Meeting minutes
Accessibility of Machine Learning and Generative AI draft.
jasonjgw: There has been some helpful discussion on the mailing list. We have used LLM to analyze past meeting minutes.
… This is feeding into work that janina is doing.
<janina> Link to email discussion: https://
janina: Appreciate your extensive comments Scott. Feedback welcome from everyone. Suggest we use this email as the template comment / basis for building the table of contents.
janina: Focusing on this for the next couple of days.
Scott: I thought it was good looking at your comments jasonjgw, different views, but also quite a lot of symmetry. Great exercise in fleshing out the thoughts around it. Good stuff to work with!
janina: Is there anything we should discuss here specifically, Scott?
Scott: There were distinct points from both jasonjgw and I on scope (mostly closely aligned). When it came to general reflections on things, I think there's some nice commentry there to draw on.
jasonjgw: Lots of agreement. Not sure how comprehensive the issue list is, but it seems to cover the things we discussed quite well.
janina: Is there anything missing?
Scott: This will probably become apparent during editing.
jasonjgw: I find it easier to spot missing things from a draft; looking forward to it.
jasonjgw: To summarize: Janina is editing the feedback together. Now's the time to comment. Something early next week.
Observations from ICCHP and IETF conferences.
jasonjgw: The point of this agendum is to consider these recent conferencs that Janina attended with respect to any issues or technological developments that relate to the work of this group in particular.
janina: Very pleased to have had the opportunity to be at both of these events. ICCHP fully focused on accessibility; W3C is well known there. I should have open links to my paper in the near future. Level Access funded non-paywalled access. The paper negotiates 2 things: (1) trying to respond to the overlay fact sheet in the context of the
capabilities CG paper; and (2) the bigger issue that emerged out of it is why is W3C doing nothing about UA accessibility?
… Didn't go into the issues as they are still controversial, though we have a 1.0 REC that's not been withdrawn, but 2.0 was published as a Note at the last minute (though some normative language remains). We haven't been here because it wasn't possible to proceed. The paper argues we should go back and do an update of 2.0 in light of what the
overlay factsheet complains about, and the capabilities report covers.
… At least one person will be joining APA as a result of David Fazio's good recruiting work. There's more that can be said, but I should switch to talking about IETF. I see a need for us to pay attention more than we have been to what IETF is doing that may affect accessibility.
… IETF met the week following the week after ICCHP which happened to be in the hotel I was in! I extended my stay to take advantage of this. I was impressed by the size of IETF; it's much bigger than us, judging by their analog to TPAC. I think we had around 500 in Kobe for TPAC 2025, but there were 1,200 attendees in Vienna, with 500 more
online. They meet in-person 3 times per year. Next is San Francisco in November, I'm considering attending.
… In terms of work that is going on that I ran into... I could talk about what I learned about their procedures, but more important maybe is the work I saw. The one group that relates directly to our accessibility work is the Human Rights group.
… I would like to work with them - offer a presentation about what accessibility is, and be specific about it. Unfortunately none of the chairs were there in-person and they had limited time (1 hour) in the room.
… I attended a couple of other WG meetings. They're not directly talking about web, but they are talking about Internet Protocols that are getting awfully close to User Interfaces in my estimation. Bears watching. Specifically, we're beginning to write about AI and Accessibility. They're starting up some efforts relating to AI.
janina: I'm going to continue to try to follow this, through the Human Rights group and maybe others. Fully open process; one can just join their mailing list. Participation virtually is free. Participation in-person is slightly more than TPAC.
jasonjgw: Thanks for the great overview.
matatk: Two relevant groups - for us, the second is more relevant, raises issues about tools acting on behalf of people, or not (excuse the link formatting; part of another doc...)
* January 2025: IETF WG: IETF AI Preferences WG
* October 2025: IETF WG: IETF Web Bot Auth WG
<Zakim> matatk, you wanted to mention IETF AI-related groups of which I'm aware
jasonjgw: At ICCHP you would've gained an awareness of what researchers are working on - given what you've come across in terms of that work, is there anything that we should be aware of?
janina: I optimized for making contacts over attending all sessions - however I attended ones relating to work we're doing, e.g. Gottfried Zimmerman's group is doing work on maturity and there may be opportunity to combine this and APA's Accessibility Maturity Model.
janina: The session in which Lionel and I presented was full of various work on different web accessibility assessment, and challenges there, auditing related. I didn't see anything new research-wise. Lots of interest in AI of course.
janina: There was a wonderful plenary on considering social context and making room for people to be people and individuals, how that's not done in society, and needs to factor. That was lead by someone in Vienna, I will be looking for a public link. It was really well done. It was great to see people cover these subjects using different language,
and so effectively.
jasonjgw: That's encouraging. When you get an opportunity to review the table of contents for the proceedings and the conference, and find out what becomes publicly available, let us know. Everything wil be published, but access may be variable.
Game accessibility.
jasonjgw: Best way to describe it currently is an enquiry - gap analysis and exploration.
JPaton: Last time we had a question about how to identify the intentional friction from the unintentional friction. There was some discussion about whether the friction the developer intends is always the friction that gamers are using to create challenge.
jasonjgw: Would it be easiest for us to comment on that when you have some early draft text?
JPaton: I have some, in GitHub; discussed this as much as I think I want to at the moment.
JPaton: Would be useful to get feedback, criticsim.
jasonjgw: I thihnk that is a good place for us to be, and thank you for your detailed work on this.
<JPaton> https://
jasonjgw: Looking good!
jasonjgw: Comments/question for John before we go and read what he's prepared so far?
Observations from ICCHP and IETF conferences.
Miscellaneous topics.
jasonjgw: The main one I'm aware of is the draft TPAC 2026 breakout session proposals. I've commented on those. Thanks to Matthew for circulating. Anything anyone wants to add?
https://
janina: I wonder if we, RQTF, want to propose a breakout, either about AI, or the gaming work.
jasonjgw: One of matatk's proposals is AI-related; if that's an appropriate venue we could discsuss the RQTF work there.
<JPaton> 1. Accessibility and Privacy -
<JPaton> <https://
<JPaton> 2. What does good AI training look like from an accessibility perspective?
<JPaton> <https://
jasonjgw: That issue of training is pretty key to several of the issues we've been discussing. Problems caused by the probabilistic nature of the technology. How do we work to address problems ranging from misrepresentation, to technical accuracy?
… Sounds like a good theme to bring people together on in the Web and AI IG, and for the document.
… I had a brief look yesterday and found at least one reference around LLMs and correcting accessibility errors in HTML. Haven't looked in detail yet, but my main concern about published scholarship on this is that inevitably they'll be working with significantly older models than we have now.
… Things are moving so rapidly that research based on models available a year or two ago may no longer be relevant.
jasonjgw: anything else for the benefit of the group?
janina: Looking at calendar ahead, I see no impediment to us meeting every week through September
jasonjgw: I think that's right. If anyone has travel plans coming up, keep us informed. Given what Janina's doing and what John's working on, I think we'll have ample material for a meeting next week with a similar agenda to today.