Physical buttons outperform touchscreens in new cars, test finds (2022)

Honestly I hate touch controls just about anywhere. On a smartphone they make sense. But for example I bought a set of BT earbuds with a physical button and it is so much superior (that said, didn't try the Apple earpods yet, maybe that would make me change my mind, they just seem so expensive and I'm not in the iPhone ecosystem anyways). Also just got an M1 MacBook Pro and definitely. don't miss the touch bar!

Touch controls on everything are so 2010s...

The safety thing is what boggles my mind. How on earth did anyone think this is a good idea.

It's actually the biggest reason I won't buy a Tesla.

A surprising number of "no shit shirlock" theories fall apart under real testing as well so it's always better to have real data over intuition.

Sure, but the testing needed to have been done before touchscreens were added to cars, not after, not many years after.

Not only it feels obvious, but it has already been studied many times, with always the same conclusion. But more studies are needed at least until people start caring.

Still, aviation is a domain where they care, and they are starting to put in touchscreens in the cockpit, but doing so reasonably. That is: the touchscreens are not necessary to fly the plane and they mostly comes in addition to physical buttons, not to replace them. It is especially important in turbulant conditions.

Saab always viewed ergonomics as critical to driver safety. In what world is a hazard button 3 menu levels deep a sane choice?

Yep. Not to mention everything now is actively sending constant “metrics” to home base etc.

I’ve physically disabled the onStar on my Volt just so there’s no phoning home at all. But it’s damn near impossible to do that on anything newer.

And people eat the shit up.

This is how i feel about my xterra. I can work all the audio controls without looking. Contrast with my wife's crv: changing the "fader" requires clicking through 3 different touch menus.

> is even in the right mode so the button is there at all.

This is actually really important, and it's something that even physical button-based systems can get wrong. Modal interfaces require careful thought because they change things out from under the user, making them unpredictable. Obviously, not all modal interfaces are bad (nobody would suggest that forward and reverse shouldn't use the same accelerator pedal), but in cars, most of them are terrible.

It's why those old radios with 15 buttons and 7 dials are a cinch to use, but the newer radios that hide everything behind menus are terrible. Even something as simple as bass/mid/treble/balance/fader/volume/tuner are an absolute pain to use when you have to click a button to switch between modes, and those features might as well not exist due to how hard they are to operate. I've tried tweaking them while driving in some newer cars and it's always a terrible experience, but my current 20-year-old car has 7 separate dials for those functions, and I twiddle them all the time without even thinking about it.

> This seems obvious to me, but it's still useful to have such research performed because there's a lot of gadgetphiles out there who can't or won't think clearly about it.

When it comes to UI things that seem obvious sometimes turn out wrong so it is a good idea to do the research.

Two examples, both from research I read of long before everything was on the internet so have no chance of tracking down citations.

First involved a system with a command line interface. It was found that adding delay to some of the faster commands to slow them down made users perceive the overall system as being faster!?

Apparently we aren't all that good at actually telling how long things take without actually measuring them. We instead judge fastness and slowness by how things compare to each other. We seem to notice slowness a lot more than fastness. If most commands are about the same time but a few are noticeably faster then we think of most commands as slow and that bothers us. If all the commands take about the same time, we don't perceive any as fast or slow. It just all seems normal and doesn't bother us.

The second involved menus vs. keyboard shortcuts. Pretty much everyone feels that using a keyboard shortcut is faster than using a mouse to select a command from a menu. But in the system and users they tested they found it was around the same speed. Everyone tested was sure they were much slower with the mouse.

Apparently what was going on here was that to do a command you have to first remember what you have to do to actually invoke the command and then you have to do that. With the menu those two tasks are (1) remember where the menu is you need and where the command is within that menu, and then (2) the moves and clicks to actually get the mouse to that menu item and click it. With the keyboard shortcut those two tasks are (1) remember what key combination invokes the commands, and then (2) pressing those keys.

The second part was faster for the keyboard shortcut, but the first part was faster for the menu (my speculation is that because the menus are at specific locations that you have to move the mouse to we remember them spatially but keyboard shortcuts have to be remembered like a list of words, and humans are really good at the spatial memory stuff).

The subjects still perceived the keyboard as way faster because the first part, recalling what you need to physically do, is mostly subconscious. They only really consciously perceived the part where they are actually doing the mouse stuff or doing the key pressing and since that part was faster for the keyboard they perceived the keyboard as faster.

Note that these results might not generalize beyond the system and users they tested. I remember doing my own test and I found for the commands I tested with I was definitely faster both by the clock and by perception on the keyboard, but I was testing with commands that I used a lot. (I used cut/copy/paste). I can believe the result might have been different with more rarely used commands especially in a system with a large number of commands.

Attached a two axis CNC machine to the Tesla touchscreen, with a pad attached to a solenoid.

Attach physical buttons to various parts of the interior within reach of the driver.

Use some popular hardware + software stack to map the buttons to x,y coordinates on the CNC machine.

Publish to YouTube etc.

Bask in the glory.

"It's not like they're going to suddenly add a new turn direction so need to make that part dynamic."

omg thank you

I rented an early model Prius once, and this was my biggest gripe. So bright it reduced my low light vision.

Had an emergency situation where I had to drive a friend to the hospital. It took like 45 seconds to figure out how to turn the car on and get it in drive... yikes

> The alternative to [x] [is] looking at my phone.

This is the biggest failure I see in a lot of car apps. Spotify and VLC both limit what I can scroll through when scrolling through my libraries on my media screen. Being alphabetical I could scroll through a lot without looking, but they truncate it. All this does is cause me to disconnect my phone, open the app, search, and plug it back in. When the failure mode is significantly more dangerous than the "safety feature" being created, your app is dangerous.

There are so many good intentions that do not consider the failure modes. Yes, there is voice alternatives, but people still don't like using those. I know there are app designers here, PLEASE PLEASE think about how your apps fail! Sometimes the cure is worse than the disease. Good ideas and clever designs can still fail in spectacular ways. You cannot ignore your failure analysis. Dogfood your apps. Make your friends eat it too and be honest. Figure out what they do when things fail. Driving apps aren't your run of the mill smartphone app where you can "move fast and break things," they are apps where failures can lead to people crashing and getting injured or worse. You have to come from an (physical) engineering perspective rather than a software perspective. I'm amazed that even big names frequently get these designs wrong (e.g. the above Android Auto complaint).

If you want to use an app, you should pull over to the side of the road.

The tactile interaction that a physical button gives means I don't have to wait for an opportune time when I can take my eyes off the road. I can operate any function in my 32yo car without ever taking my eyes off the road. Obviously it doesn't have any self driving features, but being able to activate the wipers in less than a seconds because my windscreen is suddenly completely obscured by the car coming the other way that splashed a puddle up, or lower the volume because something has happened that needs my undivided attention, or just adjust the heating/cooling without having to think or look away is always going to be superior.

> Autopilot the technology is already to the point where it's safe in the right conditions to take your eyes off the road for some number of seconds at at time.

How do you determine which seconds are actually safe to take your eyes off the road? If this can't be answered definitively, the answer is you should never have your eyes off the road. And if your answer is, 'I've done it heaps and I was fine', that's simply survival bias.

How do you feel about the lack of a physical latch to open the glove compartment? I just LOVE having to go through a fucking menu tree to accomplish a task that’s been effortless and discoverable for the last N decades.

This will be controversial but...

Saying "this will be controversial" doesn't make it not wrong and merely a reasonable difference of opinion and any criticism dismissable because it was predicted.

> Things that aren't covered they've made automatic: lights, wipers, garage door, seat heaters. Things that aren't buttons and aren't automatic voice handles well (ex. "set climate to 72").

This is my problem with this argument. I want a good alternative for when voice or automated features go wrong (as they so often do). I don't want the fall back to be a shitty touchscreen. I want the fall back to be something that is safe to use when driving.

Yes, but they need to get their EV game on. They're behind.

The easy fix for this would be to regulate that touch screens must be locked while the car is in motion. Advanced functionality which requires a touch screen can be used while the car is stationary.

Not only that, the disabled features can more easily be hidden. Most cars used to have blanks for all the buttons for features that didn’t make it to your model. Now you’ll never know that you don’t have it - or that you really do have it but it’s behind a pay wall.

I always thought it was because a screen is far fewer parts. Easier to source. Easier to get the fit and feel right. It may be surprising to some just how hard it is to get a console of buttons looking and feeling good and not chintzy.

It’s a cost thing. Lowering BOM part count is a huge way to lower service and manufacturing cost.

Came here to comment exactly this. The first time I saw my friend's Model 3 I was thinking "wait, how is this better?"

As proof, I am typing this comment on a Planet Computers Astro Slide.

I think they’re just using and testing the touchscreen UI while the vehicle is lagged parked

and housing in the flood plain is still being sold as water-front property