Re: Proposed draft of an AI policy

Hello XPath friends,

Here for your consideration is my attempt at a policy statement.

It keeps Norm's first two sentences verbatim, adding one word. Then it goes
in a different direction.

---- please edit ---

Contributions must not include content generated by large language models
(LLMs) or other probabilistic, non-deterministic tools (including tools
which are often colloquially categorized as "AI").

This policy applies to the specifications, tests, issues, comments, pull
requests, and any other contributions to QT4CG.

In order to help enforce this policy, all PRs will be subject to the
following questions, as criteria for evaluation:

- Who produced the PR and what are their goals in doing so, stated and
otherwise?
- What level of effort is required to evaluate the PR properly? (where
'properly' means to the satisfaction of current CG members and future users
and developers, not some abstraction)
- Who is accountable and answerable if a PR has unwanted effects on CG
activities, productions, or publications?
- Where the PR makes objective claims or external references, have these
been examined and verified by people?
- Where the PR proposes definitions, have these been considered for
consistency with the other definitions given in the family of
specifications, by people qualified to know?
- Where example algorithms are provided as implementation guidance, have
these been tested and verified, by people, to be correct and appropriate?
- Where formal tests are offered, who designed and implemented the testing,
and who else verified the correctness and usefulness of the tests?

Contributors and CG members may note that these questions have always been
applicable; this policy merely states them, and does not exclude other
questions.

--- ende ---

I imagine it could be longer by a few points, but it needs to be short to
be useful. Indeed I think it could also be trimmed - but here I am trying
to get the point across.

Thanks for reading!
Cheers, Wendell




On Sat, May 30, 2026 at 1:48 PM Bethan Tovey-Walsh <bytheway@linguacelta.com>
wrote:

> Patrick,
>
> I would happily wager on the amount of *good quality* accessibility
> generated. Would you like to wager on how often LLMs will be used as an
> excuse to downgrade or outright dump human accessibility work? And how many
> disabled people will be unable to afford LLMs and so will be locked out of
> the supposed accessibility benefits of using them? Would you care to wager
> on how the quality of the "accessibility" generated will change over time,
> as models degenerate when they run out of training data?
>
> And would you like to think about how much accessibility is being
> generated for people who aren't Anglophones, or how the decline in
> accessibility as a human-based task in the Anglosphere will affect
> capacity, training, status, and resourcing for accessibility work in other
> languages and cultural contexts? If disabled, white Anglophones gain some
> low-quality accessibility while it disappears for other disabled people,
> will you notice? Will you claim you won the wager if people like you think
> the accessibility has increased?
>
> Demands have been ongoing for decades, and as you note, only "...meagre
> tools
> are offered to them."
>
>
> And we should keep fighting for that to change. Accepting the slops from
> the AI bucket as a substitute for real accessibility is not a win. Not for
> anyone. It's a distraction; a way to pacify us so we'll stop demanding real
> change. Worse than that, it's a way to persuade the non-disabled that
> accessibility is improving, so that they'll be less inclined to listen to
> us when we ask for more.
>
> As I said, I'm not trying to say that every disabled individual should
> take an absolute stance against LLM use if it's the only way they can gain
> some accessibility.
>
> But I am pleading with this community to stop repeating the lie that there
> is any genuine desire to improve accessibility in what the LLM companies
> are doing, or that their slops are an acceptable substitute for proper,
> human-authored, human-focused accessibility. Just please stop claiming
> disabled people as an excuse to justify the existence of LLMs.
>
> As my physical abilities diminish, I'll take accessibility. (full stop)
>
>
> Feel free to keep thinking only about your own short-term needs. I'll keep
> fighting for the rights of all my fellow disabled people, as I have been
> doing for many years.
>
> Regards,
>
> Bethan
>
> ****************************************************
>
> Dr. Bethan Tovey-Walsh
>
> linguacelta.com
>
> Golygydd | Editor geirfan.cymru
>
> Croeso i chi ysgrifennu ataf yn y Gymraeg
>
> On 30 May 2026, at 16:45, patrick@durusau.net wrote:
>
> Bethan,
>
> On 2026-05-30 11:30, Bethan Tovey-Walsh wrote:
>
> I don't blame individual disabled people for taking advantage of the
>
> meagre tools that are offered to them. But anyone claiming that it is
>
> *a good thing* for LLMs to be used in this way, instead of demanding
>
> proper funding and attention to accessibility work, is displaying an
>
> ableist attitude towards accessibility.
>
>
> Would you care to wager on the amount of accessibility created by LLMs
> over the next five years, versus demanding proper funding and attention
> to accessibility work?
>
> Demands have been ongoing for decades, and as you note, only "...meagre
> tools
> are offered to them."
>
> As my physical abilities diminish, I'll take accessibility. (full stop)
>
> Hope you are having a great weekend!
>
> Patrick
>
>

-- 
...Wendell Piez... ...wendellpiez.com...
...pellucidliterature.org... ...raventracks.org... ...
github.com/wendellpiez...