2.4.11: Adversarial Conformance

I made a demo for WCAG 2.2 Success Criterion 2.4.11 Focus Not Obscured (Minimum). I built it strictly to capture screenshots, so while it is not a good demo as a standalone (which is why it is not embedded here I added it) it is handy to demonstrate something.

Conformance

Three of the following screenshot examples pass the SC but only one is useful to users. It comes down to this SC language (emphasis mine):

When a user interface component receives keyboard focus, the component is not entirely hidden due to author-created content.

User interface component is defined by WCAG as a part of the content that is perceived by users as a single control for a distinct function. I argue each button in my demo is a single control.

A group of buttons in a fieldset with the group label “Pod bay doors”; the buttons are completely hidden by a fixed footer, with the focus outline of one barely visible above the footer; the footer has the text “copyright 2001 A Space Odyssey.”
This fails the SC because the button is hidden, even though its focus style is visible just above the fixed footer.
The same group of buttons with only the top row of pixels visible above the fixed footer.
This passes the SC because the topmost row of pixels from each button is visible. Note that there is not enough information for a sighted user to know what the buttons do.
The group of buttons with the very top edges of the letters of the button’s text visible, but not legible.
This also passes the SC. Now we can see each button has text, but because we only see the very top of the letters we have no idea what they say.
A group of buttons with the bottoms of each clipped by the fixed footer, but entirety of the text within is legible; the buttons are “Open,” “Close”, and “Hal?”
This passes the SC and has the benefit of being legible to a sighted user.

Two of the three passing examples are terrible. That last passing example is not much better, however. Throw your favorite 1.4.12 Text Spacing bookmarklet at my 2.4.11 demo and watch all the buttons disappear.

I added the example on 29 December 2023:

Adversarial

Since two of these buttons could result in doom for Dave, it stands to reason that HAL designed this interface — creating a de facto anti-pattern to potentially drive a particular outcome. Or maybe just for the lulz. HAL had an interesting sense of humor.

This is something I call adversarial conformance. HAL the UI designer can claim that it followed WCAG even though it’s clear that HAL expended effort on minimally conforming while negative outcomes are seemingly obvious.

That extra effort is the tell, even if it is not always obvious HAL had to do more work. Heck, sometimes HAL would not recognize it did more work just to avoid implementing what it felt was an ugly design decision.

I have experienced these kinds of efforts over many years of consulting and I figured I would start documenting them.

In the very specific case of HAL the UI designer, its effort is clearly malicious conformance. That is going a step further than I am outlining in this post (since trying to kill the user is generally not a goal), but it is still a genuine concern.

Update: 2 November 2023

WCAG issue #3529 Clarification for Understanding 2.4.11 Focus Not Obscured (Minimum) to ensure consistent testing, filed today, seeks to confirm that the focus indicator is not part of the control.

In July, co-editor of WCAG Alastair Campbell noted that including the focus style in the component size would have messed up the calculation for SC 2.4.13: Focus Appearance (creating a loop), so the focus style was never meant to be included in the definition of the component:

What is perceived as the user interface component or sub-component (to determine enclosure or size) depends on its visual presentation. The visual presentation includes the component’s visible content, border, and component-specific background. It does not include shadow and glow effects outside the component’s content, background, or border.

Success Criterion 2.4.13 Focus Appearance, Note 1

If the focus style is considered part of the control, then my first example would pass and still be useless to most users.

Probably a good idea to watch for the outcome.

Update: 26 December 2023

Semi-transparent content sitting in front of other content is nearly always a hassle for readers, not just because of overlapping text and text contrast issues. But it would likely pass this SC.

A group of buttons completely behind a barely transparent footer, but the entirety of the text within is legible; the buttons are “Open,” “Close”, and “Hal?”
This passes the SC, but the text fails 1.4.3 Contrast (Minimum) and the focus style fails 1.4.11 Non-text Contrast.

Update: 29 December 2023

I embedded the demo.

3 Comments

Reply

Thank you for demonstrating how annoying “sticky” things can be covering stuff. I loathe sticky nav and sticky headers — and the versions that appear when I scroll up a tiny bit because I want to reread something.

Also, Hal the UI designer brought me joy.

Meagan Eller; . Permalink
In response to Meagan Eller. Reply

Thanks. I was trying not to turn HAL into a larger commentary on fake-AI code generation and design automation, but yeah.

Reply

Very much looking forward to you documenting examples of adversarial compliance!

Thea; . Permalink

Leave a Comment or Response

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>