Myndex
commented
2 years ago Where such divisions are made, it should be noted that future technology may change what can be automated. As an example, you mention al- text — whether or not it's there in the html is an easy, binary, automatable conformance point.
Qualitiative
When it comes to the "quality" of the writing, while there are probably no tools today that are specifically for alt-text, there are today tools that can, for instance, judge and assess the readability of a paragraph for instance:
https://wooorm.com/readability/
The technology in that could conceivably be used to asses readability of alt text... though of course we wold probably want alt text to not just be readable, but to do well describing the contents of the image. As there are technologies now that can examine an image and create a textual description, we can say that a technology that can assess the readability and the appropriateness for a given image is certainly a possibility.
The Font Weight Challenge
In terms of visual contrast, one of the key challenges is that there is no standard for either font weight nor the x-height of a font for any given font property setting.
At the moment, the method is manual, as described here: https://github.com/Myndex/SAPC-APCA/discussions/28#discussioncomment-1610289 But it won't be manual forever as there is technology in development to specifically address this issue.
Not Necessarily Binary
On the above examples, when they become machine determinable, that does not demand the result be binary. Looking for instance at the https://wooorm.com/readability/ page, the text is colored somewhere from a green to a red, depending on how well it fits a given age group.
If the results are determined through automation, I don't think this means they should be forced into a binary result, particularly if WCAG 3 is using a sliding scale scoring.
In response to an AGWG survey on issue to change "text alternatives" to "meaningful text alternatives", I responded:
I ask the Silver group to consider and respond to a tactical approach of dividing requirements between those that are machine-determinable for pass/fail, and those that are qualitative and are less easily expressed as "true" or "false".
There are a variety of ways of addressing qualitative requirements, from design approaches and practices, to adopting solutions in the wild, to user testing (or all three). To the extent that WCAG 3 can find ways of differentiating these requirements using current technologies and practices, it should be easier to establish conformance 'levels' (for example "Passes all machine-testing requirements").