Guiding Policies For Changing Driver Behavior

Watching a video of Google’s self-driving car is a lot like science fiction: it can navigate the winding streets of San Francisco with ease. It recognizes pedestrians from distances outside the human periphery. Compared to a driver, it can stop in a fraction of the response time. Eric Schmidt, Google’s CEO, summed up the project’s vision at Techcrunch’s Disrupt conference in 2010: “Your car should drive itself. It just makes sense…it’s a bug that cars were invented before computers.” Self-driving, automated cars have arrived. With successful test runs, policy makers can both consider the realities of automation, and prepare for a transition from manual-only driving.

Google’s driverless car outside at TED

This is the final of three pieces we have written about the future roles digital technologies could play in shaping how we drive. A collaboration by three strategists here in the office, each piece explores complementary but different aspects of the problem. Here are parts one and two.

Schmidt is betting on the benefits of automation, a pursuit shared by GM, Volkswagen, Volvo, Audi, and BMW. Imagine roads saturated with self-driving cars; drivers – who account for most automobile fatalities due to errors of judgement, distraction, and recklessness – would no longer directly affect traffic. The result is a significant decrease in traffic accidents (and, when cars are connected to each other, reductions in congestion once road capacity increases).

The challenge for automation, however, is not technology, but law – automation is illegal in the US. For automation to reach scale, each state must pass an amendment to the entire driving code. Despite automation’s promising benefits, policy makers have done virtually nothing, failing to address the recent progress (and its implications) of Google and automakers.

Before automated cars are unleashed on the streets, policy makers must address a number of challenges:

1. The transition from manual-only to automation will revolutionize the act of driving, creating new behaviors and in-car experiences.
2. Several generations of drivers will need to learn new driving practices.
3. Emerging technologies will challenge safety standards.
4. Significant collaboration will be needed to encourage interoperability among automakers.

Besides designing a new traffic code, policymakers must set forth strategies to overcome these challenges. They should create an organization to foster collaboration among automakers, evolve the driver education system, and adopt new safety and technology standards. A holistic approach to automaker and driver regulation could make automation a 10-year goal rather than a 100-year goal.

Challenges

The transition to automation will involve many challenges, based on decades of data from technology advancements in automobiles and other industries.

Education

Should drivers believe that automated features will perform 100% of the time? What happens if the computer freezes? Even at 99% success rates, policymakers must consider the implications of new technologies on driver behavior over time.

One concern is that automation will affect a driver’s ability to improvise. In commercial aviation, system failure improvisation – extraneous events that software designers cannot account for – is built into a pilot’s simulation sessions, allowing them to gain helpful experience on how to handle emergencies. Today, drivers don’t benefit from such training; technologies that activate computer-assistance during emergencies bypass crucial lessons for drivers. If a computer prevents a near rear-end crash by slowing a car, for example, or even takes over in the half second before a potential crash, drivers may lose helpful defensive driving training and experience. Drivers as a result must require more intensive training to make them adept at operating both manual and automated vehicles.

Interoperability

In the dawn of automation, policymakers can anticipate a significantly different and more complex driving experience. Operating a new automobile system may be as difficult as learning a new computer program. Further, automakers compete on phone-to-car syncing technologies – coming from companies like BMW, Ford, Mercedes, and Hyundai – lowering incentives to collaborate on a common platform.

For vehicle-to-vehicle communication, the chosen programming language could pose challenges to reach scale. If BMW and Ford are operating on different languages, the benefits of automation will be compromised. If a common language is to be supported, policy makers must be sure to help prevent a competitive battle among companies based on financial incentives or patents, as observed with Blu-ray vs. HD-DVD.

An open source language, common in non-automobile software development, could be the answer. Such a choice, however, has implications on safety. Standards will be needed for who can develop and modify automation software, without discouraging innovation. Given the stakes (i.e., a car crashes) if code is not up to safety standards, guidelines will be needed for amateur developers who wish to add features or advance performance.

Behavioral Concerns

As new automated technologies proliferate, healthy driver behaviors cannot be not sacrificed. Policy makers must manage the trade-offs between innovation and safety. The human factors of driving, like driver’s judgement of risk and attention, are affected by new technologies, often in unintended ways that could compromise safety.

Risk

Risk influences a driver’s understanding of safety. When anti-lock brakes were introduced in the 1970s, the benefits of brakes were offset by unintended driving behaviors, like closer driving distances and poorer response time. Similar results were observed for seat belts. Based on numerous studies, data showed no correlation between seat belt legislation and reductions in traffic injuries or fatalities. Traffic engineers justify this with a phenomenon called risk compensation: when driving gets safer, drivers become more reckless. The opposite is true too: when driving is perceived as riskier, drivers exhibit safer habits.

As a way to curb reckless behavior, traffic engineers purposely design around a drivers’ perception of risk. Risky environments, such as winding roads and poor visibility, heighten awareness, often leading to a considerable drop in traffic accidents and consequently better driving habits. Traffic engineers term such conditions as “dangerous by design.” Conversely, elements like yellow lines, signs, and barriers, act in the opposite way, calming the stress levels of drivers. The highway, with its bright painted lines and concrete barriers, purposely mitigates perceived risk to make driving at 70mph with dozens of cars bearable. Automation is a similar mitigating factor to driver stress levels. As with anti-lock brakes and seat-belts, drivers may feel irrationally safe if computers take over in moments of danger. That is, the very technology promising to make drivers safe could lead to riskier driving behaviors.

Attention

With the potential for riskier driving, drivers may become more distracted too. Drivers, at least intuitively, understand the risk of taking their eyes off the road. Consumer demand for new driving apps is growing (e.g., Waze), and app development is proliferating: most automakers have technology that syncs automobiles with smart phones, tablets, and laptops, allowing for an immersive driving experience with endless possibilities. Future automobiles will be hubs of multi-tasking for work and play; evolving the driver into a passenger.

Regulating driver behavior has been challenging for policy makers. Consumer software, is not subject to testing or safety regulations. The risk is that multi-tasking will increase with automation; reckless activities, like texting while driving, will be perceived as safer due to computer-assistance. When boredom and monotony set in on a familiar, empty, and straight road that requires little attention, drivers will keep their minds busy with something else, like texting or listening to music. If automation frees up cognitive load for drivers to focus on other tasks (i.e., a car can steer itself from cross-lane drifting and auto-brake on a split second), a few seconds of tapping away on a center console could be deemed, on a subconscious level, reasonable.

As automation reaches automobiles, policymakers will need to re-evaluate existing regulation. New technology will require standards to slow the growth unsafe driving behaviors.

Guiding Policy

Testing
Policymakers must overcome their historical risk-aversion to innovation in automobile technology, which has likely hindered industry progress toward automation. Rather than sweeping regulatory bans, policymakers should adopt an approach of diligent testing. To understand which technologies lead to behavior changes, policy makers can use existing data tracking tools to aggregate telemetry and geo-location information from drivers. A small sample of drivers voluntarily donating such data could help inform the most appropriate regulation on new emerging technology.

Many states have considered legislation to ban distracting technologies for drivers, such as cell phones or driver-accessible screens playing video. These laws may be unnecessary if automation is designed to allow for safe operation of a car while multi-tasking. Conversely, if technologies such as adaptive cruise control could lead to riskier driving, policy makers can begin to set informed standards for automakers. Rather than requiring rigorous pre-launch testing or a ban on all technologies, an informed approach will encourage innovation while ensuring driver safety. Policymakers can anticipate that emerging technology will compromise safety; standards will help maintain the credibility of automobile transportation until full automation is achieved.
Learning
Driver’s training hasn’t changed much over the past decades, but then again, neither has driving. Operating a 56’ Mustang isn’t much different than a 2011 model. Policymakers should anticipate that progress toward automation will certainly change the driving experience, from operation of automated features (e.g., adaptive cruise control) to the multi-tasking with an automobile’s web-enabled interface and set standards for training accordingly.

An evolved, digital-focused Department of Motor Vehicles (DMV) includes a new curriculum that recognizes the changes to driving over the coming decades. As drivers adjust to a flow of new technologies, policymakers must consider a process of continuous learning, a sharp change from the current one-time, adolescence-focused approach. Digital-based education could allow policy makers to quickly evolve learning modules to account for quickly emerging technology. Similar to an aviation type rating, incremental training and certification may be needed to operate vehicles that acutely deviate from the typical driving experience.
Automaker Collaboration
With competing incentives potentially hindering progress, policymakers can look to other organizations for best practices. Of particular interest are non-governing, independent bodies like the World Wide Web Consortium (W3C), which holds no official authority but creates harmony among stakeholders. W3C is a consortium of industry leaders that sets web standards, most popularly for HTML and CSS; their work has helped the web maintain interoperability (i.e., pages will work on most browsers), allowing web design and functionality to evolve seamlessly. Similarly, the Financial Accounting Standards Board (FASB) sets accounting standards in the US, allowing for consistent financial reporting (a common approach and language allows financial reporting to quickly and systematically change when business practices evolve).

An organization for automation will offer similar successful collaboration, allowing automakers to develop a common in-car experience. Drivers should not have a confusingly different user-experience among automakers or vehicle type, thus policy makers should encourage automakers to design interfaces and functionality with a high degree of commonality. Policymakers should work with automakers and consumers to develop a common language for vehicle automation as well, along with certification for development. Given the success of arriving at a common language in other industries without government-imposed guidelines, it’s likely that a decentralized approach will yield progressive results.

While automobiles have a history of mechanical tinkering, amateur developers will require proper education and software testing before hacking into the code behind automation. Automakers and developers that are accountable to regulatory testing may be only parties licensed for software development. A free for all app-store for automated vehicles, due to safety concerns, may not be feasible in the future.

The Future

On one extreme, policymakers could fiercely regulate automation, micro-managing the creation of infrastructure and requiring that automakers automate vehicles by a designated year (e.g., 2050). Safety could be guaranteed under a pharmaceutical industry-like approach to development and innovation testing.

On the other extreme, policymakers could allow independent innovation from automakers to continue, accepting the risks to drivers with a laissez faire approach.

A middle ground, structured on diplomacy and supported by rigorous education and testing, will help ensure progress toward automation. By working closely with automakers, consumers, and traffic planners, policymakers can begin to plan for the digital age of driving without preventing it.