Hacker News new | past | comments | ask | show | jobs | submit login
The Elevator Button Problem (jgc.org)
61 points by jgrahamc on June 22, 2010 | hide | past | favorite | 63 comments

The postscript mentions a paternoster, which I'd never heard of and is absolutely fascinating. It looks incredibly dangerous.

Wikipedia: http://en.wikipedia.org/wiki/Paternoster

Example of one: http://www.youtube.com/watch?v=OXSnNzGJDdg

The one in Oxford was brilliant to ride. To the hacker mind it's a great optimization.

And yes, I did ignore the warning signs and make the trip around the bottom.

What happens if, say, you slip and bonk your head on the rail, and end up laying there with your leg sticking out the door?

A couple years ago I was in a building in Vienna that had one... it was brilliant!

It was a little scary stepping on for the first time (probably like stepping on an escalator for the first time when I was a kid, not that I remember) but it's a really unique system.

Vienna, Austria? (not the one in Ontario or the ~dozen in US) Do you remember what building that was? I'd love to check it out.

Another option to the lift UI conundrum that I've encountered in Vienna is to have dedicated staff operating it. It's extremely rare these days, one example I can think of is the Staatsoper (State Opera). You still need a button to call it, but the operator decides what floors to visit in what order. Cost efficient? Probably not, but considering how hard it is to get tickets to the Opera (by subscription or depositing €250 to be entered into the ticket lottery) I guess they can afford to keep up the tradition.

EDIT: I'd love to know why I'm being downvoted.

Definitely Vienna, Austria. It was at the "Haus der Industrie" on Schwarzenbergplatz 4.

Thanks, I'll have to take a look. :)

Does anyone know of any paternosters in the U.S. or Canada? My new life goal is to ride one, and it'd be swell if I didn't have to go far to do it...

There's one in the arts tower at Sheffield university, it's a 1960s building and quite ugly but it has a paternoster (http://en.wikipedia.org/wiki/Arts_Tower). I remember going all the way around on it with my friend Mary when we should have been in lectures. A very fun ride!

The one at Leicester Uni was cool, kind of felt dangerous, have to be careful when you get on.

Neat. Before I read the comments and realized this actually existed I was thinking about something like this; a physical "token ring"!

That's not how token ring works. In a token ring a single token circulates and is captured by a node wishing to transmit. That node then transmits its frame. When the transmitted frame is received back by the transmitter (it has gone full circle around the ring), the token is released and continues circulating.

Sweet! There are several in Vienna. I'm soooo going to check one out.

Vienna has a lot of ancient and/or newer-but-micro elevators. Quite often going to the doctor or meeting somebody in their apartment, we end up in an old all-wood elevator with a bench, and no inner door. Typically they have wrought iron or brass external doors, and sometimes carpeting.

Some newer ones from the 70s and 80s even have no inner doors - but some kind of (possibly infrared) sensor so, if you stick your hand in the space, it will grind to a stop.


I'm fond of re-telling the following story, which I first heard when I was an undergrad 30+ years ago. (I'd love to be able to cite a reference. Is this familiar to anyone?)

A brand new office tower was unveiled, the architect was on hand, the tenants moved in. Everyone was enthralled with the new digs: spacious offices, sensible layout, all the modern conveniences. But everybody - I mean everybody - complained about the elevators. They were so slowwwwwwww.

The building owner consulted with the architect, who was plainly amazed. He hadn't scrimped on that aspect of the project either, and pointed out that the elevators were as fast as anything you could buy in the marketplace. They'd done modeling studies, and the simulations showed that the elevators could keep up with peak demand. There couldn't be anything wrong with the elevator system.

Yet people complained loudly and uniformly. The elevators were way... too... slow!

It took a human factors guy to figure it out. The first floor elevators had an indicator over the door that told which floor the elevator was currently servicing, but ONLY the first floor had such indicators. The remaining floors only had 'up' and 'down' buttons, and nothing more. "Put indicators above the elevator doors on EVERY floor", the human factors guy said. "People don't mind waiting, as long as they know where they stand."

The building owner installed indicators and that fixed the problem.

I've heard of mirrors being installed to accomplish the same thing. People don't pay attention to the speed of the elevators because they're too busy primping.

I would like to know more about the person who thought they should 'steer' the elevator by pressing the button which could lead it to the person's floor for pickup, and the specifics of the lobby where they were.

How old were they?

Why had they never observed, or been instructed by adults about, typical elevator operation?

Had they never seen an elevator, but had played video games? Was the display of the elevator's current floor close enough to the buttons, or perhaps stylized enough, to suggest it could or should be directly controlled like a video game character?

(Perhaps JGC was performing an experiment with a young child or recent immigrant?)

I think the traditional design was probably optimal at the time it was designed. As others have noted, a single call button adds unnecessary stops -- and at the worst times, when the elevator and lobby are most busy.

Many times there is no indicator of where a elevator car is. Or, in different buildings, variable numbers of cars, unrelated to the standard up/down buttons. So anyone with experience with a few different elevators will find interpreting from their location, rather than the elevator cars, the more consistent approach.

Apparently the state-of-the-art is to have buttons for every floor to be outside the car, in the lobby. Each person declares their destination with a press; as cars arrive they display or announce where they'll stop so the right people can board. But elevators are an old technology; so many buttons and such optimal grouping/scheduling were impractical when the current conventions arose.

Late 30s. Very educated British woman. Didn't ask why they didn't know this.

Wow. With those demographics, I might have wondered if some sort of medically-triggered confusion had beset her. I would have expected such a person to have ridden elevators hundreds of times.

Though I suppose if spent most of your life in a rural area, and only rode elevators occasionally with strangers, you might continue with such an initial misunderstanding indefinitely.

I share your confusion, but I think someone should have a rough idea of how an elevator works even after a few uses.

That said..

I would have expected such a person to have ridden elevators hundreds of times.

If she didn't travel much internationally, it's very easy to avoid or not come across elevators in the United Kingdom especially outside of London. I use an elevator perhaps a couple of times a month, nearly always in multistory car parks (which women, in particular, are often not keen to use in the UK, thanks to scare stories back in the 80s).

“I would have expected such a person to have ridden elevators hundreds of times.“

If you don’t know which button you should press is it likely that you will figure it out on your own? Elevator behavior always seems very random to me, to the point where I think that it doesn’t really matter which button I press. I suppose that in the end, pressing the right button is more effective, but probably unnoticeably so.

My reasoning is: the first time most people will encounter an elevator is as a child, with a parent or teacher, and the proper use will have been explained then.

Failing that, they will often encounter elevators with other people present -- and observe what they tend to do. Hopefully, as either a child or adult, the novelty to the experience will cause them to pay close attention to figure out a new thing.

Failing that, they will reason it out or experiment on their own. They are aided in the 'right' conclusion by several factors. First, many elevators don't show where the car or cars are, so there's no chance you'd have to guide them to your floor -- you don't have the information to do so. Requesting a desired direction from your current position is the only sensible interpretation. If you encounter a mix of elevators, some with car indicators and some without, but all with the exact same up/down buttons, you would have to strain to believe that you must consult an indicator that is not always present before choosing the right button.

Second, at least in buildings (as opposed to car parks), most elevator trips begin on a ground floor that may only have an up button. No matter how vivid the car-level indicator, or how closely positioned to the button, it's hard to interpret that single up button as a command to move the elevator up from its current level (as opposed to calling it for an up trip).

Finally, if you have the 'wrong' model in your head, and encounter elevators more than a handful of times, there will be a lot of hints that your presses have been counterproductive. Empty and full elevators going the wrong direction will stop on your floor, and then the call-light on the button you pressed will go off. Only sometimes (when there's not a pending call past you) will pressing the alternate button then reverse the elevator.

Altogether, that's why I think this person must have had only a handful of experiences with elevators, or some very uniquely impaired (and possibly even recently damaged) way of learning about the world. Perhaps it's a dyslexia-like difficulty understanding causal sequences of events or modeling non-human systems. (I'd hope it was offset with talents in other areas, such as artistry or empathy.)

>> Elevator behavior always seems very random to me, to the point where I think that it doesn’t really matter which button I press.

And that's with a relatively naive system running the show. Imagine waht elevators will be like when the Sirius Cybernetics Corporation finally deploys their Happy Vertical People Transporter...

Apparently the state-of-the-art is to have buttons for every floor to be outside the car, in the lobby. Each person declares their destination with a press; as cars arrive they display or announce where they'll stop so the right people can board. But elevators are an old technology; so many buttons and such optimal grouping/scheduling were impractical when the current conventions arose.

I've been in an office building in New York that uses this system, it's supposedly very efficient in both time and energy, as it should be. BUT, they have had to station an attendant in the lobby to basically operate the system for visitors, since it's just too obtuse for the average elevator passenger.

"Apparently the state-of-the-art is to have buttons for every floor to be outside the car, in the lobby." - Not sure if this is state-of-the-art but i used such a system in Singapore 10-12 years ago and it works very well. The only flaw is the lone-ranger problem who has to go to a floor no one else has to. In peak hours, such a person may/will have to wait for a much longer-than-usual amount of time.

Agreed. There are many ways to improve an elevator in objective and perceptual ways... and what the author proposes is definitely not one of them.

This is one area where the Japanese really have us beat. In a Japanese elevator lobby, the moment that you press the call button, you hear a gong and a light comes on next to the elevator that will be picking you up. Thus, the system instantly acknowledges your button press and provides you with information about what to do next. Very occasionally, if the system is very busy, the gong will sound again and a different light will come on (for example, if your designated car gets tied up with a slow stop), but that doesn't happen often.

Plenty of bigger or high traffic American buildings have this, too. Examples I've seen: Bear Stearns' old building (well, the one they were in when they were no more), the NY Times new building.

In the NYT one, you choose your floor before you get on, it specifies your elevator, and then... there are no buttons inside.

It's kind of horrifying.

The Bear Stearns one still had buttons inside but you were definitely directed to a specific elevator of about 8 available. The building was 43 floors and an uninterrupted elevator ride to the 42nd, where I worked, from the ground floor took about 3 minutes. It was so fast it never failed to make me motion sick, and and my ears popped.

Sounds like a case of 'and now you have two problems'. The proposed solution addresses a problem that doesn't practically exist (few people, once they learn the meaning of the buttons, are confused about their function) by making elevators more inefficient by removing the directional call. Taking away the elevator position display also removes the elevator's progress indicator so on top of elevators now being slower for everybody, you can't estimate how long you might have to wait for one. Seems more like UI pessimization.

Indeed the proposal is not UI simplification since you reduce the action of pressing one of two buttons with the action of pressing one button and then read direction for each elevator which stops after your pressed the button.

Having people make choices based on information they have to read should be avoided when possible.

point #3 isnt exactly correct. On most Otis elevators the direction you call the lift with mandates an operation. If opposing sequence of requests occur, the lift doors will close and then re-open. You'll notice the directional light above will switch upon the doors reopening. The only way to prevent this is to hit the manual reset (commonly pressing two 'secret' buttons at the same time).

When working in hotels, I use to request an elevator going down with passengers, jump inside, perform a manual reset and force everyone to ride up with me.

When working in hotels, I use to request an elevator going down with passengers, jump inside, perform a manual reset and force everyone to ride up with me.

Ah, this is a system I can get behind! I get to go where I like and screw everyone else - the perfect interface. (I'm not even being sarcastic, I have an evil streak like that ;-))

Your evil streak is pathetically mundane, Peter! Aim higher!

FWIW, to do this you press 'door close' + 'desired floor'. this can also bypass other requests (not just change direction).

I really want to try this some day but I don't know if I could be that rude. :(

Wow. I never thought about it. The UI is wrong. You need to tell the elevator where you want to go. The elevator needs to tell you when to get on and off and maybe how long you have to wait. That is all.

The list of floors should be on the outside. Selecting the floor should act as the call. A display beside the button should count down until the elevator going to that floor arrives. (And in multi-person, multi-elevator situations, perhaps a color or number should tell you which door it will be).

The existing UI expects you to understand the elevator algorithm in order to operate it. That's wrong. It's unnecessary mental work for the users and locks you into one algorithm. What if you want to optimally handle this case?

  5: Guy going to floor 4
  2: Guy going to floor 7
  1: Elevator

Ah, but if you had all those buttons, which would you push? Where you are, or where you're going? The same logic which causes some people to push down to go up would cause some people to push the floor they're on, because they want the elevator to come to them.

I, personally, would love to have that sort of display. But the added cost of the buttons & lights & damages / abuses probably wouldn't be worth it. (anyone, anywhere, can cause the elevator to visit every floor. It's almost as bad as putting control of the elevator onto a public website)

Just have it so that a floor wouldn't have a button for itself. Why would you need it anyway?

And the same person will stare at the panel until the button for the current floor appears. I've seen this behavior in humans before. It's not a pleasant experience.

If a person cannot understand that a button will not spontaneously form on a physical panel, cannot understand the basic laws of reality and physics, then they shouldn't be let outside or out of the care of others. There is nothing you can design that would work for someone with that kind of mental disability.

Aside from the "not letting out" part, that's essentially the point.

The current system works wonderfully for 99.9% of people. Redesigning everything for the last 0.1% isn't worth it, since they can either be simply taught by a passer-by or they're incapable of learning it.

I dare say everyone is guilty of at least one massive oversight like this. If everyone guilty of one were quarantined, there'd be no-one left.

One way around this is to have a label for the current floor, but no button next to it, just metal plating. Then it's obvious that there is no way to select the current floor.

The miconic 10 system does this (http://www.schindler.com/group_index/group_kg_tech/group_tec...).

I have it in my building and although it is intuitive to use, sometimes I find myself waiting far longer than I think i should be for normal 'up down' button systems.

NYT article describes it here: http://www.nytimes.com/2006/12/10/magazine/10section3b.t-1.h...

My building has it too. It is sometimes confusing, such as when you're in a pre-coffee daze and step into an elevator without punching in your floor. Because there are no floor buttons inside, you end up riding to wherever you can get off and use the elevator correctly. That said, it works as it's designed to. It's a rare event to share a car with people going to other floors.

"Design an elevator" also makes an interesting interview question.

One of the best designs that actually saw was an elevator without calling buttons. To go somewhere you pressed the desired floor on the panel _outside_ the cabins. The cabins only had open/close/etc buttons and you were notified on the panel which cabin to take.

I've always wondered why elevators don't have a "cancel" button. Many, many times, I've have pressed the wrong floor button forcing me to stop and then continue to the right floor. Not mention the smart ass kid that presses every button just before I enter the elevator ;)

If they offered a 'cancel' button, the kid can also hit the 'cancel' button on his way out and destroy all the other floor choices people already picked.

I've seen elevators where if you quickly press a button couple of times, the button will turn off (if it were already on). This of course doesn't protect you from the mischievous kid who presses every button possible.

That's easy to fix, though, they just press their floor number again. If all the buttons are pressed there is nothing they can do (unless they know the secret reset password).

Another nice example is the <intuitive> steering wheel of a car.

If you know cars, you will see: Hey. I need to turn the wheel to the left, so the front wheels of the car turn towards the left, so the entire car begins to turn to the left.

If you are a sailor, however, who only knows simple hand-steered rudders, your intuition is the opposite, because in order to steer your ship to the right, you need to turn the rudder to the left in order to get the right currents flowing.

These problems are _really_ interesting, and _really_ hard to solve. (whereas I think, the first results from the second). :)

"One interface optimization would be to replace the up and down with a single call button. Passing elevators would stop and indicate which direction they were traveling."

You might want to ponder that one a second longer.

I thought the same thing, but remembered the elevator in my grandmother's building: since there are only 6 floors without much traffic, once you are in the elevator, it's all yours for your travel to another floor. The elevator can be "busy" and you can call it to your floor only when the current travel is over. You get in, you press the floor you're going to, and only then does it take new requests.

So, one button works in that particular configuration. It's not absolutely efficient, but the traffic is low enough that it's not, in practice, very inefficient.

This is actually could be a decent optimization if there is only 1 elevator shaft and not a lot of floors- something that bugged me about my old apartment.

In my 3 floor apartment, if you're on the 2nd floor with the down button pressed and someone gets out on your floor while the elevator is heading up, you could have gotten in while the car is stopped and avoided the stop on the way down. However, the call wouldn't cancel until the elevator was actually on the way down and it would stop a 2nd time.

In my building you push the single call button, get on and press the floor you want to go to and let the elevator do its business (it travels to the lowest floor called then up or vice versa depending on which direction it was just travelling). It's either that or take 16 stories of stairs.

Why? The elevators in my college dormitory worked precisely like that. If the door opened but the "up" light was on and you wanted to go down, you'd just stare at the people in the elevator and wait for it to continue its way up.

You answered your own question. Why would you want to stop the elevator for people who have no intention of getting on?

I have always wondered what sort of optimization, if any, is done to make buildings with multiple elevators handle button presses most efficiently.

There are optimizations in most modern group dispatch systems to attempt optimal traffic routing and predictive placement of idle elevators. The fancy stuff OTIS has played with gets into AI, but the simple algorithms work nearly as well.

It's the unpredictability of the system (people holding doors, elevators going out of service, faulty mechanical switches, etc) that makes the algorithm design fun.

Feel free to ask me anything. I used to design elevator group dispatch systems as a side job in college.

The New Yorker had a fascinating article about elevator design a while back:


a call button does not give the system adequate information to take the necessary action; If you press it and you get a lift that's heading down and you want to go up, you just wasted time.

One possible answer would be to simply not give an indication of where the elevator is, this would prevent instances of the user thinking that they are telling the elevator where to go from it's current position rather than telling then system where they want to go, but this seems like an annoyance to the other 99.9% of users that actually know how a traditional elevator UI works.

But hey, a lot of modern UI design held up as an example of mastery has that exact same drawback, but pandering to the ignorant is somehow more important than not screwing over the adept. I'm looking at you, OS X.

> a call button does not give the system adequate information to take the necessary action; If you press it and you get a lift that's heading down and you want to go up, you just wasted time.

In my building there's only one passenger elevator and the service elevator, you're not wasting time by getting on the elevator regardless of the direction it's presently travelling.

This only becomes a problem when the call button calls multiple elevators to stop. However, this could be resolved simply by having one elevator travel from floor the lobby to the top floor and then perform an express to the ground floor. The other elevator does the exact opposite. You then mark the elevator doors UP and DOWN. If you accidentally get in to the UP elevator instead of the DOWN, you'll still eventually hit the floor you want, it'll just take longer and you'll figure it out sooner or later.

The problem arises from the fact that the elevators are acting erratically, they're jumping from floor to floor on no predetermined pattern. You have to pay attention to the elevator, which is a foolish form of pedestrian transport. I don't have to pay attention to the stairs, I just walk down them. I don't have to pay attention to where the bus is running, I just go to my stop and climb aboard and get off when I want.

It's kind of pitiful that multi-call elevator systems are more complex than even the poorest public transit systems.

I'm not sure what the problem is. Push the button, the elevator comes. Problem solved. Occasionally particularly dim people press the button wrong, but even then the system works.

Are people really that puzzled by how elevators work?

Listen, here are how 99% of the elevators I've used work.

You press a button for where you want to go. Up, or down.

You wait. Then, when the elevator arrives, generally the elevator will have some indicator that your car has arrived. Usually a sound and an arrow that lights up indicating the direction the elevator is heading next.

You get in. Press your button, and when you get to your next stop, you get off.

Now, you might see an elevator open up, and you'll get on. You'll press a button, and it will go in the opposite direction. That's because most elevators operate on FIFO. Obviously, with more complex systems, they can dispatch based on distance, but basically, if the elevator is coming up from the first floor, and a guy on the 5th floor pressed the button first, then you, being on the 3rd floor will just need to wait as the elevator passes you by. Of course it might open. Someone needs to get off.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact