If we extend this idea, why not do it for airplanes? It seems like a bad idea to be changing the layout of mission-critical systems. I'm sure it'd wreak havoc if the cockpit of a 787 was all touch screens, with UI changes pushed at the whim of developers.

I drive a car with analog controls. The buttons for controlling the radio, hazard lights, the A/C, etc are all static. I wouldn't want it to change. It gets the job done; it doesn't need iteration, and messing with it would reduce safety. I don't have to look at the controls, because the buttons are of various shapes/sizes that I know. Tactile feedback is value in keeping eyes on the road.

An A380 is already full of screens where touch input could be introduced. All major airplane manufacturers are definitely experimenting with touch screens in the cockpit [1]. I guess not for flight critical systems for different reasons, because of missing tactile feedback and limited usage during turbulent flights [2]. But maybe that's just a question of time...

[1] https://www.flightglobal.com/news/articles/airbus-aims-for-a... [2] https://aviation.stackexchange.com/questions/22729/why-are-t...

How about spacecraft with touchscreens: https://www.spacex.com/crew-dragon ?

Compare to the cockpit of Space Shuttle: https://i.redd.it/9mj5bcbnzsx21.jpg

I know from people inside SpaceX that they have covers for the essential switch panels which they put on when Musk is around, because he objects to them for confused silly reasons (like tesla cross-marketing). The astronauts and capsule engineers know better, of course, so this is the compromise.

I hope you're arguing in favour of the space shuttle there. It has no contextual buttons. I'll always argue that contextual controls do not belong in critical systems. Every button on that shuttle will do the same thing every time you press it. A touch screen looks nice but isn't safe. Mission control is telling you to press X button, but X button isn't actually on your display right now?? That's a big problem.

The crew dragon is designed to be completely autonomous. 99% of the time, the astronaut is just a passenger, so there is less requirement for physical controls.

In terms of mission control telling you to press X button, it is almost certain that mission control would have a simulation of the cockpit running that would show them exactly what the astronaut is seeing on their display. So they will always be able to direct the astronaut correctly.

I definitely think a touch screen in this situation can be safer, because it can display information in a way that makes it quicker for the astronaut to understand the situation. It in theory requires less training because there is less requirement to remember the position and function of every single switch.

Space missions need to account for the times when things really hit the fan, that's when design really matters. If shit goes haywire why should I assume my touchscreen is now displaying the same thing as mission controls? If it's not then instructions go out the window. If shit goes haywire in the shuttle, mission control can safely assume they're looking at the EXACT control scheme the pilot is, and can read out controls accordingly. I get that automation is intended, but for emergencies a touch screen really doesn't cut it.

>Every button on that shuttle will do the same thing every time you press it.

I agree with you in spirit, but would disagree with this particular point. Apollo had an erroneous "Abort" signal because a particular switch didn't do what it was supposed to, because it had some interesting failure modes in zero-gravity.

I think is dangerous to assume that one design is flawless over the other; both touch-screens and mechanical systems have their own unique failure modes. Maybe one is more reliable than the other, which I think is currently the case here.

Ok perhaps I could change that to "every working button on the shuttle will do the same thing every time I press it".

Your mention of zero gravity has brought up another consideration for me. Assuming it's a capacitive touchscreen you need to ensure no conductive material ever floats into contact with it by mistake. Switches and buttons have covers and rails to prevent accidental pressing but how do you manage that with touchscreens? Apple manages it on your phone because a false negative is ok in that case, I don't think you can allow that on a spacecraft. How is a floating glove finger differentiated from a glove actually attached to hand?

I don't think the "every working button" buys you much here because is could just as easily be stated as "every working touchscreen does the same thing every time I press it". The point from reliability engineering is that everything works until it doesn't. The difficulty with software systems is that they often have complexity that is tough to understand all the paths let alone test/mitigate them (see Boeing 737 Max as a recent example).

You bring up really good points on the zero gravity considerations. It would be interesting to see SpaceX's FMEA on this system to see what all they've considered.

?? Touchscreen interfaces almost always have contextual displays. So no touching it doesn't always do the same thing every time I press it. The space where a spacebar is on your phone is the same space that is sometimes the camera button. I think contextual design does not suit important interfaces.

I meant it in terms of pressing a working contextual display. I think we're saying the same thing. Software introduces many more failure pathways so it's often not suitable for primary hazard mitigation.

Elon Musk's fascination with touchscreens is absurd and dangerous.

Even in the future world where everything is autonomous, knobs and buttons will always be needed for mission-critical interactions, whether it be driving in dangerous conditions or having to perform emergency maneuvers on your "autonomous" aircraft.

> on your "autonomous" aircraft.

It seems that the FAA (and Garmin) does not agree with you:

* https://www.youtube.com/watch?v=A8o0Nk8E0S4 * https://www.youtube.com/watch?v=OnuRFAAKdfg

I like Dynon's systems a bit more, as they have a few physical buttons to allow one to go to certain top-level / important parts of the menu system in a jiffy:

* https://www.youtube.com/watch?v=xjADTgwCtUU

Just because there is a design moving forward does not mean there is not a considerable amount of discussion/dissent/disagreement on the approach.

This has far more to do with customer expectations than it does with safety. It would be much harder to sell a car with a different kind of steering mechanism than a wheel. The wheel is a leftover design element. Just because it's what we have doesn't mean it's good.

They are.

https://buy.garmin.com/en-US/US/cInTheAir-cAvionics-cMFDs-p1...

Have you seen the F-35 cockpit? It's just one giant touchscreen.

It has a large screen, but it's definitely not "just a giant touchscreen"

Some things are controlled by the helmet movements.

https://qph.fs.quoracdn.net/main-qimg-e6311865f9688fdb18cb69...

And, sensibly, stuff you would use while dogfighting or landing are switches and buttons. That, and things that have to always work like the radio and artificial horizon.

You drive car with pedals and steering wheel. Everything else that is ternary (music, climate, nav) is on touch screen.

So yes, you can have cabin light and temperature controls on touchscreen on an airplane.

Climate is not ternary, it's definitely important for driving in cold climates. To the point that if you can't see ahead, you need to stop on the expressway in order to adjust the airflow to windscreen, so that you don't collide into the other guy who has stopped to adjust the airflow in his car.

Climate can be controlled by steering wheel too.

Of course. The point is that you need to have those dedicated controls with ergonomics that support driving, not touch screen.

Think I have seen flight attendants use a touch screen on the intercom and the cabin lights / infotainment. I’d assume everything on the cockpit is life of death for ~300 people, so different rules apply. Imagine what a sick joke it would be if the recent Boeing crashes were caused by pilots fiddling with a touchscreen while the plane went into a dive. It’s difficult enough to work all the controls accessible without tabbing or scrolling.

> So yes, you can have cabin light and temperature controls on touchscreen on an airplane.

Which means, annoyingly, that you can be locked out of those controls during crew announcements.

> having a flexible interface makes a lot more sense

For business, not the user.

And honestly I expect the answer is simpler: this enables them to subcontract out the UI part and run it in parallel with the development of the rest of the car, because the UI can no longer affect anything else in the car design.

Which allows them to have the UI for controlling the car be done by a team without strong communication links with the team designing the rest of the car's UX.

Is that a genuine benefit, or one that shows up more early in planning and who's downsides show up late?

It's a benefit to the business. There are way more variables of interest to consumers in a car than just UI; few if any will take a crappy touch UI as a dealbreaker (especially if everyone does crappy touch UIs). So it's a pure win for business - the downsides never reach them, they only materialize on the user end, large enough to frustrate people and maybe even cause accidents, but small enough to not cause a purchasing decision change the next time one is in a market for a car.

(I need a term for those cases. Psychological externality?)

Vested disinterest?

That sounds awfully like Boeing's 737-MAX MCAS debacle.

Because who know when the volume control might get iterated out of existence

I prefer to think of cars as cars. A dock for a smartphone takes care of entertainment and for everything else discrete knobs switches and levers. Preferably clicking with a loud click.

I agree on the flexibility part. Interesting thought. Although that doesn't necessarily mean it has to be a touch screen. There are many non-functional buttons in my car right now, which might be wired but just don't perform an action. The car manufacturer could make it more easy to reprogram these.

You can see it a bit as a Playstation controller (or whatever platform you like), it has multiple programmable buttons. A car could basically have the same. A touch screen is not required.

It gets tricky as soon as you need labels and/or more complicated visual feedback attached to "dumb" buttons.

The touchscreen is "lazy" in that it can defer thinking hard about your problem space and I can see why that alone would feel scary in a car.

On the other hand, there is only so much foresight you can have when you are innovating. Developing a Tesla might simply not be feasible without the flexibility a touchscreen gives you.