The Semiotics of the Road

Human beings are not stationary creatures. We spend significant portions of our lives in transit, whether on foot, public transportation, or by car. When we travel from Point A to Point B, we invariably encounter other human beings. These encounters – and the way we navigate them – are governed by a very complex system of rules. Many of these rules are unconscious, many are a function of cultural norms, many operate under physical or artificial constraints, and many are decided instantaneously.

One of the most common methods by which humans travel is by car. When we drive, we encounter other drivers in their vehicles; we also encounter pedestrians, bicyclists, and various road signs. What I’m interested in exploring in this post is how drivers (and pedestrians and bicyclists) navigate this world, and what signals – conscious and unconscious, overt and covert – they use to guide their behavior, communicate their intentions, and infer the intentions of other drivers.

Why it matters: Understanding driver behavior is already important for improving road user safety. But as autonomous and semi-autonomous vehicles become increasingly prevalent, it also becomes more important to understand how people communicate on the road (Risto et al, 2016; Norman, 2007). Finally, understanding road user communication can offer interesting scientific insights for models of human communication more generally.

The Language of Roads[1]

Human behavior is messy, and cars are both very large and very fast, so it’s only natural that dangers should arise when a bunch of humans drive in cars. To keep us in check, road designers add all sorts of overt signs and signals along the side of the road: stop signs, yield signs, “School zone” signs, speed limit signs, no left turn signs, no parking signs, bike lane signs, walk signals, traffic lights, and so on. These signals have overt, fixed meanings – “green means go” – and essentially function as directives to drivers. The reasoning is that if all road users (drivers, pedestrians, bicyclists, etc.) obey these directives, the road is a safer place to be. In this way, overt signals are a kind of “overt semiotic code” responsible for governing the behavior of road users.

There are two problems with this:

  1. Humans don’t always follow the rules.
  2. Road situations are dynamic.

The first point is obvious; the existence of law enforcement is proof that sometimes humans break the rules they put in place. People drive above the speed limit, don’t fully stop at stop signs, and usually don’t change speed at all when they see a “School zone” sign. This isn’t necessarily because we are “bad”; sometimes we break the rules because it’s convenient, and sometimes we break the rules because we don’t even notice or understand them.

This speaks to the second problem, which is that road signs are largely static, but road situations are dynamic. So even if a rule makes sense most of the time, and even if people follow the rule most of the time, there will be situations in which the rule doesn’t make sense to follow, or could even make things more dangerous. Law enforcement is generally understanding of these situations, but the fact that they arise at all suggests something non-optimal about the framework we use for road design.

Monderman, Woonerven, and “psychological traffic calming”

I’m not arguing that road signs should be abolished; they’re obviously important. But there is a growing movement among road designers and traffic engineers to design systems that implicitly govern road user behavior. These engineers argue that rather than implementing overt signs and directives along the road, actual roads and roadsides should be designed in such a way that the road users naturally adjust.

Hans Monderman, a Dutch engineer, played a central role in this movement. Having observed the increasing monopolization of roads by motor vehicles, he advocated for reframing our image of the road as a “shared social space”, which all road users could enjoy. The idea might sound pie-in-the-sky, but it’s actually been implemented successfully. Woonerven are “living streets” built in many cities in the Netherlands, with the intention of transforming residential streets into these shared spaces[2]. In the woonerven model, streets are conceptualized as rooms of a shared household (e.g. the neighborhood), which all residents – pedestrians, cyclists, motor-vehicle-operators – must work to keep clean, safe, and hospitable.

Monderman’s idea stems from earlier models from “traffic calming”, which is the practice of designing roads to encourage or discourage particular driver behavior. Common examples include: speed bumps, narrow roads, and “rumble strips” along the side of the freeway[3]. Traffic calming is kind of the “blunt force” version of this approach, but Monderman saw in it the kernel of a more general notion regarding street design. In the 1980s, Monderman was called upon to rework the town of Oudehaske; without the budget to implement traffic calming measures, he simply suggested that the road be made more “villagelike”. The product resulted in a massive drop in driving speeds, a change that has lasted through today, and a corresponding rise in the pleasure of walking along the streets. This subtler approach can be called “psychological traffic calming”.

Human-Centered Road Design

So how is this stuff about woonerven relevant to my larger point? I mention this because I think Monderman’s insight can be abstracted further to make a point about the optimal design of systems. Consider another example of intelligent road design – the roundabout – and its obstinate cousin, the four-way traffic light. A whopping 50% of road crashes in the USA occur at intersections with traffic lights; intersections that have been converted to roundabouts, on the other hand, result in a total crash reduction of 40%, and a fatal crash reduction of 90%. Proponents of roundabouts argue that they eliminate the two most dangerous moves of an intersection: crossing directly through, and making a left turn. Instead of relying on overt signals, drivers negotiate their way through a more natural system.

So I think the idea to take away from woonerven is that systems (or “interfaces”) can, and should, be designed in ways that are more intuitive and more effective. The cognitive scientist Don Norman has written extensively about the notion of human-centered design in books such as Design of Everyday Things and Design of Future Things; I think this notion of human-centered design can actually be abstracted beyond product design, and applied to the design of systems.

This includes physical systems like roads, as well as more intangible systems like systems of governance, law and legislation, public health, and more. The design principle in question here is: Systems ought to be designed such that the design of “interface” of the system encourages a particular mode or modes of behavior, e.g. makes that mode the optimal mode. In other words – instead of telling people to do X, Y, and Z, construct a system in which doing X, Y, and Z in the most intuitive, logical, and easiest thing to do[4].

In terms of road design, then, I’m advocating that in addition to including overt signals to guide road user behavior, we design roads so that the design guides user behavior. In this way, a more natural, implicit semiotics or “language” of the road can develop.

This leads me to my second topic: how do people communicate on the road?

Language on the Road[5]

Overt guidelines, such as road signs, help constrain driver behavior, but as we saw above, these constraints frequently break down. In cases of uncertainty, road users need to infer the intentions of other road users. How do they do this?

Researchers in the Automation Group at UC San Diego – as well as various automobile companies – are asking this very question. It seems like the signals used in road user communication can be broken down into several classes:

  1. Situational rules: these are the types of overt signs described in the section above, in which the driver does not need to signal. Rather, there are prescribed rules, with customary outcomes, such as a green light.
  2. Conscious, legal ordinances: these are the types of overt signals used by drivers to signal their intentions, and which have a legally prescribed meaning, such as a turn signal.
  3. Conscious, culturally determined ordinances: these are the types of overt signals used by drivers to signal intentions, which have culturally determined meanings. Within the framework of a given culture, they have a generally agreed-upon meaning, but there may be some variability due to the lack of legal prescription. Similarly, the same signal can sometimes mean different things, even within the same culture – and certainly across cultures. This includes signals like: honking, flashing your brights, waving someone on, or even “flipping the bird”.
  4. Covert signaling: this type of signaling probably still has “intentionality”, but the difference is that road users may not be explicitly coding the signal, and other road users may not be explicitly interpreting it. That is, the level of interaction seems to be different from more explicit signals. This includes signals like: stopping short at a crosswalk, lingering[6], slowing down to allow someone to merge, and so on.

Note that (2) and (3) above might use the same media of communication – lights, gestures, etc. – but the chief difference is in how their meaning is derived. One is legally prescribed and generally has a fixed meaning, while the other is an agreed-upon rule of the road, and can have multiple meanings even within a given culture.

Remarkably, road users get by. Car accidents are, sadly, far too frequent, but unless people could successfully communicate on the road (for the most part), driving wouldn’t even be a viable option.

Autonomous Vehicles and Road User Communication

The task that arises, then, is how to get autonomous vehicles to understand these different levels of communication, as well as design them to effectively broadcast their own intentions. Many automobile companies are racing to attach extra signaling devices to their autonomous vehicles – such as flashing lights to indicate different intentional states – but I think a good place to start is to determine more precisely how it is that human road users interpret the intentions of other road users, and then proceed from there.

First question: can a human road user interpret the intentions of an oncoming car just from its movement patterns, e.g. the covert signaling described above?

  • If not, we know we need some additional level of overt signaling (such as the eye gaze or gestures of the driver); this will impact our implementation for autonomous vehicles.
  • However, there does seem to be evidence that humans do get by with covert signaling; so if that’s the case, we proceed to the next question.

Second question: Even if covert signaling facilitates human-human communication, is this sufficient for communicating intentions between humans and autonomous vehicles? For example, can a human road user interpret the intentions of a driverless car based on its movement patterns?

  • If not, we know we’ll need more overt signaling.
  • If yes, we know this is sufficient, and proceed to a more refined question.

Third question: Even if covert signaling is sufficient or “enough”, is it beneficial to add other signals? Does more information put humans more at ease? Or does it distract from the overall intention signal?

There are, of course, additional relevant questions to ask that go beyond cars communicating intentions. For example, there’s a general concern among the human-robot interaction community that making robots too human-like will cause other humans to make incorrect inferences about the robot’s abilities. This is referred to as a the habitability problem: how can we design robots in such a way that humans accurately assess their capabilities, and thus use them to their full extent, but do not expect too much from them?

Conclusion

None of this spells doom for the commercialization of autonomous vehicles. However, it does suggest that the integration of autonomous vehicles into our existing roadways may be a more complicated process than we’d like to think. And since autonomous vehicles will doubtless already face resistance, it’s important that when they are integrated, they are done in ways that are both safe and naturalistic.

 

References:

 

Traffic: Why we drive the way we do (and what it says about us). Vanderbilt, Paul. 2009. Knopf.

Norman, D. (2007). The Design of Future Things. Human Factors and Ergonomics in Manufacturing (Vol. 18). http://doi.org/10.1002/hfm.20127

Risto, Malte; Müller, Lars; Emmenegger, C. (2016). The social dilemma of autonomous vehicles. UBICOMP, 352(6293), 1573–1576. http://doi.org/10.1126/science.aaf2654

[1] Note: almost all of the information in this section – apart from the insights at the end – has been adapted from Paul Vanderbilt’s 2009 book Traffic: why we drive the way we do (and what it says about us).

[2] Lest you think the implementation is Dutch-specific, I’ll add that they’ve been implemented under different names in other countries, even the USA.

[3] Which apparently reduced run-off road crashes by 70%.

[4] This is obviously difficult, and it’s doubtful that an ‘optimal system’ could be developed. But I do believe that if people designing a system – in any domain – keep this principle in mind, the system will be more successful.

[5] Note that many of the insights in this section come from meetings with the Automation Group at UCSD, as well as discussions with Colleen Emmenegger, one of the researchers there.

[6] When a pedestrian lingers on the sidewalk to make it clear they’re not trying to cross.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s