You’re dead if you aim only for kids. Adults are only kids grown up, anyway.
On 10 January 2000, Crossair Flight 498 crashed ~2 minutes after taking off from Zurich. All 10 people on board were killed. The accident investigation report blamed human errors, including the fact that the captain might have been under drug influence, as major causes of the crash. However, it was evident that the cockpit interface design and the use of words in communication between the pilots and the air traffic controller also played a role in the accident. These communication and interface design errors revealed some simple yet crucial human user interface design principles. These can be seen and be compared with the interface design of OSX Mavericks.
First, allow me to detail the accident of that faithful day. The plane involved was a Saab 340, which was a Swedish built commuter airplane. It was scheduled to take of at 5:55 pm that day. However, right after taking off, air traffic controller made a slight change to the flight path. The new path called for a left turn, looping the plane around the airport, before heading north to the destination. However, mysteriously, the captain started turning the plane to the right. Air traffic controller saw the unusual movement on his radar screen and questioned, “Crossair 498, confirm you are turning left.” The copilot replied “standby”, but the right bank angle kept increasing. The captain struggled to control the plane. The plane eventually fell out of the sky and slammed onto the ground, killing everyone onboard.
Artificial Horizon: A crucial instrument to tell the bank of the airplane with respect to the ground. In this image, the plane is flying levelled.
Now, image what the horizon would look when a plane turns right from the of the pilot. The horizon will bank left as shown in this real POV photo from a glider turning right.
In a western built airplane, this sight from the pilot’s perspective is replicated in the artificial horizon. On the display, the plane is fixed on the diagram and the horizon bank with respect to it. For example, in a right turn, the indicator will show:
Artificial horizon during a right turn for Western planes.
However, Russian’s design of the same instrument took a different form. Instead of fixing the airplane and tilting the horizon accordingly, the horizon is fixed and the plane tilt.
Russian artificial horizon when a plane is turning right.
Now, it is not hard to see that, at a glance, a right turn on a western instrument looks similar to a left turn on a Russian instrument. Both diagrams contain a similarly looking “axis”, but convey opposite information! As both pilots were trained under the Russian design but were flying a western built aircraft on that day , this conflicting interface proved to be fatal.
As the Russian pilot turned right, instead of left, he was told by the copilot about his mistake. However, judging from instinct when looking at the artificial horizon, he thought he was really turning left, just as he was supposed to do. Thinking that the plane might not be responding to his left turn input (when in fact he was turning right), he turned left even more (when in fact, he turned right even more). This spiralled into a negative feedback loop which led the fatal right bank.
This is an exemplification of the importance of consistency of interfaces, especially when the inconsistency would lead to opposite information or actions. This is a well known principle in interface design. For example, in OSX, when you delete an item, be it in finder, iTunes notes or reminders, the deletion confirmation dialogue has the same exact layout. The confirm action button is always on the right and the cancellation button is always on the left, unlike the artificial horizon which took different form on different airplane designs.
Many of us has built-in muscle memory of the button layout of the deletion dialogue. Imagine how disastrous it would be if suddenly a dialogue pops up with the confirmation button and cancellation button swapped? Consistency in interface design is of paramount importance.
2) Use of Language in Warnings
As the pilot made the deadly turn, air traffic controller’s dialogue only fuelled the confusion. He questioned, “Crossair 498, confirm you are turning left.” when he saw the plane turning right on the radar. In retrospect, this didn’t sound like a problem, however, you must remember that, at that moment, the pilot really thought he was turning left when he was in fact turning right. This conversation only re-affirmed the pilot’s wrong interpretation of the artificial horizon indicator.
Instead of saying "confirm you are turning left.”, the air traffic controller should have said ”You are turning right, please turn left.” or something similar. This careful wording is very important in dialogue design. When a user is performing a bad habit, instead of being informed about what he should do, the fact that he was doing it wrong must be re-enfored first. This is evident when you try to remove a drive without dismounting it first in OSX.
When a user eject a disk improperly, more often than not, he didn’t know that disks have to be dismounted before removal. So, the phrasing “please remove disk properly" wouldn’t help, just like how "confirm you are turning left” didn’t help the pilot at all. Instead, OSX worded the dialogue as “disk not ejected properly”, which enforced the fact that the disk was removed improperly, just like how the air traffic controller should have said “You are turning right”.
3) Location of Information on Screen
Part of the air crash investigation went into discovering why it took a long time for the copilot to find out that the pilot was turning in the wrong direction. During the take off, it was agreed between the pilots that the captain would be responsible for the flying and the copilot would be taking care of other take-off related tasks, such as raising the landing gear or keying in flight level information into the flight computer.
As we all know, the cockpit is a very complicated interface. The artificial horizon is usually placed just in front of the pilot and the copilot. However, for the take-off tasks that the copilot had to perform, he needed to look and reach out to various corners of the cockpit (i.e. away from the centre and, thus, away from the artificial horizon). As a result, it took an awfully long time before the copilot caught on the artificial horizon and realised that there was somethings wrong with the direction of bank. By then, it was too late already.
This is an important lesson. An interface should present critical information together and group tasks together so that the user can be informed about the entirety of the system at a glance with minimal effort. For example, whenever you looked at the time or turn wifi on and off on a Mac, you have also “unintentionally” glanced at the battery level, a critical piece of information. By putting all information —- bluetooth, wifi, input methods, time, battery level, time machine backup status, volume etc. —- on the upper right corner of the screen, the user can comprehend the system status with the least effort. Taking a step further, you will notice that all buttons and menu bars of windows on a Mac are positioned at the top. This leads user to look upward consistently. By having everything on the top, the design allows the user to be more focused on a smaller area, reducing the chance of error, improving efficiency when conveying information and enhancing functionality discovery.
As we can see from the investigation report of Crossair Flight 498, cockpit interface design strategies can be generalised to many different interfaces, most notably desktop OS interface design. More often than not, poor interfaces will only lead to user confusion, yet, in the case of Crossair Flight 498, will can see that the result can be fatal.
You can learn more about the crash of Crossair Flight 498 by watching the new (dramatised) documentary from National Geographic:
Nowadays, almost all MacBook Pros and all MacBook Airs come with SSD drive. SSD drives’ speed and performances are impressive. However, the only downfall is the capacity. To be honest, it is really hard to live with a computer with only 128 or even 256 GB of internal storage. So, I take every opportunity to expand the storage of my MacBook Pro very seriously.
Having external hard drives are possible solutions, but is inelegant and inconvenient. Instead, the Nifty Drive solution really caught my eye. It uses the SD card slot that comes with all MacBook Pro as a slot for storage expansion. Essentially, the Nifty Drive is a micro-SD card reader that fits nicely into the SD card slot. The dimension is specifically designed not to stick out of the computer and the colour matches the beautiful aluminium enclosure almost perfectly. With a 64 GB micro-SD card, I have expanded my 256GB MacBook Pro storage by 25%! And, I have no ugly drives and tangling USB cord to carry round all the time!
On the rare occasion that I’ve to use my SD card slot to transfer photos from my DSLR to the Mac, I can simply dismount and pull out the nifty drive.
The nifty drive requires the nifty tool to be removed. The nifty tool is just a nicely designed and made hook that allow you to pull out the drive with ease. I always keep the removal tool, along with my iPhone SIM ejection tool in my wallet, just in case I need to use my SD card slot for whatever reason.
As far as the Mac is concerned, the nifty drive is still an external SD card slot. So, it will be mounted as a separate drive. This is very convenient as I have set the nifty drive as an time machine external backup storage that backups all the files in my document folder (i.e. NOT the entire mac as the drive wouldn’t hold everything. Just the critical documents.)
The read and write speed of the drive of course depends on the speed of the SD card. Using a class-10 SD card guarantees maximum speed.
Theoretically, having a SD card mounted constantly does drain a little bit more battery. However, as far as I can tell the difference is very minimal. So much so that I can’t really tell the difference!
All-in-all, I think nifty drive is a very genius solution to the limited storage problem of SSD drives. Yes, indeed, it costs, especially considering that it is, after all, just an SD card drive. But the design and seamless experience and look do pay off. As a matter of fact, I think you will have paid more on buying the micro SD card then the nifty drive anyways! I am happy to recommend it to everyone, except for those who own a 11” MacBook Air simply because, well, the 11” MacBook Air doesn’t contain a SD card slot!
Though, the nifty drive left me pondering a few questions:
It is incredible to realize that the personal computer revolution started only 30 years ago.
Nowadays, the internet is decentralised and unorganised. Users interact dynamically almost randomly by clicking, sharing and creating webpages. However, it is not hard for one to imagine a global structure emerging from this disorderliness: E.g. people are organised in a spider-web-liked structure on Facebook.
The internet is an example of self-organisation in which macroscopic structures are formed out of chaos. This phenomena might be counterintuitive as random systems should naturally become more and more randomised. After all, a messy room can only get messier over time unless some mechanisms intervene, such as a person cleaning up the room.
While the self-organisation mechanism is well-studied in many regime of physics such as crystallisation and fluid dynamics, those in the area of astrophysics are generally not. For example, supernova remnants, solar wind shocks and astrophysical bodies like the Herbig-Haro object are results of global structures (e.g. collision-less shockwave) emerging from turbulent plasma that are not well understood.
Although these phenomena have long been studied using satellites, their mechanism and even the necessary pre-conditions are still unknown. Researches have now tried to recreate stable and macroscopic structures in plasma in laboratories, hoping that greater parameter control could generate more insightful models. For the first time, physicists could make accurate measurements of plasma self-organisation processes that take place deep in the cosmos.
2 Counter-streaming plasma were created by focusing high energy beams on target disks. The intensity of the beams was a quadrillion time more than that of the sun on earth’s surface. The counter-stream meant high relative velocity, which rendered inter-beam ion exchange rare and facilitated the formation of collisionless shock — an important macro-feature of interest. The interactions were recorded using proton imaging.
After 2 nano-seconds — the time needed for light to travel 60 cm — turbulent fields developed. Shortly afterwards, sharp caustics structures started to form. Caustic structures were similar to pattern of light formed when sunlight shines through rippled water. Then, the structure arranged itself into two horizontal regions, leaving an empty centre.
Remarkably, these structures spanned distances much longer than any other plasma feature. They were also so stable that they persisted during the entire experiment window. Given these time and length, these features were definitely “global” structures that emerge from the chaotic plasma beam.
A paper published a year later hypothesised a few explanations. Some models relied on electrostatic phenomenon while others explained using the formation of toroidal (doughnut-shaped) magnetic fields. The latter successfully accounted for the general features but relied on too many assumptions.
Further investigations are currently underway with improved data collection and tomographical techniques. Indeed, merely analysing proton imaging data is inconclusive. Methods that examine density gradients and magnetic field patterns have to be deployed. Future experiments can also be conducted at the National Ignition Facility, which allows researchers to produce these plasma structures at a much larger scale.
Ultimately, models of plasma self-organisation developed in the laboratory can help astrophysicists to understand the emergence of cosmic structures better. In the long term, the implications go beyond astronomy and facilitate a better understanding of self-organisation processes from microscopic to cosmological scales.