Distracted Drivers

I try to pick subjects where I can either provide new (or at least not well known) information or a different perspective.  The bulk of this post will be rehash but I do have something new to add.  It's down at the bottom.

There is a computer term:  multitasking.  Over time the term has moved beyond its initial use only in a purely computer context to now frequently being applied more broadly.  As we will eventually see, it now applies even to the subject at hand.  But let's start with it in its original context.  Old computers were really slow compared to their modern counterparts.  But there was still a problem.  Generally speaking data would be pulled into RAM and processed.  It was then spit back out, typically in a new form.  But where was it pulled from or pushed to?

The old technical term was "peripheral devices", gadgets that were connected up to the "computer" part of the computer.  For modern computers these might be a disk drive, network card, etc.  Or they could be a human interface device; something like a keyboard, mouse, or touch screen.  Consider for a moment a keyboard.  A good typist is capable of typing 60 words per minute.  That's one word per second.  And, for the purposes of measuring these sort of things, a "word" was 5 characters and a space.  So in our example the computer would expect to see a new character every sixth of a second. 

But even a very slow computer can perform a million instructions (computations) per second.  This means that the computer can perform over 150,000 instructions while waiting for the next character to appear.  You ought to be able to do something useful with 150,000 instructions so it seems like a waste for the computer to sit around idly doing nothing useful while it waits for the next character to come in.  The solution was to have the computer work on two or more things "at once".  That way it could beaver away on problem #2 while waiting for the next character aimed at problem #1.

Now pretty much every peripheral device can handle data faster than a keyboard.  But the above example illustrates the problem.  And early computers were fantastically expensive.  They cost several millions of dollars each.  Wasting even a little computer time amounted to wasting a lot of money.  So various techniques were devised to allow the computer to have multiple tasks available and to be able to quickly switch from one task to another.

Most of the time any one particular problem (or task) is waiting for some I/O (Input/Output - a read or a write) operation to complete before it can continue on with the job at hand (executing instructions).  This "fill in the otherwise idle time with useful work" idea is the driving force behind many of the early networking efforts.  If several terminals are hooked up to the same computer then the computer can switch from working on the task associated with one terminal to working on the task associated with another terminal whenever the first task is hung up waiting for I/O to complete.

So the benefit was obvious.  Multitasking could keep the computer busy doing useful work more of the time.  But there was a cost.  In the very early days of computers RAM was very hard to make and, therefore, expensive.  A reasonable amount of RAM might cost a million dollars so every effort was made to keep RAM requirements to a minimum.  You needed a lot of RAM to have enough to keep the critical pieces of several tasks (say one for each terminal) resident in RAM at the same time.  But the price of RAM dropped and the cost of getting enough RAM to enable multitasking soon became manageable.  But there was another cost.

The computer doesn't really do multiple things at once, or at least computers couldn't in the old days.  So a piece of software called a "task switcher" was necessary.  This software kept track of all the tasks currently loaded into RAM.  It also kept track of what "state" each was in.  A task could be "running" or "waiting to run", or "waiting for a specific I/O request to complete" (the most common state).  The running task would go along until it needed to perform an I/O operation.  Then it would turn things over to the task switcher.  The task switcher would schedule the I/O, put the task to sleep, then look around for another task that was waiting to run.  If it found one it would wake that task up and turn things over to it.

It turns out that it takes the execution of a lot of computer instructions to do a task switch.  I have skipped over a lot of detail so you'll just have to take my word for it.  The modern "task switch" process eats up a lot of instructions doing its job.  So what's the point?  The point is that these instructions are not available for use by running tasks.  An old and slow computer from my past would often dedicate up to 45% of all instruction executions to task switch and other overhead processes.  If without task switching you can only keep the computer busy 10% or 20% of the time this overhead is a good thing but it is still expensive.

And that cost associated with multitasking applies to other contexts like people.  Most young people multitask all the time.  They might be sitting in class and simultaneously monitoring Twitter feeds and updating Facebook posts.  They are doing the same thing computers do.  They devote a small slice of their attention to one thing, say the lecture.  But then they quickly switch to focusing on their Twitter feed, but again not for long.  Because they almost immediately switch to the Facebook post they are creating.

If you ask them they will say that what is happening is something akin to what happened on computers in the olden days.  They are able to task switch quickly enough and often enough, and efficiently enough that their net productivity goes up.  The problem is that everyone who has actually studied the net productivity of multitasking people finds that their net productivity goes down, sometimes by a lot.  They think they are doing multiple things well at the same time.  But the studies show that they are actually doing multiple things badly at the same time.  So how does all this apply to driving?

It turns out that it is directly on point.  This same model of trying to do multiple things at the same time means you do them all badly.  And you especially drive badly.  People think they can task switch quickly enough and efficiently enough that nothing important happens on the road in the small time their focus is away from driving.  But every study says they are wrong.

And it turns out that there is a model for what is going on, driving drunk.  Drunk drivers aren't multitasking.  But they are doing the same thing multitaskers do.  They fail to sufficiently focus on their driving.  In the drunk driver's case their mind tends to wander.  They are not switching from one productive task to another.  They are switching from a productive task to a non-productive one, effectively day dreaming.  This behavior pattern and its impact on driving was recognized decades ago and the response has been Mothers Against Drunk Driving.  An emphasis on getting drunks off the road has cut down on crashes.

But with the advent of the cell phone things changed.  You didn't have to be drunk to drive badly.  If you were switching your focus between the road and your phone the results could be similar to driving drunk.  Lots of drunk drivers believe they can drive well while drunk.  Similarly, lots of cell phone users believe they can drive well while using their phone.  And it turns out that to some extent they are right.  What?

Put simply, there are times when driving requires laser focus and times it doesn't.  If you are driving on a straight road in good weather and there is no one else on the road driving does not require much intellectual effort.  On the other hand, let's say there is a lot of other traffic on the road.  And the speed of the traffic changes drastically and frequently.  And say your sight lines are impaired (or the weather is bad) so that things can "come out of nowhere"; a driver pulling out of a parking space, a pedestrian hopping between cars outside of a crossing zone; a bicyclist cutting in and out of traffic.  Then driving requires a great deal of attention and it requires it pretty much continuously.

The basic question to ask is "how many decisions per second need to be made".  Coupled with this are "how much time is there for the decision" and "how many items must be factored into the decision".  If the number of decisions per second is low and the amount of time permitted for decision making is long and the number of items is small then a great deal of "free time" is available without a significant diminution in your quality of driving.

Looked at this way we can see why the first scenario is easy.  Few decisions per second are required as not much is going on.  The good sight lines mean that there is a relatively long time within which to make the decision.  And few factors, perhaps one or two, need to be allowed for.  This results in few decisions needing to be made and not much effort being required to reach the correct decision and implement it.  And what this means is that in this situation a lot of time can be spent with your focus away from driving without risking any harmful consequences.  With a little discipline a cell phone conversation can safely take place under these conditions.

And also looked at this way we can see why the second scenario is hard.  This scenario involves a much higher potential decision rate.  And that's really the same as a high decision rate.  Deciding there is nothing that needs to be done right now is a special kind of decision.  And there are a lot of moving pieces to monitor.  The car in front is not far away (there is a car in front because there is a lot of traffic and it's not far away also because there is a lot of traffic) so it needs to be carefully and continuously monitored.  If it's a multilane road then cars in the other lanes need to be monitored for potential lane changes.  Blind spots need to be monitored for the unexpected.  And if something changes a decision needs to be made and acted upon quickly.  And it is possible, even probable, that several things will change at approximately the same time.  So the decision may not be a simple "slow down"/"speed up" decision.  Perhaps a swerve needs to also be thrown in so it's complex.

In the first situation if we are switching our focus from driving to the cell phone for short periods of time we will probably still be ok.  Something may have changed while our attention was away.  But the change will be simple and we should still have plenty of time in which to decide what to do and to do it.  In the second scenario the chances that something will go wrong while our focus is elsewhere is much greater.  In this environment the time and effort to task switch away from something else and back to driving is a luxury we can't afford.

I think that if we are being honest all of us would agree that talking on the phone while driving is a bad idea.  But a lot of us think we can get away with because we are good at task switching and we will be careful and only do in at "appropriate" (situation 1) times.  But people usually give themselves too much credit in these situations.  Let's move on.

The "fix" to the above situation is to only permit talking on the phone while driving only if we are using a "hands free" device.  There is something to this but not much.  The hand's free device allows us to keep our hands on the wheel and our eyes on the road.  This is an improvement but not much of one.  The decision to allow hands free cell phone use was a political one that was based on no scientific research.  The broad  availability of blue tooth hands free devices meant that many cell phone users had already gone hands free.  The industry could see sales of hands free accessories increasing so they decided to go along.  But the benefits of hands free are small.

The problem that hands free does not solve is the focus problem.  Where is your attention focused?  Theoretically hands free makes switching focus back to the road quicker.  But lots of people drive one handed so the fact that you are using one hand to hold the phone to your ear rather than doing something else with it doesn't really change things.  And it is easy to talk into a hand held phone while keeping your eyes on the road.  So there is no real difference during the conversation part of cell phone use.  Hands free plus voice activation does help with dialing and hanging up but that's only a small portion of a typical call.  In the end there is very little difference between hands on and hands free.

The discussion of talking while driving started when cell phones became common.  But a couple of generations of phones later a new threat arose:  texting while driving.  You pretty much need to look at the device and use both hands to text.  This removes your focus, your hands, and your eyes from where they are supposed to be.  It also means that the length of time you spend with your focus switched away from driving is much longer.  It require more of your brain to text than it does to talk and it takes longer to type a phrase than say it even if you are using all the cute texting shortcuts.  That means there is a much longer continuous interval where you are not monitoring the driving environment.  This is worse, much worse, and the statistics bear this out.  Texting while driving is way more dangerous than talking while driving.  And unfortunately there is a significant population that thinks they can get away with doing it anyhow.  They are a danger to us all.

So far I have mostly covered ground that everyone is familiar with.  Okay, I might have thrown is some computer stuff that is unfamiliar to most of us.  But for the most part people have gotten to the same conclusion I have by one means or another.  So where's the original content?  Coming right up.

I bought a new car about six months ago.  I purposely got a car with all the new "electronic assist" goodies.  I wanted a backup camera.  I wanted a blind spot monitor.  My car came with those and lots of other goodies.  If the road has decent lane markers my car will warn me if I start to drift across the lane markings without first putting on my turn signal.  It also has a alert that warns me that I might not have noticed the car in front of me slowing down.  It has another alert that warns me that the car in front of me has started moving and I haven't.  It has a bunch of more alerts too but I think you get the idea.

My old car was a 'mid-price 99.  It had some gadgets on it but nothing like what my new car has.  You could set my old car so the head lights stayed on for a while after you exited the car.  It had a compass built into the rear view mirror and cruise control.  That was pretty fancy for the time.  But my new car leaves all those old gadgets in the dust.  My new car has an "automatic" setting on the headlights.  They turn on and off automatically based on how much daylight  there is.  And, of course, it automatically delays shutting the headlights off when you exit the car.  My new car has a deluxe cruise control system and a full up navigation system to compliment the compass in the rear view mirror.  I'm not trying to brag here.  There are lots of other cars that come equipped with a similar (or perhaps an even more extensive) set of gadgets.  I am just pointing out that my new car has gadget after gadget after gadget.  And it's not just the sheer number of gadgets.  Each individual gadget is much more complex.

Let me give you an example.  I now leave my headlights set on  "automatic".  But once the setting got changed without my noticing it.  So now I'm driving around in the dark without my headlights on.  After a while I figured that out.  But now the headlight control is surrounded by a bunch of other controls.  So there is no way I am going to be able to fix the problem and drive at the same time.  Once I got where I was going I turned the cabin light on and spent about thirty seconds getting the setting fixed.  But while I was doing this I accidently turned the high beams on.  Again it took me a while to figure this out.  But that led to the "how do I turn the high beams off" problem.  I am so used to everything on my new car working differently than it did on my old car that the obvious solution of doing the same thing I would have done on my old car literally did not occur to me.  Someone had to suggest it to me and you'll be happy to know it worked fine.

I want to make two points.  First of all my new car is regularly alerting me to something or another.  Some of these are wonderful.  It will alert me to cross traffic when I am backing out of a parking slot.  With my old car I often couldn't see a thing until I was in the middle of the street.  By the time I was far enough out of the slot to see it was too late.  The crash would have already happened.  With my new car I get an alert, frequently before I have even started backing up.  The alert tells me when the traffic has finished passing by and I can back out safely.  So that alert is all to the good.

With the other alerts some of them are more of a mixed bag.  With the stopping alert I am most likely on top of what is going on.  I just want to start stopping a little later than the car does.  The car, for good reason, is quite conservative about when it thinks I should start to slow down.  Remember, it needs time to warn me, decide I am not going to heed its warning, and after that still have time to slow the car down on its own.  With the starting alert what's usually happening is I can see something (i.e. a pedestrian) that the car doesn't so there is a good reason why I haven't immediately imitated the car in front of me.  And so on.  The point is that my car now fairly frequently makes some noise that breaks my concentration and thus interferes with my focus.  So that's one thing.

The other thing is illustrated by the head light story.  Things that used to be simple to do are now often much more complicated.  A couple of weeks ago I was using the navigation system to get me home from an unfamiliar location.  The trip involved two parts.  The first part was "get from the starting point to the freeway".  For that part the navigation system was invaluable.  It did a great job of navigating me along a complicated path on unfamiliar streets.  But once I was on the freeway I no longer needed or wanted the navigation system.  All I wanted to do was turn it off.  But that is not a simple process.

It involved using a touch screen.  That involved taking my eyes (and one hand) away from their driving tasks and working my way through the process, all while driving in such a way that I did not run off the road or crash into anybody.  It turned out that what I did didn't turn it completely off.  I just put it to sleep.  So twenty minutes later it woke up and started alerting me to the fact that I should not take an exit that I already knew not to take.  So I had to fiddle with it some more to get it actually turned off.  And again it was important to not drive off the road nor crash into anybody.

I'm sure I will get better at this sort of thing as time goes by and I get more familiar with all these new gadgets.  And I'm sure that some of my fellow drivers thought I was nuts while I was distracted dealing with my navigation system.  But then we all see people doing nutty things on the road all the time now.  Why?  Because they are on the phone or texting or whatever.  In any case no one thought what I was doing merited a honk so I guess my behavior fell into the range of what now passes for relatively normal driving.

So my new contribution to the subject is to be the first to identify a new source of distraction that drivers can be subjected to:  new cars with a lot of electronic gadgets.  With my old car I was pretty much on my own.  I knew it and acted accordingly.  I didn't expect the car to be of much help but on the other hand the car did not routinely engage in distracting behavior.  And you could do simple things like turn the headlights on or change the channel on the radio without having to take your focus off of driving for longer than a brief moment.

With new cars things are quite different.  The car will routinely engage in distracting behavior.  Sometimes this is a totally good thing.  I love being distracted by a cross traffic alert when I am backing out of a parking stall.  In other cases the advantage is less clear cut.  Most but not all of the time the car is alerting me to a situation I am already aware of.  If I am already aware of it then it is a needless distraction that did not exist before the advent of modern "driver assist" electronics.  And I can't get a car with the long list of goodies my new car has without also getting a car that has much more complicated controls than cars like my old one.

And its not just me.  A few months ago Tesla put out an update to the software on their model S.  The update provided an extremely sophisticated cruise control.  You could take your hands completely off the wheel for long periods of time.  The car would automatically compensate for traffic conditions.  It would read the speed limit signs along the side of the road and act accordingly.  It would even change lanes if that seemed appropriate.  Way cool, right?  Apparently too cool.  Almost immediately Tesla released a new download that dialed way back on how much the car would do on its own.  Apparently It wasn't quite as ready for prime time as Tesla thought.

Finally, let me fold a "robot car" update in here.  Back when my old car was new we had a simple situation.  The driver drove the car with little or nothing in the way of intelligent assistance from the car.  That put the responsibility squarely on the shoulders of the driver.  Once we get to robot cars we will again be in a simple situation.  Then the driving responsibility will be squarely on the shoulders of the car.  Where we are now is some kind of a middle zone.  Driving responsibility rests primarily with the driver.  But the car is making a significant contribution.  This makes the situation more complicated than either pure extreme.  And, as the Tesla experience demonstrates, navigating this middle ground is not going to be 100% smooth sailing.

And the general public has already picked up on this.  Surveys show that a large majority of people are not ready for a completely robotic car yet.  People definitely want a person to be able to take control back from the "robot" any time they feel the need.  That means it is going to be a while before there are cars on the market that do not include a steering wheel, brake pedal, etc..  As far as I can determine the Tesla problem resulted in a few scares but no actual accidents.  Tesla rightly dialed things back as soon as problems became apparent.  And I expect that is how things are going to continue to evolve for some time.

Development of robot cars is proceeding apace.  There are now multiple well funded groups with access to deep pools of technical expertise actively working on the problem.  It won't be long before someone finds a way to get robot cars into the hands of ordinary consumers.  And this will happen in the very near future (2020 or sooner).  Then consumers will be in a position to make an informed decision.  From Frankenstein to Jurassic World, all of us have long been exposed to "technology gone horribly wrong" themed movies.  The public is very attuned to the possibility that robot cars will somehow go horribly wrong.  So people want to take it slow for now, and with good reason.  But once ordinary consumers get their hands on these cars they will either work well in the real world or they won't.  I think before long they will work and work well.  At that point the situation will switch from movie plot to real world experience.  At that point I expect public to quickly become comfortable sharing the road with robot cars.  Time will tell.