Tag Archives: psychology

Unskilled and unaware of it (re-post)

It was pointed out to me that I haven’t posted anything for weeks. It will be a few more before the project that has been taking up my time is out of the way. Here, then, is something I wrote way back in November 2010, which seemed relevant given the latest Department for Transport “please play nice on the roads” marketing campaign. If I were writing it today, I’d mention a million other things, and probably do away with the sarcastic transport mode tribalism that amused me so when this blog was young. But I’m not.

In the War Bulletin this week I mentioned a study that found drivers to be at fault in 87% of car/bicycle collisions.  According to the press release and coverage, the study included (but was not limited to) giving cyclists in Melbourne helmetcams, and analysing the footage of 54 “events”, including 2 collisions.  It sounds like the study has a number of limitations — it’s difficult to draw general conclusions about collisions from only 2 of them, and the results were only ever going to apply to the helmet-law and vehicular-cycling environment of Melbourne, and even then only to experienced cyclists who (presumably) were aware that that their own behaviour was being recorded.

The study was conducted by the Monash University Accident Research Centre, who I am sure did a good job.  But unfortunately nothing resembling it appears in their reports and publications, and I can find no evidence that the original research has been made public yet.  (Allowing the world’s media to uncritically churn your press release without being able to see the actual details of the work — and perhaps more importantly, before your fellow academics are allowed to review what you have done — is rather bad form.)  So there’s not really anything more we can say until we can see the study itself, and we may yet find that everything that has been said was wrong.

But the reported findings do fit with what we already know about accident causes and driver behaviour.

The Motorist attitude to their own collisions and near-collisions is a particularly interesting field.  When one suggests that speed cameras might be a good thing, for example, somebody will always pop up to declare that they have been driving at 90mph for decades and never caused a single accident, because they are a perfect driver who knows exactly when speed is appropriate. And it might be true: some people are good drivers and some people are bad drivers.  Trouble is, the driver himself can never know which he is: all drivers believe themselves to be above average.  Everybody is seeing bad driving, but nobody admits to doing it.

In Traffic, Tom Vanderbilt documents the details of the phenomenon of drivers unable to recognise their own lack of skill.  A large part of it he puts down to a lack of feedback.  For example, in the Monash helmetcam study, there were a mere 2 collisions, but there were 6 near-collisions and 46 “other incidents” (the classic Heinrich triangle).  These “other incidents” correspond to those situations where we notice people driving badly.  They occur because the driver failed to spot a hazard or failed to recognise as a hazard something that they did see.  By definition, if they did not see or did not recognise, the driver will never have been aware of the situation.  They will reach their destination assuming that they had done a great job, oblivious to the bad driving that had been recorded.  That’s probably what happened in 52 out of the Monash group’s 54 “events”.

And when the driver does finally notice that they have just been in a near collision, they can congratulate themselves for having the skill to have avoided an actual collision.

Thus reassured of their own driving skills, on the few occasions when they do get some feedback, they find ways to dismiss it.  That horn honk wasn’t aimed at me, or if it was, it must be because the other driver is an impatient egotistical bad driver who wouldn’t recognise good driving like mine.  The police pulled me over because they have a quota to fill, or because they’re anti-Motorist, not because I was driving dangerously.  After all, I already know that I am not a dangerous driver.

And then they crash, and it was an accident, bad luck, a momentary loss of concentration, beyond one’s control.  They couldn’t have caused it, because they already know from their experience and their long record of not causing accidents that they must be a good driver.

The evidence from driving simulation experiments shows that drivers can’t accurately remember what was happening in the lead up to the crash — what they saw and heard, who else was on the road and where and which order and when they appeared; what they were thinking and where they were looking and when they last checked their mirrors.  So they can unconsciously fill in these details with whatever makes them feel the least uncomfortable.

When drivers are shown videos of their driving (from helmetcams, or, as Vanderbilt discusses, Drivecam), most of them are surprised to discover that they have many more bad habits than they were aware of.  And that can create some uncomfortable cognitive dissonance for them, with attempts to deny or justify their behaviour, or, as with speeding, attempts to redefine it as safe.

It’s important to know these things about driver psychology if you’re trying to create a marketing campaign to make drivers be nice, or design ways to rehabilitate careless and dangerous drivers (how does sending a dangerous driver on their way with a £60 fine help anybody when the driver doesn’t have the skills to figure out what they are doing wrong?), or wondering whether to send your helmetcam footage to Roadsafe to be passed on to the offending driver.

And it’s important to know these things about driver psychology when deciding whether motor vehicles can ever share nicely with vulnerable road users.

Risk compensation and bicycle helmets

Some months ago I left a series on bicycle helmets hanging while I got distracted with other things. We had looked at what the best evidence for the efficacy of helmets in preventing injury in the event of a crash is, and some of the reasons why we should be cautious about that evidence. We found that if you’re unlucky enough to have been hospitalised while riding a bicycle, you’re less likely to be there with a head or brain injury if you were wearing a helmet at the time of the crash. We noted several ways in which this protective effect is exaggerated and used to mislead, we noted that reduction in injury is from a very low level anyway, and that the research so far done fails to provide any sub-analysis of very different riding styles, such as racing cyclists, mountain bikers, and utility cyclists.

We also made careful note of the fact that a reduction in the rate of head injury in the event of a crash is a different finding to a reduction in the rate of injury and death of bicyclists. We briefly began the exploration of what this means by considering the fact that helmets are not much defence against a motor vehicle.

How could a reduction in head injury in cyclists who crash not mean a reduction in injury and death in bicyclists? Well, helmets could be causing other kinds of injury in crashes. Or they could be causing crashes — particularly the worst kinds of crashes.

The latter is a particularly interesting avenue. The idea is risk compensation or risk homeostasis, a phenomenon documented in fine detail by John Adams in the 1985 book Risk and Freedom. Adams showed that advances in road safety — seatbelts, motorcycle helmets, safer vehicle designs and wider, straighter, safer road designs — are never followed by quite the cut in injuries and deaths that is predicted; that while road “safety” has improved crashes are no less frequent, and that bystanders — pedestrians and cyclists — are butchered at an ever increasing rate.  There is a set level of danger that people are willing to tolerate, and so motorists were compensating for the new safety features by driving faster and taking more risks. To put it in Adams’s technical terms, potential “safety benefits” were instead absorbed as “performance benefits”.

James Hedlund reviewed the evidence on risk compensation and came up with a set of rules for when people are likely to compensate for a safety intervention:

  1. They know it’s there.
  2. They know it’s a safety feature.
  3. There is a potential performance benefit to be had.
  4. There is freedom to realise that performance benefit.

Well cyclists know whether or not they’re wearing a helmet, they know that helmets are meant for safety, there are potential performance benefits — riding faster, through smaller gaps, in more hostile traffic, or with less caution in conditions that would otherwise advise it — and cyclists are generally free to ride more furiously if they want to. (Indeed, you may be wanting to cycle faster, in which case go ahead and use a safety feature as a performance benefit if that works for you.)

But that’s only a hypothetical reason to expect risk compensation by cyclists wearing helmets, not evidence that it actually happens. And very little effort seems to have been put into researching that — perhaps because it’s difficult to devise a properly controlled test. A study of cyclists in Spain attempted to test the idea by comparing the rate of helmet wearing in traffic law violators to the rate in non-violators, finding that law breakers were less likely to be helmet wearers, the opposite to what they say should be expected if there is risk compensation. However, this study could not control for all possible differences between the populations (“confounding variables”) — for example, helmet wearers are already a population of safety-conscious conformists, less likely to commit traffic violations, and so a better question to ask would be whether those helmet wearers acted even more cautiously when their helmets were taken away from them, and whether the non-wearers behaved even more recklessly when given a helmet. (This study is, embarrassingly, the British Medical Association’s sole reference for their dismissal of risk compensation.) A more recent study observed a set of participants behaviour with and without a helmet, using speed as an indicator of risk taking and heart rate variability as a proxy for risk perception. This study found that when helmet users had their helmet taken away, the risk taking (i.e. speed) reduced to keep the risk perception stable. However, the study only looked at 35 people, and only looked at proxy variables. Neither study is very convincing — the limitations I describe here are just the tips of the icebergs — and certainly nowhere near strong enough or specific enough to guide policy. We still have a mere plausible hypothesis with no good evidence as to whether or not it’s true.

The authors of the Cochrane review acknowledge the suggestion that risk compensation by cyclists could affect their crash rate, but believe that is unlikely. It’s interesting to see a hypothesis dismissed with the argument from personal incredulity in a Cochrane review.

What is not touched on in the review, and which is potentially far more important (given the fact that crashes with motor vehicles are more likely to kill or seriously injure), is the risk compensation effect not of cyclists themselves but of the other road users around them — i.e., of the motorists. Look again at Hedlund’s rules. Motorists can see whether a cyclist is wearing a helmet; they know that helmets are supposed to be a safety feature; they can potentially find performance benefits — they think they can squeeze through tighter gaps when overtaking against oncoming traffic, or pass more quickly, or shoot in front while turning, because if they hit the cyclist then no harm is done; and there is nothing to stop them realising that performance benefit, since the police, if there even are any, are rarely even aware of the relevant traffic rules, let alone bothered with enforcing them. There is therefore a plausible hypothesis that motorists will take more risks around cyclists who wear helmets than around cyclists who do not.

This hypothesis is made all the more plausible by the fact that, in addition to potentially making cyclists seem less vulnerable, helmets make cyclists look more competent: in surveys of motorists’ beliefs, most assume that cyclists who wear helmets are more experienced and more “responsible“, meaning that they may be driving more carefully around non-helmeted cyclists who they expect to do something silly. And motorists overwhelmingly think that cyclists should be forced to wear helmets — presumably so that the motorists can get the performance benefits of driving more dangerously around them.

The motorist risk compensation theory has famously been tested by @IanWalker in one of the most delightful experiments in the field. Walker rode around Salisbury and Bristol on a bicycle fitted with an ultrasonic distance sensor measuring the effect of a number of variables on passing distance, including rider position in road, type of motor vehicle, and whether he was wearing a helmet. Analysis of over 2,000 passes showed that motorists tended to give on average around 5-10 cm less space when the rider wore a helmet. It’s not much difference, and the effect of motor vehicle type, perceived rider gender, and rider’s distance from the edge of the road were all stronger.

But it’s important to note that there is always a distribution of passing distances — a bell curve. There are a few motorists who give a lot of room, a few who scrape past, and a lot clustered in the middle, giving a little over a metre distance. When wearing a helmet, the bell curve shifts in a little bit. The cautious drivers give a little less space, the average drivers give a little less space, and the dangerous drivers give a little less space.  It’s the latter who are now more likely to drive into you.

Walker’s research, delightful as it is, is itself not without limitations. Most important amongst them is that, when it comes to answering questions of cyclist safety, it suffers the same limitation of measuring only proxy variables: passing distances rather than actual risk of crashes and injuries. But it tells us that there is a very important reason to study more than just the isolated risk of head and brain injury in the event of a crash.

Helmets are a medical intervention, exactly like a drug or surgical procedure. They are a preventative intervention and they are a physical intervention, but neither of those are alien to medicine and to the modern methods of evidence-based medical science. And risk compensation is just a side-effect of this medical intervention, like the side-effects of drugs. The side-effects of drugs that make it to market are by definition outweighed by the beneficial effects; but ten times as many drugs are discarded during development because the research finds that either the side-effects are so big or the beneficial effects are so small that the harm outweighs the help.

The authors of the Cochrane review defend their dismissal of risk compensation by saying “the fundamental issue is whether or not when bicycle riders crash and hit their heads they are benefited by wearing a helmet.” And that’s fine if you’re in the preliminary stages of developing an intervention and you are so far only concerned with whether it has beneficial effects. But the authors go far beyond that early stage in their conclusions, recommending that this intervention be compulsory — despite there being very good reasons to suspect that there are potentially major side-effects of this intervention. They can’t have it both ways. If you haven’t bothered studying the side-effects you can’t license the drug. It might kill people.

Queuing

I’ve been meaning to write a bit more about the M4 bus lane, but haven’t had the time, so here’s a rather crude brain dump while I sit in the dark on a bus somewhere on the A9 in the snowy Cairngorms.

The Dutch infrastructure minister recently announced that speed limits on some stretches of motorway would be raised.  This would not help Motorists get to their destinations any quicker, she noted, but it was a change worth making because it would make the motorists feel better.  Aside from being a delightfully refreshing piece of honesty from a politician, it highlights again that behaviour and psychology should not be ignored when designing transport policy.  Philip Hammond should have been this honest when abolishing the M4 bus lane — instead of the weak nonsense about improving journey times, just tell the truth that it’s a cheap way of making Motorists feel better.

The M4 bus lane was designed to cut the journey times of Motorists entering London — to make their journeys faster and more reliable, and thus to cut the <insert absurd made up number here> billions of pounds that the Institute of Directors like to claim is lost because of their Jags being caught in congestion their supply chain being delayed by congestion.  The Motorist probably thinks that he too would like his journeys to be faster and more reliable.  But this is not quite true.  The Motorist would like his journeys to seem faster and more reliable.

The M4 bus lane was hated not because it increased car journey times or made journeys less reliable.  It didn’t.  As previously explained, the bus lane was a clever hack to the layout of a road with a bottleneck.  It made a tiny and irrelevant cut to journey times, while cutting lane changing and accident rates and thus greatly improving consistency in journey times.  The bus lane was hated because motorists thought it increased their journey times.

Part of it was the problem of common sense.  The likes of Jeremy Clarkson and Terry Wogan despise those scientists and academics with all their fancy facts and data — the problem with these researchers is that they don’t have any common sense, and common sense tells Clarkson and Wogan that taking away one lane of the M4 must have caused traffic jams.  No amount of your facts can change that.

Another part of it was recall bias: all of those massive pre-bus lane jams begin to blur into the distance, whereas this jam that I’m sat in right now is real — and hey look, there’s a bus lane.  Coincidence?

But it was more than this.  It was about people’s perception, and particularly people’s perception of queues.  Since I’m on a bus with no reference material and limited battery life, I’ll put it in bullet points:

When sat on a Motorway in a traffic jam, Motorists usually believe that their own lane is going the slowest.  It’s simple: when their own lane is moving freely, they’re concentrating on driving, and don’t notice that the other lanes are stationary; when their lane is stationary, they have nothing better to do than stare at all the vehicles which are moving freely in the other lanes.  So even if over time all lanes even out, the Motorist perceives that the other lanes are moving better — especially if the jam is severe enough that they spend more time stationary (observing others moving) than moving themselves.   (Hence all the futile changing of lanes in jams, which just makes the jams worse.)  This is the same reason why in the Post Office — wait, do blog readers even still use those?  OK, this is the same reason why in the ticket office at a major station, you have a single queue serving several windows, rather than independent queues.  Independent queues make people nervous about their decisions.

This perception leads to Motorists overestimating their time sat in traffic, and it’s made worse when they can see moving traffic — if the opposite carriageway is moving freely, or there’s a parallel un-jammed road, then the sight of moving cars merely serves to remind the poor stationary Motorist of their own lack of motion.  Drivers asked to estimate how long they were stuck in traffic consistently over-estimate the jam if they see other traffic moving freely.

So the M4 bus lane was about the worst thing you could do if you wanted Motorists to perceive that they were spending less time in queues.  Now when they were sat in a queue they weren’t just sat there with nothing better to do than get paranoid over the relative speed of the two lanes of traffic: they could also sit there watching the buses and taxis and prime-ministers go past at speed, constantly highlighting the fact that the Motorist was going nowhere.

The research shows this — have drivers estimate their queuing time with and without visible moving traffic nearby; or compare the driver and passenger experience of a stop-start motorway jam. It’s just another of the many fascinating little quirks of psychology — one of the bizarre things our brains do when confronted with absurd man-made situations like traffic jams.  You can make Motorists happily spend more time sat in traffic jams, simply by making them sincerely believe that it is less time.

(Somewhen I’ll try to find some interesting references, but 3G has just dropped out…)

Engineering, psychology, and a bus on stilts

This week I’m trying to clear up the loose ends of threads I began and never finished, and get rid of some of the draft posts that I started but never polished…

Last week I posted about tracked hovercraft and straddling buses — a tongue-in-cheek look at how through the ages engineers have proposed ever more overcomplicated engineering solutions in an attempt to manage our out-of-control transport problems.  I assumed that my learned readers would get the point without labour.  WordPress.com very kindly picked it as one of their daily front-page features, though, which led to it receiving around 4,000 spam comments, including several dozen along the lines of “wow that bus looks awsum and wood solve all our problems make one for america!!!?! (p.s. here’s a link to my blog!!!11!)”.*

Well, actually my guess is that the straddling bus will be just another absurd transport solution that fails to achieve the things that it is designed to achieve.  The stated purpose of the bus is not to get cars out of its way, it is to get the bus out of cars’ way: the designers complain that the frequent stop-starting of buses means that they hold up the traffic behind.  It will probably fail to achieve much in the way of making car drivers’ lives easier because the designers are obsessed with engineering and don’t consider Motorist behaviour.

Here are a couple of random fascinating psychology factoids.  I wonder to what extent the bus backers have considered them in their models?

  • When you make road lanes just a little but wider — as you will surely need to do if you are to accommodate the bus safely — people drive faster.  They’re not doing it deliberately or rationally, perhaps not even consciously, they just do it.  It feels right.
  • Drivers slow down for tunnels, and things that feel like tunnels — tree-lined avenues and close high walls.  Even if there’s nothing telling them to, and no rational safety reason to do so.  They just do it.

The cause of traffic jams is traffic.  Too much of if, behaving erratically.  We like to pretend that it’s bad engineering, because we can always fix engineering by replacing it with some different engineering.  And we like to pretend that it’s not the volume of traffic and the behaviour of drivers, because acknowledging this would mean giving up hope that one day the traffic jams will magically be solved.  But that’s the way it is: too many cars, too badly driven.  The straddling bus will probably not help congestion — at least, no more than a conventional bus on a conventional bus lane — because it will change driver behaviour in a way we can’t easily predict, but which (as described) will likely involve them slowing down and speeding up in chaotic waves as the bus passes them and they pass the bus.  It doesn’t sound like much, but these effects have a habit of amplifying themselves: the traffic between lanes will cease to be smooth, so cars will be changing lanes more, and this lane-changing contributes further to slowing things down, and also greatly raises the risks of accidents occurring.

Perhaps that effect will be marginal given all of the other existing complications and currents in the traffic flow.  Perhaps we’ll see other interesting unforeseen behaviour changes in the Shenzhen trial.  All that we can say for sure that everybody will be predictably surprised when drivers don’t behave in a simple rational manner.  Just like they were the last ten thousand times the solution to congestion was discovered.

The main reason the bus will fail, though, is the same reason that all urban roadspace provision schemes fail: create a new space for cars to drive in, and an equal or greater quantity of car journeys will be created to fill that space.  The cause of traffic jams is too much traffic.  Double the capacity for traffic and all you’re doing is doubling the size of the traffic jams.

Put a conventional bus on a conventional (parking and taxi enforced) bus lane.  It’s easier.

* Not that I’m not grateful for all your valuable contributions to our discussions ;)

Unskilled and unaware of it

In the War Bulletin this week I mentioned a study that found drivers to be at fault in 87% of car/bicycle collisions.  According to the press release and coverage, the study included (but was not limited to) giving cyclists in Melbourne helmetcams, and analyzing the footage of 54 “events”, including 2 collisions.  It sounds like the study has a number of limitations — it’s difficult to draw general conclusions about collisions from only 2 of them, and the results were only ever going to apply to the helmet-law and vehicular-cycling environment of Melbourne, and even then only to experienced cyclists who (presumably) were aware that that their own behaviour was being recorded.

The study was conducted by the Monash University Accident Research Centre, who I am sure did a good job.  But unfortunately nothing resembling it appears in their reports and publications, and I can find no evidence that the original research has been made public yet.  (Allowing the world’s media to uncritically churn your press release without being able to see the actual details of the work — and perhaps more importantly, before your fellow academics are allowed to review what you have done — is rather bad form.)  So there’s not really anything more we can say until we can see the study itself, and we may yet find that everything that has been said was wrong.

But the reported findings do fit with what we already know about accident causes and driver behaviour.

The Motorist attitude to their own collisions and near-collisions is a particularly interesting field.  When one suggests that speed cameras might be a good thing, for example, somebody will always pop up to declare that they have been driving at 90mph for decades and never caused a single accident, because they are a perfect driver who knows exactly when speed is appropriate. And it might be true: some people are good drivers and some people are bad drivers.  Trouble is, the driver himself can never know which he is: all drivers believe themselves to be above average.  Everybody is seeing bad driving, but nobody admits to doing it.

In Traffic, Tom Vanderbilt documents the details of the phenomenon of drivers unable to recognise their own lack of skill.  A large part of it he puts down to a lack of feedback.  For example, in the Monash helmetcam study, there were a mere 2 collisions, but there were 6 near-collisions and 46 “other incidents” (the classic Heinrich triangle).  These “other incidents” correspond to those situations where we notice people driving badly.  They occur because the driver failed to spot a hazard or failed to recognise as a hazard something that they did see.  By definition, if they did not see or did not recognise, the driver will never have been aware of the situation.  They will reach their destination assuming that they had done a great job, oblivious to the bad driving that had been recorded.  That’s probably what happened in 52 out of the Monash group’s 54 “events”.

And when the driver does finally notice that they have just been in a near collision, they can congratulate themselves for having the skill to have avoided an actual collision.

Thus reassured of their own driving skills, on the few occasions when they do get some feedback, they find ways to dismiss it.  That horn honk wasn’t aimed at me, or if it was, it must be because the other driver is an impatient egotistical bad driver who wouldn’t recognise good driving like mine.  The police pulled me over because they have a quota to fill, or because they’re anti-Motorist, not because I was driving dangerously.  After all, I already know that I am not a dangerous driver.

And then they crash, and it was an accident, bad luck, a momentary loss of concentration, beyond one’s control.  They couldn’t have caused it, because they already know from their experience and their long record of not causing accidents that they must be a good driver.

The evidence from driving simulation experiments shows that drivers can’t accurately remember what was happening in the lead up to the crash — what they saw and heard, who else was on the road and where and which order and when they appeared; what they were thinking and where they were looking and when they last checked their mirrors.  So they can unconsciously fill in these details with whatever makes them feel the least uncomfortable.

When drivers are shown videos of their driving (from helmetcams, or as Vanderbilt discusses, Drivecam), most of them are surprised to discover that they have many more bad habits than they were aware of.  And that can create some uncomfortable cognitive dissonance for them, with attempts to deny or justify their behaviour, or, as with speeding, attempts to redefine it as safe.

It’s important to know these things about driver psychology if you’re trying to create a marketing campaign to make drivers be nice, or design ways to rehabilitate careless and dangerous drivers (how does sending a dangerous driver on their way with a £60 fine help anybody when the driver doesn’t have the skills to figure out what they are doing wrong?), or wondering whether to send your helmetcam footage to Roadsafe to be passed on to the offending driver.

And it’s important to know these things about driver psychology when deciding whether motor vehicles can ever share nicely with vulnerable road users.

A momentary attraction

Vaughan of the ever fascinating Mind Hacks discusses a recent study that looks at the effects of having an attractive passenger in the car on driver errors.  The researchers put forty people in driving simulators and made them drive through hazards while talking to passengers.  The basic finding was that the drivers failed to spot pedestrians and hit more motorbikes when talking to those passengers that they considered to be more attractive.  Driving with somebody you fancy is dangerous.

More interesting, though, is the detail.  The drivers who were distracted by their hot companions were more anxious and drove slower; their eyes continued to look in all the right directions.  The errors that they made were “looked but failed to see” errors: they were looking at the road and mirrors as usual, but not processing the information.  I imagine they thought that they were driving well, at appropriate speed, and making all of the checks that they needed to make…

Link to PubMed entry for study; link to Vaughan’s post.