Friday 11 October 2013

Embodiment and design; the affordances of pedestrian crossings

I'm a sucker for good design. I'm interested in things that work well because they are designed with the right task in mind. Take the greatest potato masher of all time, the Spudnik. It works well because it mashes as the result of a very natural action with the arm, easier and less effortful than the more traditional device. Good design works with the user and the task (rather than trying to impose a behavior) because behaviours that are supported by the task and the environment will be stable, reliable and easy to maintain. 

Design is interesting for embodied cognition because it's an attempt to artificially manipulate the environment to create affordances for some but not other behaviours. Sometimes it's just a matter of getting the physical layout right (see the Spudnik). Some of these required behaviours, however, are quite complex and not the sort of thing that would typically create information if left to their own devices. A great example is the design of streets to promote safe driving and pedestrian behaviour; a lot of the rules being designed for are human conventions, not physical necessities, and so without someone intervening and building something there couldn't be perceptual information anywhere about that rule. In addition, the designed element often creates information about something other than itself (see the Aboutness dimension in Sabrina's information taxonomy). This in turn creates the possibility for there to be more than one way to create an environment that produces relevant information and can therefore shape behaviour, and in turn, this allows the possibility that some designs will be better than others.

With this in mind, let me introduce you to two examples of design that I would like to change; the staggered pedestrian crossing, and puffin crossings. Both of these artificially restrict access to useful information in ways that mean well but that I think fight too much against human behaviour. I actually started preparing a grant to empirically investigate these designs from a more embodied perspective, and the EPSRC thought it was in their ballpark. More pressing concerns intervened, but I would love to actually do these studies and would like to hear from anyone who might be interested in collaborating ("Dear Pamela lab..."). I think our embodied cognition approach (Wilson & Golonka, 2013), with it's focus on task analyses and information, could really have an impact on an interesting part of our day to day life.

Staggered crossings
Note that in the real world, there's also a lot of fences forcing people to cross at the lights
A staggered pedestrian crossing is one in which you cross the road in two stages. You cross one side, walk down a little island, then cross to the other side. The lights for each stage are independently controlled. The general guideline for building these is to make you walk so as to face oncoming traffic as you cross the island in the middle; this is so you can see what's coming up traffic wise. This is only a guideline, however, and there are many staggered crossings in the UK where this isn't possible for some reason, and the stagger goes the other way.

The basic idea is not a bad one; use a series of fences to guide pedestrians to walk in a particular way across a road. (NB Missing from that picture are the many fences required to make pedestrians enter the crossing at the right point!) However, staggered crossings include a design feature that fights against what I'm pretty sure will be the pedestrian behaviour that emerges from this layout, and it drives me slightly mental whenever I encounter it. 

The feature is this: blinders on all green signals after the one controlling your first crossing.
These blinders are arranged so that you cannot see the green signal until you are facing almost directly opposite. The logic is that because you must cross two sections of road and the lights are independent, being able to cross one does not mean you can cross the other. They put blinders on all green signals after the first one so that you won't accidentally keep crossing if only one turns green. The design imperative is to restrict your access to information that might mislead you. You can always see the red light when it's on.

A schematic of the staggered crossing light setup

In order to be able to obey the lights at a staggered crossing, you must walk straight across the first crossing, turn, walk down the island until you are nearly directly across from the next signal, turn, wait till it's green then cross. In effect you have to walk as depicted by the arrows (as the imposed trajectory). 

I see two problems with this. 
  1. There are large portions of the trajectory where you have no information about whether you can cross. This leaves the pedestrian in limbo about what is coming up until very late, and given that it takes time to move and change direction our perceptual systems are typically hungry for information about what's next, rather than what;s going on right now. 
  2. You are being asked to walk to a point, stop, turn, walk, stop, turn, check and only then go if it's indicated. People don't typically walk like this; turning on a right angle is actually hard when you're moving because of your inertia, so real locomotion trajectories change direction as smoothly as possible. I've added in a dotted line to indicate my guess as to the preferred trajectory people will try to take given the affordances (spatial layout) of the staggered crossing. The path cuts corners, it wants to turn early and it therefore wants to know whether it can go as soon as possible. (Pedestrian control designers know about these kinds of paths, which they charmingly refer to as 'desire lines').

The designer reply is 'so what? I'm trying to make people behave responsibly'. This is all well and good but it's fighting with people's preferred locomotion (a preference that is supported by the spatial layout of the stagger and by the control requirements of walking) and my bet is that in that fight, design loses nearly every time.People will cheat and make bad decisions rather than obey the imposed trajectory.

My project idea: I think that the relevant desire lines emerge from people moving through a specific spatial layout (i.e. this layout affords some and not other paths to moving people). I think that restricting access to the green signals is an unnecessary precaution that leaves people operating in a 'no information' setting (and this can be disastrous). I think you can build a staggered crossing that works with human locomotion, and not against it. 

I would first like to collect data from real staggered crossings and measure people's behaviour (what are the desire lines given this layout?). I'd then like to model that behaviour (perhaps by adapting an existing model, e.g. Fajen & Warren, 2003, or Wilkie & Wann, 2005). I'd give the model the layout of the staggered crossings and replicate the observed trajectories. Then, I'd adjust the layout presented to the model to see what happens and if we can identify a better form. Then I'd record real people walking through real layouts in the new forms (and old forms with continuous access to green light information) to see if they behave more usefully. Warren at least has gotten his model working in multi-agent settings (he gave a great talk at VSS2013 on this stuff) and it's specifically about walking; Wilkie & Wann are more about navigating at higher speeds (e.g. driving). I think this task would make a kick-ass test bed for testing these models, though.

The main thing I'd need (besides access to crossings I could film and the technology to do so, problems I had in principle solved when I was working this up) is the ability to test novel layouts. A gym and some fencing might do the trick, but if anyone reading this works at the Pamela lab, call me; I may yet call you!)

Puffin crossings
In my research for the above project, I found out about the different kind of crossing systems. There are pelican crossings (the familiar kind you see most everywhere), toucan crossings (for both bikes and people - 'two can' cross, get it? Ha!) and puffin crossings. Puffins are the new big thing and are supposedly getting phased in throughout the UK. 

I hate them and I want them to die.

The idea, again, is not a bad one. Puffin crossings put the green light next to the waiting pedestrian, rather than across the road. This makes it easier to see, especially on bright days and for visually impaired users. They also have some sensors and changes to the way they cycle that make them a bit more responsive. 

But again the problem is about removing access to information. With a puffin crossing, as soon as you leave the sidewalk, you can no longer see any lights, and you have no idea whether the crossing is still good. Actually, the information about whether it's safe to cross is not where people typically look - at their upcoming path across the road, and not down the road at the traffic where puffin crossing designers think they are looking. 

There are two common cases where this is a problem; when you've run onto the crossing in the middle of the cycle and you don't know how much time you have left, and when you are slow for some reason (for instance, older adults) and you might want to know if you're about to hold up all the traffic. I saw an elderly lady nearly get killed on one of these; she started crossing very slowly and immediately had no idea whether she was still allowed to cross; no green signal anywhere. She stopped, turned, and came back very distraught and had to try again. 

When people are moving they don't like to simply get going and then cast themselves adrift from information about how things are going. All actions entail ongoing, real time monitoring of progress and at a crossing, green signal lights are part of that. I'm not especially slow and even I find these crossings very disconcerting; it's like the green light has turned off once I head out, rather than just swung out of view. 

My project idea: For this one I'd mostly like to find out if I'm right in my intuitions that the absence of a visible green light throughout the crossing affects people's locomotion. Video from an established crossing (ideally before and after the the transition from pelican to puffin crossing) would be about all that's needed. I suspect these things are a terrible idea and would love to be able to back that up. Impact case study for the next REF, anyone?

Embodied cognition means, fundamentally, getting serious about the environment and information about that environment. When you do, you see very different solutions to problems that (so far) typically work much better than the old solutions (see our review of the successes of embodied cognition in our Frontiers paper). Safe pedestrian locomotion through a paved world designed for cars is a place where this work might really make an interesting difference, and one day, I will run these studies!

Fajen, B. R., & Warren, W. H. (2003). Behavioral dynamics of steering, obstacle avoidance and route selection. Journal of Experimental Psychology: Human Perception & Performance, 29(2),, 343-362. Download

Wilkie, R. M., & Wann, J. P. (2005). The role of visual and nonvisual information in the control of locomotion. Journal of Experimental Psychology: Human Perception & Performance, 31(5), 901-911. Download

Wilson, A. D., & Golonka, S. (2013). Embodied cognition is not what you think it is. Frontiers in Psychology, 4, 58. Download


  1. Speaking of embodied cognition and traffic, this is worth the short read:

    Essentially Monderman redesigned a traffic circle by removing any signage in order to improve navigability by revealing the information that is already perceivable, he subtracted the noise.

  2. I was thinking about this issue and the article from the first comment over the weekend while seeing all these TV commercials from car companies that tout vehicle safety technology (active braking system, blind spot notification). Will this technology lead to more 'idiot' drives that don't have to always monitor information while driving? Do you think drivers would become reliant on this information? Or just use the information in critical situations?

    1. People use what you give them; I think people would become very reliant on these systems. I know lots of American drivers who rent cars or drive here in the UK and get freaked because there's no parking camera.

      People need good information in order to produce good outcomes, and that can come from technological systems. The issue is, how good is the information from these systems and under what conditions would it fail?

  3. Hmm.... this seems like the type of study that could be confirmed or denied pretty easily with a few days of just sitting and watching the cross walks (says the animal behaviorist). You would still want to do modeling and testing for other reasons (i.e., to be able to recommend specific improvements), but it would be easy to know if you were right or not on the broad question before you began.

    1. I had some ideas for video crossings and either coding or digitising the result to produce estimates of dodgy behaviour or weird locomotion forced by the design. Anecdotally, I see a lot of stupid behaviour on these crossings.

  4. Separate buttons for slow crossers? Have you seen this? Not sure how best to share image:

    1. An ok idea but a) it requires too much labelling so it breaks the Donald Norman 'if you have to label it you designed it wrong' principle and b) it still doesn't solve the problem on the puffin crossing of not being able to see the green light once you're on the road.

    2. Agreed that it doesn't solve the problem of the puffin crossing. I'm not sure about the labels. I agree in principle, but I suspect no one reads the labels more than once. I thought you would get a kick out of it when I saw it. It was posted on facebook by Leon van Gulik, who spent some time at Clark while I was a postdoc there, before returning to his native Netherlands.

      I was oscillating between thinking that the slow-button was a good idea in general, and thinking that something was wrong with the design of the normal system if it did not permit enough time for someone slow to walk across. Imagine how much harder the puffin crossing would be if you were very slow to get all the way the point where you could see the next light (i.e., the cost of not being able to take the diagonal is greater for the slow).

    3. The other thing about labels; it's telling you that you have the same system with two modes that function quite differently and that you must really get it right. Too fragile!

      Pedestrian crossings aren't made with pedestrians in mind; they're made with car traffic in mind. What's the shortest time you can expect to hold up traffic at that light?

  5. Puffin-style nearside indicators for pedestrian traffic cannot be seen when someone is standing in front of them so are unsuitable when more than one pedestrian is present at a time. They require a 180-degree head turn to view vehicle traffic coming from the left so are unsuitable for two-way single carriageways. There is no signal visible to the pedestrian during crossing so are unsuitable for jurisdictions where jaywalking is not illegal. Pelicans have none of these drawbacks and can be further improved by countdown indicators and nearside repeaters, and could be fitted with the same auto-detection as Puffins. Puffins are a disaster. Joe May14

    9 May 2014 09:23

  6. In 2011, TRL completed a study of 50 crossings that were converted from the older Pelican design (which has the pedestrian signals on the far side of the roadway) to the Puffin design (with pedestrian signals on the near side of the roadway). There was a 17% reduction in personal injury accidents with the new design, most likely because fewer people attempted to start crossing late in the cycle. The offset central refuge is also the result of decades of pedestrian behaviour research. In the US we don't do these things, and consequently we kill and injure a much higher proportion of pedestrians. Here's a link to the 2011 study.