It's far worse to give users a false mental model of the system. If the underlying system is complex, but that's out of your control, then it will create confusion and frustration in users to paper over it. Please don't do this!

That is, unless you can create a nearly leak-free abstraction. The example given in the article is anything but leak-free.

(This kind of design error reminds me of the skeuomorphism debate nearly a decade ago. Though I do miss beautifully detailed app icons and overall whimsy, in-app false affordances and leaky abstractions both give users an incorrect mental model of the system.)

Make it as simple as possible, but no simpler. Anything else is deceptive.

Then you (and your users) only need the toggle, if at all, for edge cases.

(If the user cares, they likely will have set system settings the way they like them, and if they don’t care to, it’s pretty unlikely your web site is special to them. It would really just be for users who care but are somehow forced to use a system that doesn’t support the settings they want, and for dev/testing.)

But I agree with the conclusion of the article. If you do have a toggle, it makes more sense to just have the two options. Users click it to say, “I want it the other way right now.” After that, it makes a lot more sense to assume they want it to keep being the way they want than not. And how do you know what that is? Because they’ve taken the time to specify that in the system settings. Or they haven’t. Because they didn’t care to. In which case you still aren’t doing something they didn’t want.

A UI where you just whack it until it does what you want (and then keeps doing it) is perfect for we tailless monkeys. One where you have to stop what you’re doing to think about your previous life choices and whether you ought to continue to make them is maybe a little too much for each web site.

Personally I'm totally fine with light/dark/auto but if you really must simplify I see only two true ways:

1. Just offer light and dark. No auto. Whether auto works is out of your control anyway. Users who know and want auto have to switch manually.

2. Just offer auto on/off. Default is off and off is light mode. User who want dark mode will figure out how to configure their system.

Now there's no UX, it just does something reasonable regardless, and if the user has already opted in to color mode management, it respects their declared choice.

Users who are power user enough to want to override their light/dark mode setting per-site can use a browser extension to let them do exactly that on the websites they want to target.

Am I missing something?

unlike everybody else in this comment section, i like the two-mode toggle. but that's just my opinion. without user testing numbers, an article saying "users like this better" is useless.

In auto, none of the buttons is pressed (and preferable a small label that says auto is shown). Pressing dark/light will force that mode, enable that button and remove the auto label.

In dark/light mode, you can press the other button to enable the other state. Or you can press the current button to disable it, and switch to auto. This last one is the confusing-at-first interaction, but makes sense if you think about it (it's like saying: I want dark / I want light / I don't care)

With this, there will suddenly be a bazillion frustrated users who want to force an exception to their timed dark/light mode cycle, but can't because this proposed system hides how to do that. And even if they look up how to do it, they have to either wait for a specific time of day to make their change, or they have to temporarily change their system settings in order to change the program/website-specific setting.

Meanwhile, people who don't care or don't understand will just leave it on the default and probably not even look at the settings page. And everyone is used to seeing "system default" as an option at this point, anyway.

Net effect: no benefit, and some users now upset.

Why change what isn't broken?

(I don't get the "everything needs to support dark theme" thing either tbh)

> “Whatever the system says” is not just one extra option. It’s also one extra weird option. It doesn’t feel like the other two.

It's wrong to conflate the `prefer-color-scheme` media query with “What the system says”, meaning the OS-level dark/light option. For most people this is probably true, but I prefer sites that work this way because I use two Firefox extensions to set it globally or on a per-site basis regardless of what my OS is set to.

- https://addons.mozilla.org/en-US/firefox/addon/toggle-dark-m...

- https://addons.mozilla.org/en-US/firefox/addon/darkreader/

Do not hide what the machine is doing from the user in this way, this is how so many of the bad things happen.

This is not a novelty but a built-in feature called “Auto” Appearance in MacOS, iOS, iPadOS. Similar feature exists in Android and KDE.

"For example, a keyboard could include a special-purpose power management button (for example, “Power”) that, when pressed during a system sleeping state, generates a wake event. Alternatively, the button(s) on mice and other pointing devices could be used to trigger a wake event. "

That's a CSS @media not

dark -> dark

light -> light

os preference -> os preference

into

dark -> impossible to know with just this info

light -> impossible to know with just this info

os preference -> impossible to set via the app

if you absolutely have to account for the third case, then you need an explicit option for it and should be able to tell the user what their current os preference is. all the information i need to make a decision should be present in the interface, and the underlying code should do what the fuck i tell it to, not what it thinks i meant.

For this scenario to unfold the following has to happen:

1. OS is set to auto mode 2. It's late afternoon 3. User finds they have (for some reason) in the past turned on the Website dark mode (presumably in the daytime) 4. User wants it light now.

This is a pretty tortured hypothetical though. I admit as an engineer, who understands that there's a light/dark signal being passed down, and there's an ability to override that signal, I appreciate separating the "should override" control the way a "3-way" switch allows.

But I have to admit, I cannot imagine a real person getting bit in any way by the implementation the author describes. 90% of people set their OS to the "mode" they want, whether it's full-time or sun-based, and that group never even needs to touch or look at this control (We do default to auto, right?). The other 10% are most likely here because they don't understand they can set this in the OS and they apparently want to toggle to the opposite mode. They might be thinking "I want this dark mode until morning" but "we" in the website actually can't really help them with that unless we start prying into what time it is there and build our own time-based mode. So the best we can do for that user is to shift them into "latched dark mode" and let them come back (as they've demonstrated a knowledge to do) when and if they want the 'lights' back on.

I'm the first to criticize UI designers who rob users of choice in the name of minimalism -- the Alan Dye approach, but this actually screams "Designer thought hard about the actual intent of the users" to me and I like it.