Simulations predicting behaviour in traffic assume we are rational. But it appears we are not. Guy Walker discusses how our attachment to queuing etiquette causes huge delays in traffic at road works.
I don’t think I’ve ever been involved in such a controversial ergonomics topic. What could it possibly be? Something hard-hitting like nuclear power? Healthcare? No. Roadworks. A little while ago some research of mine appeared in the press under the title ‘Polite drivers cause queues’, and within nanoseconds I was receiving emails and phone calls from motorists. Some were reacting positively to the story, some rather less so. I think the phrase “waste of tax-payers’ money” might even have been used. So let me explain, because behind the headlines and sound bites is a genuinely fascinating ergonomics story.
Imagine you’re driving along the motorway, with three lanes of emptiness ahead of you. Then you see signs warning of roadworks and lane closures. As the traffic thickens and the point arrives where the closed lanes have to merge, what happens? Does everyone make maximum use of the available road space and allow others to merge at the head of the line with a friendly wave and a spirit of mutual cooperation? That’s what traffic theory would predict. It would say the lanes would become fully loaded, even at the point of merging, because each motorist is looking to maximise their utility in terms of time savings and other cost variables. Does this sound like real life though? I don’t know about you, but drivers in my locale head for the nearside and open lane as soon as humanly possible, and there seems to be a critical distance from the merge where people decide not to let other motorists in. They’re pushing in aren’t they? The British love a good queue and there are strict rules to follow.
What we have here is a classic case of emergence, where small, low-level behaviours magnify to truly daunting levels, giving rise to counter-intuitive large-scale effects out of all proportion to what’s causing them. Closing the gap and preventing someone from merging, or heading for the nearside lane as soon as possible might seem pretty inconsequential but we can lose up to a lane of capacity – that’s in addition to the lanes we’ve already closed – purely due to queueing behaviours like these. On the subject of wasting tax-payers’ money, bear in mind that it costs something like £3 million per mile to put those lanes there, and a further £43 thousand per year on average to maintain them. Combine these figures with the fact that we already have an acute capacity challenge, and you can see why we can no longer ignore the problem. And it is an ergonomics problem. Here we have an engineered environment where we expect people to behave in certain ways, except they don’t. Sound familiar?
An incredible amount of effort goes into planning roadworks. A cutting-edge approach employed up here in Scotland, and elsewhere, is something called traffic microsimulation, where a faithful virtual model of the road network is populated with lots of little virtual cars, which in turn follow some fairly simple rules. When the simulation is set running, what emerges is an incredibly lifelike representation of how real traffic will behave. This simulation is quite a break from how transportation engineering used to do things. In the past you would use ‘macroscopic’ models built on origin/destination matrices and some fairly formidable algorithms to calculate coarse-grain flow rates and speeds and so forth. The benefit of using traffic microsimulation is that you can look at individual vehicles, in very small time frames, giving you very fine-grained analyses of traffic conditions.
Because microsimulations can deliver such a detailed analysis, Transport Scotland commissioned the Clyde Strategic Microsimulation Model, a model of the entire metropolitan Glasgow area: 250km of road, 50 grade-separated junctions and no fewer than 1.5 million simulated vehicles. The purpose of the model, in use since 2010, is to help design and test roadworks for the trunk and motorway maintenance programme in central Scotland. Obviously, the model has to be accurate as it will help determine when, where and how many roadworks will be performed as well as informing the travel advice given to the millions of motorists who use the network. As luck would have it, the opportunity arose to test the model with some real-life roadworks on the M8 motorway to the west of Glasgow, to see how real and virtual drivers compared.
Working with SIAS Ltd, an Edinburgh-based company who developed the Clyde Strategic Microsimulation Model, we were able to make use of automatic traffic count data from a range of recording stations around the roadworks site over a three-year period. With several million data points we can be fairly confident in what we were observing, which was a marked discrepancy in what the microsimulation model was predicting in terms of traffic flow and what was actually happening. For some of the most critical rush hour flows we were losing a substantial amount of capacity because of the way people queued and, moreover, the microsimulation model wasn’t capturing it.
The capacity reductions due to driver queuing behaviours were substantial. The average capacity reduction was 22% or something in the order of 1500 vehicles per hour per lane, which is not too dissimilar to the amount of traffic a normal motorway lane would carry. Remember, this is a loss of a lane of capacity in addition to the ones that have been closed.
It got worse, because the capacity reductions were hitting peaks as high as 55%, and it had nothing to do with hard engineering as such, it was due entirely to seemingly irrational driver behaviour. In some follow-up work in other locations it was clear some very powerful social norms were in play, which affected how people wanted to queue. Norms that were far more powerful than all the logical, rational transport engineering advice provided to motorists in the form of signs saying “merge in turn”, “stay in lane” and “maintain a safe distance”.
To some extent we have been here before. We’ve known for a while that we can get bigger than expected reductions in capacity when you reduce lanes, and of course, there are the famous Minnesota studies all about late and early merging. The key thing, though, is that while we know something about the ‘what’ we really don’t know much about the ‘why’. With traffic microsimulation, which focuses on individual vehicles and so-called ‘agent behaviour’, we need to find out what, exactly, is happening.
Driving is clearly a social activity performed in close proximity to others. In motorway driving, our other research keeps showing how important the presence of other drivers is to your own behaviour. We call this phenomenon conformity and there is a lot of it about. Research shows that drivers approach junctions faster and brake later when being followed compared to when they are on their own. Other research describes the pressure we all feel to keep up with others, sometimes even when it is not safe to do so. People the driver knows, such as passengers, tend to inhibit speed. In other situations, with anonymous other drivers, it has the reverse effect, as we can see in the early merging in response to upcoming roadworks. None of us wants to experience the aversive stimuli of being hooted at or blocked from merging, nor being regarded as a ‘typical white van/BMW/Audi/Volvo driver’. These factors all sound rather trivial, but they are clearly a more powerful determinant of behaviour than the rational optimisation we, and engineers, would like to assume. And it gets worse. Through social learning these behaviours feed back into the wider driving culture to themselves become local and national norms of behaviour, continually reinforcing what people will keep conforming to.
The good news is that we were able to help our civil engineering colleagues get a handle on all this complexity and make the world a better place. In the short term, we are able to take these insights and provide explicit guidance on how we could use the existing features of traffic microsimulation to make the virtual vehicles behave in more lifelike ways. In the longer term we’ve tapped into a rich seam of research that is providing us with small, clever, user-centred ways of increasing capacity that are orders of magnitude cheaper and more effective than stronger exortations to ‘merge in turn’ or building more lanes. Watch out for future articles.
In the meantime, let’s go back to the beginning and ask the same question: do polite drivers cause queues? Sort of. We are certainly ready to conform to socially accepted norms around queuing, and actively try to avoid trouble and offence by what we see as ‘pushing in’. Then again, I’m certainly not advocating that people become rude drivers.
What about this instead? What about using ergonomic insights to help reduce feelings of anxiety in queues? What about making the cognitive and perceptual environment one in which the behaviours we want to elicit are the least effortful and traumatic to perform? Instead of ‘merge in turn’ what about ‘drivers ahead don’t mind you pushing in’? v
By Guy Walker, Associate Professor in Human Factors at Heriot-Watt University in Edinburgh
This article first appeared in issue 532 of The Ergonomist, October 2014.