a little bit about history.
But…
Mary Ann (MJ) Jawili: Oh yeah I guess I was jumping the gun there Um right You're right. Great point.
Sambhavi Chandrashekar: No.
No, no, I just want to know, maybe there's a difference between the two.
There's something intentional we need to do for the version control.
Mary Ann (MJ) Jawili: Well, I think version… oh, sorry, go ahead.
Um I think version history is like what I was uh like you can see like maybe all the different versions but I think version control would be being able to like um.
go back to certain versions, or to revert to different versions, is how I see it, but maybe there's other, version… interpretations of that.
Sambhavi Chandrashekar: Yeah, and there could be some organizational policies or policies around ATAG itself as to…
What's a good cadence and what's a good time period?
I don't know.
Miriam Fukushima: Yeah, I can only say that when I talked about history, I meant it in the sense of version control. I'm coming from a programming background, so I thought about that kind of history that you can revert back. And also.
Even if it's just a very minimal editing program, just the basic controls, with copy, paste, and revert, are kind of,
a way to control it. But you should be. It should be made transparent in the way that you see what is going to happen, and if you've done it where it changed. So I think it doesn't matter if it's actually.
a complicated program where you have lots of versions, you see who is working on it, different people working on it, and you have actual versions you go back to, or if it's just a very simple document or text file, where just a simple key control can do something.
I think the… that ATAG should cover both extremes and should say, okay, the, uh…
the editing workflow or the authoring workflow should be as transparent as possible. And that is the first step is to prevent major mistakes and major money hazards. The second step is to if you.
Evelyn Wightman: With token/money consumption where the cost may not be clear at the start of an action, maybe require the ability for the user to configure usage warnings and/or hard stops at whatever thresholds they want.
Wendy Reid:💯
Miriam Fukushima: control your editing. It should be transparent on what happens, what do you delete, what do you change, and if you have changed it, it should be transparent what was changed.
Mary Ann (MJ) Jawili: Yes should have both undo/redo and version control
Miriam Fukushima: like throughout the whole process. And that can be different from tool to tool. And then we can like best practices or something, we can extend on that. But as an extra section, I think we need this transparency of the workflow.
with warnings beforehand, transparency of what's gonna happen, what happened.
Abhinav Kannan: Yeah, I'll make a quick point since we're approaching time. I would think, though, we might… if this discussion on history versus version control is going to go a bit…
If this is going to remain important.
And if that's going to be… if it's going to be longer, then I think we might need to define a line there. I would think that…
Anything that goes into branching or collaboration and committing.
Push, pull, commits, forking, all of that is strictly version control. But, oh.
reverting to prior stages of work would that that is somewhere along the line of history and version control.
So anything beyond that is definitely version controlled. But anything…
Anything before is probably just history.
Mary Ann (MJ) Jawili: Would also like the difference between versions to be explicitly called out. Sometimes difference is not obvious.
Sambhavi Chandrashekar: Would it also correspond to single authoring, single author environments versus collaborative environments?
Abhinav Kannan: It's a good question.
Sambhavi Chandrashekar: Not necessarily, but…
Abhinav Kannan: Yeah, version control in Git slash GitHub would essentially pertain to.
Would would it would essentially involve heavy collaboration for sure.
But in simpler authoring interfaces, I'm not…
entirely sure how how we might define that line.
Wendy Reid: I think there's an interesting aspect here too of.
When we think about history and, and,
And version control, which I do think there's, there's there is a difference. I'm trying to think of like how best to describe it. We may have to come up with definitions for that.
something that struck me was I think now, especially, again, in the AI age, one of the ways that.
at least in academic circles, that I've seen people.
try to verify how something is authored. is this essay, you know, AI generated, or did the student actually write it? One thing they'll look at is tools that essentially.
watch the creation of the document. And so, if all of a sudden you go from an empty page to 5 min later, you know, 10 10 pages of writing versus.
over a course of 12 hours, you know, you see sentence, sentence, paragraph, you know, backspace forward.
the construction of the document. It's okay, I have… I have a feeling I know, you know, you can more clearly assess, was this generated by a human, or was this generated by a computer?
Plenty of ways to get around that, I'm sure. But it is an interesting like idea of like showing the.
history and like generation of a document.
And I do like the idea of separating like single authoring versus multi-authoring scenarios, because there's complexity there too.
Sambhavi Chandrashekar: In education, Wendy, it's a very important thing when we want to preserve productive struggle.
Sambhavi Chandrashekar: for learning, and don't want AI being just a cognitive crutch. There was actually an instance where an instructor said a student finished an assignment in 9 min as a Phd.
I couldn't do it in three hours.
Wendy Reid: Amazing.
All right.
That is.
I
think all we have time for today.
I'm going to try putting some of this together into like a wiki, just so we can keep track of like some of the points we've raised. And we can follow along.