Re: Scope of AI/ML and accessibility work

Thank you, Scott, for your comments.

Is there a good definition of "generative AI" anywhere? It's clearly a 
subset of machine learning applications, whereas we are interested in a 
larger set of applications and use cases, not just this smaller 
category. For example, an evaluation tool that uses machine learning to 
compute a score wouldn't be "generative", but it would be useful in 
accessibility assessments.

I don't object to using the term "generative AI" in announcements along 
with "machine learning", provided we don't use it to define or confine 
our scope of work.

On 31/7/24 20:49, Scott Hollier wrote:
>
> To Jason
>
> Agree with comments and scope, the only thing I think has been 
> overlooked is the need to specifically reference the term generative 
> AI. Machine learning has been around for a long time, the part of it 
> that has drawn the attention of the world to this work in recent times 
> has been generative AI and based on my research, there’s less 
> familiarity outside of academic circles about machine learning as a 
> term but a lot of familiarity in the public with the term of 
> generative AI, so I think important that where applicable this is 
> referenced as taking a straw poll of my 13 staff and a few dozen 
> people in the meetup groups in the accessibility community here, few 
> are aware of machine learning but all were familiar with generative AI 
> and that’s how the discussion is framed so just want to make sure our 
> work isn’t overlooked in our messaging not specifically including this 
> term.
>
> Other than that, a great summary of the work, scope and approach.
>
> Scott.
>
> Dr Scott Hollier
> Chief Executive Officer
>
> Centre for Accessibility Australia logo 
> <https://www.accessibility.org.au/>
>
> Centre For Accessibility Australia Ltd.
> Phone: +61 (0)430 351 909
> Email: scott.hollier@accessibility.org.au 
> <mailto:scott.hollier@accessibility.org.au>
> Address: Suite 5, Belmont Hub, 213 Wright Street, Cloverdale WA 6105
>
> accessibility.org.au <https://www.accessibility.org.au/>
>
> *Subscribe to our newsletter* <http://eepurl.com/drA-ib>*__*
>
> *__*
>
> X icon <https://twitter.com/centrefora11y>Instagram icon 
> <https://www.instagram.com/centreforaccessibility/>Facebook icon 
> <https://www.facebook.com/centrefora11y/>LinkedIn icon 
> <https://www.linkedin.com/company/centreforaccessibility/>
>
> /CFA Australia respectfully acknowledges the Traditional Owners of 
> Country across Australia and pay our respects to Elders past and 
> present. /
>
> *From:*Jason J.G. White <jason@jasonjgw.com>
> *Sent:* Wednesday, July 31, 2024 10:07 PM
> *To:* RQTF <public-rqtf@w3.org>
> *Subject:* Scope of AI/ML and accessibility work
>
> Based on Task Force discussions, it is my understanding that our 
> document on AI/ML and accessibility is intended to address the 
> application of machine learning models to enhancing the accessibility 
> of Web content, as well as the accessibility of Web content which may 
> be created by this technology (e.g., AI-generated text, audio or video).
>
> This scope includes applications of machine learning used in
>
>   * Content creation and distribution (e.g., machine learning invoked
>     by the author or Web site operator, including any third parties
>     they engage). Authoring tools, code generation applications, and
>     accessibility evaluation tools are all included in this category.
>   * Interaction with Web content by the end user, including machine
>     learning technologies invoked by the user (possibly in an
>     assistive technology or as part of a user agent). Machine learning
>     applications that the user engages to enhance the accessibility of
>     content are included. In addition, machine learning models invoked
>     by the user as part of a Web application lie within our scope. For
>     example, a text or image generation application built on machine
>     learning technology that is available to the user as part of a Web
>     application would be relevant to our work.
>
> The purposes of our document include the identification of issues of 
> accessibility to be considered in the development of these 
> applications, and identification of currently applicable W3C 
> standards, as well as other suggested practices or approaches that we 
> choose to discuss.
>
> The use of machine learning to make decisions about the rights, 
> privileges or responsibilities of individuals (including people with 
> disabilities), and the questions of discrimination that these 
> applications pose, lie outside the scope of our work, even if such 
> applications happen to be available via the Web. For example, an 
> application designed to assess recidivism risk in the criminal justice 
> context raises important policy questions, but it is outside our scope.
>
> Comments are welcome, as Janina plans to integrate these 
> considerations into forthcoming draft text.
>