Risk of Subjective Editorial Judgment and Potential Impacts on Creative Content [b49b2e] #2390
Description
Summary of Objection
While I appreciate the effort to provide more clarity for SC 2.4.6 (Headings and Labels) through the proposed ACT rule, I am concerned that the current draft introduces a testing methodology that is overly subjective and counter-productive. By pinning a heading's "descriptiveness" strictly to the "first perceivable content" that follows it, it risks expanding the scope of accessibility testing into the realm of editorial strategy, where testers lack professional expertise.
Here I share my concerns as to why the examples on the proposed page need to be comprehensively reconsidered before the page becomes the official resource.
Risks Associated with the Proposed Guidelines
The backbone seems to be an invented term
The requirement relies on one statement: "Describing the topic or purpose of the first perceivable content". Some corporate websites may be adopting this trend, but I have found no global recognition for this term outside of roughly 7 WCAG-related sites referencing each other. Establishing a new, unverified term as an authoritative testing standard is a questionable practice to begin with.
Discouraging creativity, encouraging AI-generated stuff
The requirement technically means headings should almost verbatim repeat the following paragraph's content. While this may easily enable automated checks, the potential impact on creativity and originality is significant.
Introduction of this “first perceivable content” term as a rule, may practically make some large websites that have thousands of pages and most of the literature non-compliant with WCAG Level AA.
The proposed failure sentence “this h1 heading element does not describe the first perceivable content after it (the first p element),” applies to the examples below.
• Edgar Allan Poe’s “The Raven”: The <h1> would technically fail WCAG 2.4.6 because the word “Raven” does not appear until the 7th stanza. The "first perceivable content" describes a weary narrator and a tapping at a door, not a bird.
• Ernest Hemingway’s “In Our Time”: Story begins with depicting several drunk people. The first paragraph has nothing to do with "In Our Time”.
• University of Cambridge article. Why human expertise still matters in the age of AI certainty. Several headings in this professional piece do not include anything from the "first perceivable content" that follows.
The popular “users with cognitive disabilities” argument might be resorted to, but I believe using it as a supporting argument for all subjective matters is entering the twilight zone.
The percentage of controversial findings will rise
SC 2.4.6 is fundamentally about human language. Empowering WCAG testers to judge editorial quality invites subjectivity into what should be a technical audit.
Consider these variations for the same proposed example paragraph:
<p>We are open Monday through Friday from 10 to 16</p>
<h1>Business days</h1>
(The paragraph mentions days before hours. Does this fail? If this passes should hours fail? Should the<h1>have included both days and hours?)<h1>Hours</h1>
(One tester may find it “too vague”, another one may not.)<h1>Close on weekends</h1>
(Perhaps it is a supermarket that is expected to be open on 7 days a week and they want to emphasise they are not? Is this "not descriptive"? )<h1>Yes, we are open</h1>
(A marketing "hook" or COVID-era slogan perhaps?)<h1>Wonder when we are open? </h1>
(A conversational style . What if a tester does not like this?)<h1>Oepning Hours</h1>
(There is a typo. Now what?)
I argue that none of these could be failed, as WCAG testers are not editors or copywriters.
Technically speaking, when a heading is simply a repetition of the beginning of the next paragraph, this criterion is satisfied. AI can generate totally useless text that will never fail under this criterion and most of the creative writing will be deemed inaccessible.
Will these pass or fail?
<h1>Opening Hours</h1>
<p>We are open Monday through Friday from 10 to 16</p>
<p>Except that the above paragraph was a joke since we don’t even have a store. Ha-ha-ha!</p>
(Totally misleading. Since this criterion does not require anything to be true as it should not, this will not be flagged. Both the heading and the first <p> are perfect.)
<h1>Opening Hours</h1>
<p>Opening Hours</p>
(No information whatsoever. The web publisher forgot to post the actual hours. But the problem is with the first <p>, the heading is fine. Since this criterion is not about following <p>s, can the heading be failed?)
<h1> Opening opening opening opening opening hours hours hours hours hours </h1>
<p>We are open Monday through Friday from 10 to 16</p>
(The heading looks like this and it is a script-generated error, but everyone will still understand everything immediately.)
But these will fail though there is nothing wrong:
<h1>Opening Hours</h1>
<p><strong>Please be advised that we are closed this week!</strong></p>
<p>We are open Monday through Friday from 10 to 16</p>
(The owner wants to underscore first that they are closed that week, but this fails because the “first perceivable content” does not include hours, though the info the heading refers to is there.)
<h1>Opening Hours</h1>
<p><strong>We will post it once we start working, stay tuned!</strong></p>
(The shop will open the next week. This fails because the “first perceivable content” does not include hours. They did not decide about their hours.)
<h1>Opening Hours</h1>
<p>Temporarily closed.</p>
(They posted this info perhaps on the front page, but the owner wants to repeat also here that they are not open. Fails because the “first perceivable content” does not include hours.)
<h1>How do cats sleep?</h1>
<p>I may be a dog lover, but I will not talk about my dog today .</p>
<p>Cats are interesting animals.</p>
(An article in a blog starts with a joke. Its heading fails because the “first perceivable content” is neither about cats nor about sleeping patterns. The second <p> also fails, as it does not mention the word “sleep”.)
I argue that all of the above examples are editorial matters and none of the above is a 2.4.6 matter. They are all inapplicable.
Scope of testing expands significantly
The entire testing process will take longer.
Any heading from level 1 to 6 is subject to this criterion, right? Will testers start to read all headings and all the first paragraphs that follow them in the entire content?
I argue that doing this is simply impossible.
Granting additional powers to testers
Proposed examples will empower testers with the ability of recording failures to their heart’s content. As a result, there will be more and more controversial findings in reports. Since this criterion is one of the few totally non-technical ones, the website owners will easily understand the discussion and there will be much more pushback. As any controversial finding makes the report itself controversial, testers disliking headings will damage the reputability of the entire WCAG testing process.
I argue that this will result in too many manufactured failures, more no-impact findings, and it will come back to harm us.
New rules assume testers will right away understand whether a heading is relevant
Not all headings are like “Weather” or “Opening Hours”. Sometimes it is almost impossible for a WCAG tester to understand whether a heading really relates even to the entirety of an article.
Consider a hypothetical CERN article that talks about the specifications of Saturn’s rings, that has Saturn’s images, its first perceivable content is about Saturn, but its heading is made up of Neptune’s rings’ specifications. Editorial error.
The heading and the first paragraph are something like this:
<h1>Spectroscopic identification of radiation-darkened organic tholins and sub-μm silicate grains in the low-albedo Adams ring arcs</h1>
<p>We analysed Saturn’s rings in depth.<p>
Is a WCAG tester supposed to find out that the article was actually about Neptune’s rings?
What if the article began like this? This one talks about the specifications of Neptune’s rings, that has Neptune’s images, and its heading is made up of Neptune’s rings’ specifications.
The heading and the first paragraph are something like this:
<h1>Spectroscopic identification of radiation-darkened organic tholins and sub-μm silicate grains in the low-albedo Adams ring arcs</h1>
<p>This article explains how particles exhibit asymmetric albedo variations and distinct wavelength-dependent scattering properties, consistent with microstructure alterations caused by sustained energetic particle interactions and ultraviolet photolysis.</p>
Is it a WCAG tester’s role to decide what heading would be better here? Nothing in the heading is repeated in the “first perceivable content”. Would the web accessibility report recommend “What this article explains” as a better heading?
I argue that this is a new and irrelevant role assigned to testers, ignoring real world headings, based on unrealistic assumptions.
Do all examples on the proposed page really belong to 2.4.6?
Are both passed and failed examples 2 and 3 not about other criteria, such as 1.1.1, 1.3.1, 4.1.2? Are they contributing to 2.4.6?
Not enough cases for testers to extrapolate
There are 14 examples, but are there really 14 different cases? The words “Opening Hours” used in 6 out of 14 examples and “Weather” was used in 4/14. A criterion about wording could have benefited from more than 2 examples of correct and incorrect words that can or cannot be used as headings.
I can understand the idea of keeping the examples as neutral and easy as possible to avoid controversies, but many potential cases are not included.
Basic grammar errors
Or layout issues in such authoritative documents indicate insufficient collaboration and raise various other doubts.
- Screenshot reads “This h1 heading element does not describes the topic of the following paragraph.”
-
Please open the “passed example 3” in a new tab. One may not even see the text because of the unnecessarily giant 980px by 980px png image. That page has no title, does not reflow, and the reason of its existence can also be discussed. w3.org/WAI/content-assets/wcag-act-rules/testcases/b49b2e/14faa79c92b5e281d8694f8a18ec00ba0c11da6b.html
-
An “icon”: opening_hours_icon.png (980×980)
Do people with disabilities make the only impacted group, when a heading is not written as if it is for the search engines of 2000s?
A heading that is not describing the paragraph it precedes “good enough”, impacts everyone the same.
Does WCAG list any one of these as failures?
- Broken links
- Pages that show 404 errors
- Grammar or spelling mistakes
- Links that open incorrect websites
- Terrible photos
- Harmful, malicious, or illegal content
- Websites that are down
- Dummy content
I argue that headings not repeating the immediate content that follows, cannot be web accessibility concerns either.
The scope I propose
I believe heading failures under 2.4.6 for not being descriptive enough should be limited to headings that provide zero info or obviously irrelevant info. Headings should be about the topic, not about the very first sentences they are followed by.
PASS examples I propose:
<h1>Driving Licence Renewal</h1>
(For a page with a driving licence renewal form, good heading.)
<h1>Please fill out the following form</h1>
(For a page with a driving licence renewal form, just bad heading.)
<h6 class=”h1”>Driving Licence Renewal</h6>
<h1 class=”h2”>Please fill-out this form and save on your device</h1>
<form>[form content]</form>
<h4 class=”h2”>Email the form as an attachment to</h4>
[email address]
(For a page with a driving licence renewal form. Page with messed-up semantic heading hierarchy but looks normal. This example underscores 2.4.6 is not 1.3.1)
<h1>Credit Card Application </h1>
(For a page that has one form collecting user data for credit card application. Good heading.)
<h1>Get It Now!</h1>
(For a page that has one form collecting user data for credit card application. Get what? Okay, bad heading. But this is not like a “Find out more!” stand-alone link text that necessitates user interaction, the content is readily available right below it.)
<h1 role=”presentation”>Licence Renewal</h1>
(For a page with a driving licence renewal form. Passes 2.4.6, fails under 1.3.1, F92.)
FAIL examples I propose:
<h1>Home</h1>
(For a page with a driving licence renewal form or for almost all pages.)
<h1>Hi</h1>
(For a page that has one form collecting user data for credit card application or for almost all pages.)
<h1>Hello</h1>
(For a news article that is about the comparison of the performance of 2 stocks or for almost all pages.)
<h1>Fatal error: Uncaught Error: Call to undefined function getUserDetails() in /var/www/html/index.php:15 </h1>
(For any public page that displays a PHP error message mistakenly in the <h1> element for some reason.)
<h1>[Replace this with heading 1]</h1>
(For any public page on which the web publisher forgot to replace the placeholder text.)
INAPPLICABLE examples I propose:
- Pages with no headings at all
- Pages with invalid heading elements, such as
<h10>Driving Licence Renewal</h10>, are to be treated as no headings - Empty headings such as
<h1> </h1> - An
<h4>for one paragraph mistakenly pasted as the heading of another paragraph - 2 identical
<h2>s that follow each other, web publisher made a copy-paste error
There should be at least 10 distinct examples for each PASS, FAIL and INAPPLICABLE cases
Conclusion
This should be a test-to-pass criterion, listing very few cases as concrete failures.
The only failure rule should be about total lack of information
- Headings that hint nothing about anything, due to intentional choice of words
Editorial or coding mistakes should not be flagged as errors
- Incorrectly spelt words
- A clear, but clearly wrong heading that belongs to another article
- Any headings with even terrible errors specified under several other criteria, as long as their text is okay, such as this:
<span class=”h1” class=”h2” lang=”eng-CND” required=”correct” tabindex=”1” onmouseover=”playTaTada();” onmouseout=”keepBlinking(this);” title=” Renouvelez votre permis de conduire” aria-hidden=”true” style=”color:#ddd; background-color:#eee;”>Renew Your Driver’s Licence</span><br /><br /><br /><br /><br /><br /><br /><br />
Irrelevant-looking headings should be ignored
- Creative writing, even when it seems irrelevant to the page
At the very least, there needs to be a lengthy exceptions list for the types of headings that would be considered inapplicable.
Please accept my contribution as constructive feedback.
PS: As a creative writer, I am open to further discussion or contribution.