Derailing illusions that kill: misperceptions at railway crossings (2003) | Hacker News Reader

Derailing illusions that kill: misperceptions at railway crossings (2003)

latimes.com

53 pointsmryall112 comments

>"The rat is always right." Green explained: "If you set up an experiment and the rat doesn’t do what you want it to do, it’s not because the rat is stupid, but because you set up the experiment wrong. It’s the same with humans."

I think this is important to remember in our field too. Too often I hear about a user story breaking down because "the user wasn't doing it right".. blaming the user is the wrong way to go -- you just set it up wrong.

Reminds me of a story I saw on here about restaurant staff using a physical whiteboard marker to cross bookings off on the screen because the software didn't do what they want. Maybe someone can find the story (I can't).

Users will always find a way to do what they need to do. The challenge is in finding out what they need to do because most of them would rather come up with their own solutions than come to the developers with a fully-specified feature request. I think this is why spreadsheets are so popular. They are never the best tool at any job, but users can bend them a million ways to do what they need to do.

https://javlaskitsystem.se/2012/02/whats-the-waiter-doing-wi...

Some places just plain don't care what the end users want. The worst version of this is when you're stuck in a shop where features are very obviously only dictated by managers. You can usually tell this is the case when the manager insists on being a MITM for all communication.

blaming the user is the wrong way to go -- you just set it up wrong.

Sometimes the user is wrong. If a person drives their car into a lake and drowns, do we say that cars are broken because they allow this? Or do we point to the millions of other drivers out there and say that driving on roads (and not into lakes) is the expected user behaviour?

An extreme example, sure, but I think it also translates to software. Some people criticize vi/vim for being modal because it confuses beginners. They use this as the basis of an argument that vi is a bad way to edit text or that modal interfaces are bad in general. But tons of other people take the time to learn the vi system of modal editing and they love it so much that they write plugins for everything else in an attempt to replicate that experience. Who is wrong here?

That vim dumps you into a screen with no guidance and where your keypresses appear to do nothing (or apparently random things!) would be the problem here. It would take very little to have it say "type :help to learn how to use this program" in the bottom of the screen when it starts up. It's also confusing because many Linux systems are set up by default to launch vim whenever another application requests an editor, even if the user uses something they are more comfortable with otherwise. This could be resolved by asking "which editor do you want to use?" instead of dumping the user into a randomly selected editor.

You can build powerful systems without confusing people who often never intended to wind up in your program in the first place. You can even build them to show people how to use them.

That's what the EDITOR environment variable is for.

Which is great... as long as you knew about that environment variable before you got into this situation in the first place. You're not exactly told about it during the install process or "first steps" guides for any distro I'm aware of.

> It's also confusing because many Linux systems are set up by default to launch vim whenever another application requests an editor, even if the user uses something they are more comfortable with otherwise.

  export EDITOR=/my/fancy/editor

Way to miss the entire point.

You usually don’t figure this out until after you’ve already run into this issue.

In Windows if you open a new filetype for the first time, it asks you what application you’d like to set as the default type, and on top of that it tells you where to go to change that later. That’s what beginner friendly UX looks like, not just throwing the newborn into the pool like Linux does with Vim editing.

It is not that I don't understand the "problem" - after all, I ended my very first accidental vi session with a reboot as well. But that was 23 years ago, long before Linux/GNU distros would be installed with a GUI and then boot into a GUI.

If the "newborn" in 2020 insists on using the terminal instead of the much more beginner friendly options that are the default nowadays... granted, but why compare a terminal editor in unix to a GUI? That is like comparing apples and oranges and then complaining that the strawberries don't taste like bananas.

Throwing up a menu to provide context and ask you what you want to do, rather than assuming a default, is not some thing that only magical GUIs can do. CLIs with menus is not some crazy new thing.

As an example of how manually configuring the editor flag can be annoying, consider a job where you regularly ssh into many machines, each that just by default use vim. I know how to use vim, I just don't like it, but pretty much every single machine I have booted into always assumes I want to use vim, and even if you remember to change the flag doing it to many machines all the time is majorly annoying.

It would be great to have a tutorial you can run, a tutor of some kind perhaps.

>do we point to the millions of other drivers out there and say that driving on roads

Cliffs, lakes, and highways often have guardrails, even if millions of cars drive by without going over the lines.

Vim, IMO, is different because it's not meant for consumption by the general public. Vim is a sysadmin tool, in unix-like environments, where the users have a general amount of competency before even encountering it. The whole rat-experiment analogy is more about stories that are public facing.

And sometimes people just calculate probabilities wrong.

People race trains all the time because they don't want to sit for 120 seconds. The problem is that losing that race means death.

The is especially stupid for the trains cited. "Metrolink" is not some slow cargo train that you will be trapped behind for 15 minutes. A "Metrolink" is about 6 train cars and is generally moving at a decent clip--probably no less than 35mph. You're not going to be waiting long.

LA has a much more significant problem with people walking in front of or committing suicide on the train tracks. Those cases generally far outnumber traffic accidents with the trains.

> Even as the train moves toward you, and you move toward it, the train’s image maintains a relatively constant position on your retina[...]

That's funny. The very first thing you learn on a ship is to watch out for exactly this. Any ship that maintains it's position relative to yours is on a collision course!

Same for pilots

Or it's moving away from the same point.

OK, fair enough: that's why they actually talk about CBDR - constant bearing, decreasing range.

This is how old air-to-air missiles worked. They would steer until the image of the target stopped moving. This was something simple enough that you could make it work electronically in the 1960s.

> This is how old air-to-air missiles worked. They would steer until the image of the target stopped moving.

That seems to be a weird way of describing "steer so the target is always in center of the image".

Not necessarily on center. Constant bearing but decreasing range will result in a collision.

Exactly! If the target was in the center, that would mean the missile was pointed to the target's current location. The target would quickly drift out of view!

Instead you want to intercept the target. You want to reach the point where the target will be, and at the same time as you. So keep the target to the side, at a constant angle.

Imaging missiles typically have the sensor mounted on a gimbal so the target can be centered in the image even when the missile body is not pointed at the target. But yes, the line-of-sight to the target will generally be off of the missile boresight.

I implemented this once in a video game. If the missile steers so it bears directly on the target, but has limits on its ability to turn, then it will often miss the target or spin around the target.

If you instead steer so the target’s bearing doesn’t change, and ignore whether that bearing is in the center or somewhere else, the missiles are much more likely to hit.

Yes this is called constant bearing decreasing range (CBDR) and it's a core lesson when you're training for your pilot's license as well.

Proportional navigation: https://en.wikipedia.org/wiki/Proportional_navigation

It’s also how baseball players catch balls, how many animals catch their prey, etc.

There’s no need to estimate position and speed of objects and then extrapolate them to predict whether course collide and how long that will take. It’s easier to directly derive that from optical information (think of it this way: if that train is 10% as far away, and also goes 10% faster, it will still hit you at the same time. That’s a degree of freedom brains don’t have to solve for)

(Couldn’t find a really good link. You’ll have to do with https://en.m.wikipedia.org/wiki/Ecological_psychology, http://www.scholarpedia.org/article/Optic_flow#Active_and_ec..., https://www.latimes.com/archives/la-xpm-1995-04-28-mn-59922-...)

Same reason why all road vehicles have brake lights. Too difficult to detect the (decreasing) speed of a vehicle in front of you.

Its because of this that road authorities usually place "mandatory head lights on" road signs on long straight single lane pieces of road. (because if there is high traffic there will be people taking advantage of the long straight to overtake slower cars, and during the day there is the possibility that a driver will not interpret a head on collision correctly)

In the Netherlands there was a horrible accident a couple of years ago with children who'd been picked up by daycare staff in a small electric bicycle wagon where 4 of the children were killed. I still remember the feeling of reading the headline. Since then I've read that municipalities are accelerating the transition from railway crossings in urban centres to underpasses, despite the huge cost of this re-engineering. I think the problem with railway crossings is that they are just inherently dangerous, and the severity side of the risk equation if you get a crossing wrong is huge (potentially multiple people being killed).

I like the Dutch approach here: replace them entirely.

https://www.bd.nl/oss/vier-kinderen-4-tot-11-jaar-omgekomen-...

Oh fuck, apparently the bicycle malfunctioned and couldn't brake. If only the rider had had the presence of mind to just steer sideways as hard as posible and take a roll instead...

Edit: I'm not trying to but any blame on her. It's easy to think in hindsight of what could have been done, another thing entirely to do it in seconds while panicking.

Yeah of course, completely natural.

My point is that even with all these insights into how human perception works, I agree with the strategy that it's best to bypass railways completely by going under/over them. And if it's doable in a densely populated country like the Netherlands then it's doable anywhere. (Some of the underpasses that have been built in recent years have required significant changes to the whole surrounding layout of the town, like one they completed in Bilthoven near Utrecht. Yet it was done.)

> And if it's doable in a densely populated country like the Netherlands then it's doable anywhere.

There's three different "doables" at play here:

- technically doable: can you actually convert the level crossing to an underpass or a bridge? For example, it's hard to do an under/overpass when there is a river 10m away, or when everything bar the road/rail and a bit of space is filled with buildings.

- financially doable: is there actually enough budget in the city/responsible authority's coffers?

- politically doable: will the population support the necessary construction works and the limitations that come with it? Like, will there be heavy protests of people that rely on either the rail or the road to get to work and would require massive amounts of time to bypass?

It is rare that there is a combination of all three being true, with the exception of course being tragic disasters with people and especially children dying as such events usually eliminate the political and financial blockers.

Yes, of course underpasses are the categorically better solution - if you can get the budget, which is often a hard problem. At least they have the positive side effect of speeding up road traffic significantly on busy crossings.

A railway crossing is the sudden appearance of an industrial environment in an otherwise non industrial setting.

Yup. British policy is the same. No new at-grade crossings. Existing at-grade crossings to be closed where possible. Where a crossing cannot be closed, upgrade it to achieve acceptable safety margins, revisit annually.

There are three different examples near me of crossings they haven't closed with very different scenarios.

1. A large industrial estate built on the tidal river is reached via a road only barely above sea level. The road is used by lots of heavy articulated vehicles (it is after all an industrial estate) yet it crosses a four track railway that serves both freight and passenger traffic to a major city. Result? A major at-grade crossing which is closed to road traffic for about 10-15 minutes at a time as much as three times in an hour. This crossing has full barriers, and a pedestrian bridge (so pedestrians needn't wait for the barriers). There's no way to dig down (it'd flood) and a road bridge would have to be very large to carry those articulated trucks but there's nowhere to put it. I expect this crossing will remain until forces of gentrification some day mean the industrial estate turns into housing and they just close the crossing altogether.

2. Where that same railway once went through the city to deliver rich people to the ocean liner terminal itself the passenger trains these days are diverted through a tunnel under the city - however a few freight trains per month still make the journey to the docks. So a wide city road goes over an at-grade crossing that is rarely needed, most users probably have no idea it's a crossing. This is a full gated crossing, but it's manually operated. A crew will come out, switch on stop lights, block the road, move the gates, then one train drives in or out of the docks, then they unwind everything. Because it's so seldom used a bridge seems unacceptably costly, and because it's manual I'd guess the residual risk is low so I expect this to remain in use essentially forever.

3. A rail branch towards another coastal city cuts across a residential street I used to live on. This one I can imagine closing. It's currently using full barriers but has no pedestrian route except to just walk to the nearest bridge. I think sooner or later somebody will get themselves killed, clambering over the barriers or whatever and they'll just close it because there's no way to make it any safer other than closing it.

Interestingly enough, Japan has a crazy amount of rail crossings, including smack in the middle of busy districts like Shibuya in Tokyo where the gates are down for the majority of the hour; yet they seem to have less fatal accidents than peers. I wonder what's driving the difference there?

As a bonus, replacing them also increases line capacity and reliability. The level crossing down the road from me closes something like every three minutes at peak times; any issues cause delays.

This article had lots of echos from The Design of Everyday Things. I highly recommend it if you haven't read it. Before I did "design" was a word meant mostly aesthetics with a little UX sprinkled on to me. This is 100% a design problem, and we have to design systems for people that are stressed, in a hurry, confused, think they are smarter than they are, you name it. The part about differentiating markings for active and passive crossings is especially poignant for me. I ride bikes a lot and cross numerous passive crossings on smaller country roads. It's not that I don't understand how those crossing works, but I'll admit that I often subconsciously think to myself, "they designed this not to kill me if I don't disobey the bells and gates" as I tear past it after a quick glance.

Styling is the word. It's different to design. It's only about looks. Design should be about functionality and ergonomics etc.

In my native language, they still haven't really found out a satisfactory way to differentiate those.

One thing I remember from that book that drives me crazy is the bit about push vs. pull door handles. You see handles that you intuitively pull on both the push and pull sides of doors all the time. (Our HQ building is one of them. I'm usually only there once or twice a year and I can never remember to pull or push the handles on both sides of the elevator core.

In a sane (or well-regulated) world, all doors would always open outwards, in the direction of the nearest exit. In case of fire, you don't want a horde of people pressing an inwards opening door shut

They usually do and I imagine they do open out of the elevator core in this case (because the stairwells are outside the core). It's just not mentally intuitive (to me) that going "out" in this layout means going from the elevators to the offices. I'm sure it would be natural if I worked in the office but I don't so I need to think about it--or just do it wrong first :-)

ADDED: And while "out" is obvious with an exterior door, it may not be with an interior door (and in fact there may not be a clear "in" and "out".)

I would expect that it's actually a requirement and that you could get fined for that in an inspection.

146 people died in club in France due to inwards doors in 1970: https://en.wikipedia.org/wiki/Club_Cinq-Sept_fire

I don't know the scale of the fine but using inwards doors are a big no-no.

And similarly, electrical door must unlock themselves when losing power (unless special buildings like military or dangerous laboratories).

I seem to remember the classroom doors of my elementary school (I'm in my sixties) opening in to the classroom, so that during an emergency they wouldn't be pressed closed by the panicking hords rushing by in the hallway, and so that opening them wouldn't injure those panicking hords.

You also often see bathroom doors that open in; I had this random conversation with someone at a conference last year. I assume the theory is (assuming there is one) that there would rarely be so many people crammed in a bathroom that there'd be such a crush to get out that it would be hazardous. On the other hand, even day-to-day bathrooms are sometimes located such that someone quickly walking down the hall looking at their phone could fairly easily be slammed into by a door being swung out into the hallway.

I do notice some improvements in a few places over the last 17 years. Many busy highways now have overpasses. Many grade crossings have traffic lights as well. That’s good design in my opinion, like putting a stop sign bar on the school bus.

I don’t know why grade crossings without gates do not also have stop signs. We have to assume that many drivers are unaware of the rules of the road, either through lack of training, forgetfulness, local culture, or inconsistencies with other countries.

"One peculiarity of human perception is that large objects in motion appear to be moving more slowly than they really are" - Folks, never cross the roads/tracks when you see a vehicle approaching. Wait for it to pass and then cross.

Same applies to ships, especially for those in small craft.

Crossing roads is different - and it would be often impossible if you had to wait until no vehicle was approaching.

But as this article points out, you can't use your experience judging car distance/speed to judge train distance/speed.

How far away is the vehicle? How much time do you have?

From a fairly recent HN article -- see Lamport's discussion of Buridan's principle.

http://lamport.azurewebsites.net/pubs/pubs.html#buridan

The hell? Just put big automatic gates everywhere. Seems like a simple life/cost estimation exercise.

Think about the thing you know best. Imagine an article about a challenging problem related to that thing. A person with almost no knowledge of that thing reads it and says, “Seems simple enough” and presents a solution. Are they likely to have overlooked something? Misunderstood the problem? Be unaware of the challenges in solving it?

By my interpretation of the article, it seems to agree that this is the solution. The caveats seem to come mostly from paying for it.

Yes, but sometimes just bluntly stating a solution can be a request for enlightenment.

Can we just put this comment on the top of every discussion here?

"Why don't they just ..." https://blog.jgc.org/2012/08/why-dont-they-just.html?m=1

> Seems like a simple life/cost estimation exercise.

All safety mechanisms are this. The tricky bit is the value of life calculus. Is it worth an added $10/year tax for every taxpayer to MAYBE save 1 life per 5 years? You'll get different responses for different reasons.

That $10 may be a rounding error for many, but to some its the difference of a week of lunches for their kid or not, and they vote, too.

Just for completeness $10/year from every taxpayer is $1.43B/year from one billionaire. So the problem of $10 being unaffordable to some is easily handled.

Asking someone else to pay is always easy for those not being asked, sure.

I wonder if that's the most cost-effective way to use $1.43B to save lives?

With trains the cost/benefit is greater than just stopping the crash from taking a life. Train/car collisions are messy and require fixing the infrastructure that has been disrupted (a collision/derailment will most likely screw up the tracks and any equipment in the way); and the train line will be out of service for the better part of a day, at least.

Read Darwin Awards. You find even with the gates in place people get out of the cars to raise the gates and still try beating the trains.

Actually no. Cannot be too big and automatic. You need to allow a stranded car to break free. You must not crush a car that is only half way in.

Sigh.

Railway crossings are pure evil. There is no fail safe design that I know of without adding a proximity sensor inside the car (I am a railway signalling engineer). Would love to be wrong though.

I know that at least in this area some wyes and crossing points use radar (the same type used for lane occupancy on roads) as an additional safety measure---something like track circuit clear but radar detection of an object results in a very restrictive signal like a "red-yellow." Do you know if there's a reason why this hasn't been deployed for road crossings? Definitely it wouldn't be perfect but seems like it could help with stuck-vehicle scenarios, especially in more urban areas where speeds are low but visibility is poor. I guess I assume that the reason it isn't being done is because it would require some kind of upgraded cab signalling that would cost too much to roll out.

Googling "MCB-OD crossings" will bring up various pages and videos that show such crossings in the UK. They are relatively new to the UK but I believe they have been used for some time in continental Europe.

They have disadvantages including expense, and additional road closure times (compared to automatic half-barriers) as they must be closed and proven clear early enough that an approaching train does not need to slow down.

Trains are very heavy and have long braking distances at speed, in some cases over a mile. For a train to have enough time to stop, the crossing would have to be kept clear for a very long time, which would probably tempt drivers to bypass the gates since they would perceive the crossing to be overly safe (described in the article as the "cry wolf" phenomenon).

In the UK many crossings have full width barriers, and must be confirmed to be clear of road traffic after the barriers have closed before the protecting signal can be cleared for a train. These are effectively fail safe.

Confirmation comes from observation by a signaller or crossing operator either directly or by CCTV, or automatically by LiDAR and radar. These crossings tend to be used in areas of heavy road, pedestrian or rail movements, and when train speeds may exceed (from memory) 100 mph / 160 km/h.

Their disadvantage is additional installation, maintenance and operational costs. They must also be closed for longer before a train arrives than an automatic half-barrier crossing to allow for the required safety checks and ensure that signals are cleared far enough along the line that approaching trains do not need to slow down.

This is probably better than nothing, but it is amazing how often cars get stuck in them.

I was once closed in by automatic beams here in NL. Not a fun experience having to lift my bike through some hedge to get out.

For millions of years, the largest moving object humans had to deal was the size of an elephant. You can look at the time the object travels its length in and get some rough estimate of speed from there. You won't be mistaken that much.

Now, take a 50 meter long airplane or 100 meter long train. Or a 300 m long ship. It's not going to work - fatal accidents ensue.

TIL: "The rat is always right."

//edit: okay, the above is a bit minialistic, so, in full: The article is a good point on how systems have to adapt to actual human (mis-)behaviour, instead of relying on wishful thinking on how humans should behave.

A hundred years ago, tradespeople were often missing fingers. For some reason, it was really hard to sell the idea of safer equipment.

I bet it has something in common about hand washing in hospitals taking some hundred years to become mandatory - or how it was done much earlier in Hungary, yet nobody outside believed in it.

Railway crossing design has parallels in software design. Rare events, expensive safety/security mechanisms, expensive redesign, and safety/security mechanisms that are perceived as repetitive irritants rather than timely helpers.

This same set of perceptual illusion is the cause of one of the more common motorcycle crashes, colloquially known as the SMIDSY (Sorry Mate I Didn't See You) accidents:

https://motorbikewriter.com/smidsy-biggest-cause-crashes/

https://motorbikewriter.com/scientific-studies-explain-smids...

Most drivers find it very hard to estimate the closing speed of a 2 wheeled vehicle approaching an intersection.

> “Railroad crossing deaths in the U.S. have come down from 786 in 1975 to 315 in 2001. ... I think this was largely due to the U.S. government’s rail-highway crossings program which since 1978 has injected $4 billion into crossings improvements"

Cars have become a lot safer in that time. And the number of crossings changed by a factor of 0.68. Both of these could account for the reduction in railroad crossing deaths.

> the traditional American crossbuck, the simple X-shaped railroad sign that warns drivers to yield to an oncoming train.

I would guess that roadway signage in the US runs about 80:20 textual:symbolic, but that the EU runs about 20:80, with much heavier use of symbols, and use of text only to modify or explain something not covered by the available symbolic palette. Could any EU->US or US->EU migratory drivers make their own estimates?

"How did you not realize that the empty circle means no parking from the hours of 1000-1600??"

Is signage not in the driving exam in your locale?

For someone coming to a European country and driving by virtue of reciprocal driving license permission (not having to take an exam), those signs are not at all intuitive.

They're not meant to be intuitive, you're meant to learn them.

In our country we have unprotected (sign-only) crossings, red light crossings without barriers, red light and barriers crossings and red-white light crossing (with out without barriers). The blinking white light informs the safety system is working and it is safe to cross. Unfortunately, they are actually removing the white lights to match the rest of the EU.

Red light crossings seem especially disadvantagous for red colorblind individuals with the addition of white light signals.

One could say that these railway crossings are adversarial examples that trigger visual misperceptions in the organic neural networks inside our heads.

https://archive.is/Rynhm

>The tendency to blame the victim

They're only victims of their own stupidity. The real victims are the train drivers who are often traumatized, injured, or even killed in these collisions through no fault of their own.

The same goes for other occupants of the car. Pretty much everyone but the driver.

> One peculiarity of human perception is that large objects in motion appear to be moving more slowly than they really are

Rule #1: if you are at a railway crossing and you see a large object in motion coming down the track, then stop and let that large object pass before you try to cross the tracks!

I love how this article tries to make this sound like a complex affair. Only the most ridiculously stupid person would be in a situation where they are continuously trying to judge the position of the train relative to their own position as they attempt to beat a crossing. The only situation where you should ever be remotely in danger is if all of your braking and steering catastrophically fail at that exact moment and there is nothing to stop you from rolling directly into the train.

I also find it amusing that all of the stories regarding train-car interactions seem to occur at crossings with modern signaling devices, rather than "at your own discretion" crossings in the middle of nowhere. This leads me to suspect that the typical human has been conditioned to treat any signalized intersection transition as a "beat the yellow" event, but in the case of a rail crossing the brutal physics equations seem to be conveniently ignored.

Perhaps there needs to be a day in driver's education courses where everyone has to review just how heavy a freight train is and how much kinetic energy must be dissipated in order for it to come to a complete stop. Maybe make them ride at the front of a train to experience how painfully long an emergency braking maneuver takes to complete.

90 day license suspensions for those who ignore signals should do it.

Those that ignore the signals and live, that is.

In German written driving license test, there is a section literally dedicated for railway crossings.

Edit: Clarified that it is written test.

In my state of Georgia it was the only thing I failed on my driving test (we didn't have driver's ed when I a kid, you just practiced with your parents and then took one test). I yielded and looked but didn't come to a full stop. I remember it took 15 or 25 points off of my score, but since it was the only thing I failed, I still passed the exam.

The things is, I don't remember the written test I took having any content at all about railroad crossings.

> I yielded and looked but didn't come to a full stop.

Physics tells us that we should accelerate to spend the least amount of time in the danger zone (rail bed). Since the velocities are orthogonal, it doesn't matter if you are hit by a train when your car is going 55 or 5 MPH (probably, unless trees in the resulting vector of your now pulped car).

It never made any sense to me to see school busses stopping at a grade crossing, then taking an extraordinarily long time to cross those tracks. What happens if the engine fails while on those tracks and a high speed train appears.

No, my vote is to make sure you have a clear view of the tracks in both directions and punch it. Less probability of interacting with a train.

Thanks, 17-year-old me feels validated.

Here in India we seem to have a large number of such ridiculously stupid people (perhaps a consequence of having a large number of people in general) but it also becomes a street-cred thing as someone in front of you beats the train at a railway crossing and now you look like a wuss if you just stand there and wait for the train to pass.

This kind of accident is so common I remember the government commissioning a series of funny PSA ads that played on national TV showing people acting recklessly at railway crossings, getting killed and then expressing their regrets from beyond the grave.

Be careful, as many topics in railway signalling, this is actually fairly complicated and tricky. As I said in another comment, crossings are pure evil. They are many hazardous scenario and without any system inside the car, mitigations are not fail safe.

Personnally, I belong to the school of « remove them all, whatever the costs ». A little extreme, but €&@% I hate these things.

> I also find it amusing that all of the stories regarding train-car interactions seem to occur at crossings with modern signaling devices, rather than "at your own discretion" crossings in the middle of nowhere.

A large part of this is due to population density (or trip density) around the crossing. If you've got 1000x the number of trips on a crossing, a 100x safer crossing is still 10x more likely to result in a colission. Fill in your own estimates, as my numbers are clearly made up. Also, guidelines require more signalling in built up / dense areas, and at intersections with a history of colissions.

Additionally, if you have a colission at an unguarded intersection, the response is often to consider a guard, and if it happens again, the story is 'senseless officials refuse to put in guard' rather than 'sensless person disobeys guard and is injured/causes confusion and delay'

Given that it's clear that this is not how people behave in practice, maybe we should take different steps to prevent the problem? Rather than blaming the user for the system's poor UX?

"Human error" isn't a very useful "root cause" for any incident. Humans make errors.

> Given that it's clear that this is not how people behave in practice

Is that clear? Most people don't get hit by trains, so I don't think that's clear. I'd say most people are cautious around trains while a minority aren't, and a minority of those who aren't will get hit by trains.

That's a dangerous way of thinking, and is in contradiction of the article. It implies that people get hit by trains because they are stupid, and it'll never happen to me because I'm not stupid. I would imagine that most of the people who do get hit by trains don't think they are stupid.

There's a physiological reason why people have a tendency to underestimate risk in certain situations, or miss obvious pending collisions. The usual comment from someone who has had such a collision is "They just came out of nowhere". See also https://news.ycombinator.com/item?id=23024275

Airplane pilots (particularly military fast jet), and ship captains are taught specific techniques to deliberately and systematically overcome these physiological limitations. Some of the techniques are taught in some driving schools (for example https://news.ycombinator.com/item?id=18157090 ) and to be frank I think driving education should go all-out on it.

The article argues that it's difficult to judge the speed of a moving train, and I don't doubt that's true. However if you're attempting to judge the speed of a train at all while at a train crossing, you're already doing it wrong. The cautious approach is to stop and wait for the train no matter how fast you think it's moving. This is what I do every time. This is what drivers ed programs teach. It's what common sense dictates. It's what most people do, and is the reason most people don't get hit by trains (3000 collisions per year is not very many at all when you consider how many times people cross train tracks each day.)

Might I still be hit by a train I never saw at all? Sure. Somebody I knew in highschool was hit by a train on a track he believed was abandoned and consequently, never looked down (he lived after being pushed 100m down the track). It's possible that might happen to me, at least if I ever forget that it happened to him. But what won't happen is getting hit by a train I see and judge to be moving slowly or not at all, because I stop for any train I see no matter how much time I think I might have to cross the track.

> Airplane pilots (particularly military fast jet), and ship captains are taught specific techniques to deliberately and systematically overcome these physiological limitations.

The "specific technique to overcome this physiological limitation" is: Don't race trains. Don't race trains even when you feel certain you can easily beat the train. Racing trains is stupid.

> This is what I do every time

Nobody cares what you do all the time, though. In order to reduce accidents overall, we care about what the average person does. And if the average person does something else, then common sense isn't really the right term.

And even average person isn't the right bar. To really cut down on accidents, we need to see what the person who's behaviour is a couple of standard deviations less safe does in various scenarios.

> And if the average person does something else, then common sense isn't really the right term. And even average person isn't the right bar.

It's quite evidently you don't think the average person is the right bar, because the average person doesn't race trains. 3000 a year is next to nothing compared to other common hazards.

Maybe we should eliminate all intersections, since although the average person obeys signal lights and signs, the occasional idiot chooses to run red lights instead. We should eliminate crosswalks too; the average person looks both ways before stepping into the street, but occasionally people don't (this happens FAR more often than train/car collisions.) Let's ban bicycles too; the average cyclist wears a helmet but some choose not to and get their heads split open. Let's also ban self-service gas stations; the average person can pump their own gas without burning the station down but sometimes some people fuck up and cause huge fires. Doesn't that mean we should all follow the lead of Oregon and New Jersey and ban this activity? 'Nobody' cares that you can pump your own gas safely, please be considerate of the rare idiot who can't.

If there's a level crossing without an automatic barrier, that's certainly poor UX.

But if there's a working automatic barrier, and the user bypasses it? Tough to see how you could UX your way around that.

If people are reminded using novel imagery it can help a lot. A few years ago Melbourne introduced a campaign on its tram network that had a rhinoceros riding a skateboard to remind pedestrians/drivers not to cross/turn in front of a tram. It got people’s attention about the danger through a humorous analogy.

https://www.fizzicseducation.com.au/articles/science-in-publ...

https://www.youtube.com/watch?v=wdeqwNzMhJQ

Maybe this applies more to countries with faster trains? They only go ~30mph here and I never had any problem crossing tracks on foot with trains in the far distance, and the rest of my family did it too. These tracks also had a lot of curves though.

No, it’s really important advice.

The big issue with trains is that it has an inertia that our brain is not trained to.

As signalling engineers, we get to compute safe braking distances a lot and they are /big/, even at relatively low speed. This means that trains in curves -> do not have line of sight all the way to the next possible stop point <- so the driver may never have the chance to stop if you fall while crossing the track.

So don’t do it. Even at 30mph in curves, you are in danger.

I would always assume that if a train driver sees something on the track ahead, she is going to start braking. And I know that even if I believe I am going to be out of the way in time so they won’t have to stop (or hit me), merely causing a train to brake is going to cause inconvenience and delay and waste energy. So I’m not going to step in front of a train, ever.

> The tendency to blame the victim in grade-crossing accidents exasperates cognitive psychologist Green: “That lets the authorities off the hook. Then they don’t have to redesign the system.”

And why should they? Some people die, but how many times more people do cross successfully?

And then we have a nuclear incident because "everything we did worked for 30 years so we didn't do maintenance".

I mean, you could say that of practically any safety rule. "Sure, some people die of cholera, but most don't, so continue dumping the sewage into the drinking water".

Reply on news.ycombinator.com