Emission Controls, Defeat Devices, and Computerization: The Volkswagen Debacle in Historical Perspective

Volkswagen executives have admitted to the Environmental Protection Agency that the company used so-called “defeat devices” to beat automotive pollution control tests. VW’s stock dropped over 20%, and the crisis meeting its board is holding this Wednesday will likely be bloody. News reports suggest that the fiasco will cost the company at least $7.3 billion and may involve the recall of 11 million cars. The situation highlights changing aspects of the auto industryespecially the ever-increasing role of computerization in vehicles—and the need to reinvigorate American regulatory agencies, which have seen budget decreases since the federal sequester. But in misleading the EPA and the public, VW joins a long and proud tradition of companies that have tried to dodge federal standards. 

Defeat devices are not new; neither is cheating emissions tests. In 1972, executives of Ford Motor Company came to the Environmental Protection Agency (EPA) to confess that one of the company’s divisions had been faking data for federal air pollution tests. The employees of the division had been conducting engine maintenance more often than was allowed by the federal test procedure. Frequently changing spark plugs, engine filters, and the like meant that the cars ran much cleaner than normal. So, it looked like the company’s vehicles were passing the tests with flying colors when, in reality, they were frequently flunking. 

That same year EPA staff members were running emissions tests on an American Motors Corporation (AMC) car. It failed. Everyone, including AMC technicians who were observing the tests, scratched their heads. “It must have been a bad sensor,” one of the AMC techs suggested. “What bad sensor?” asked an EPA staff member. That test initiated a series of events that led to EPA administrators coining the term “defeat device.”

It turned out that auto designers at AMC were using temperature sensors to turn off its vehicles emissions controls under certain conditions. Yet, to guarantee laboratory-like reproducibility between emissions tests, federal procedures required tests to be conducted with ambient air temperatures between 68-86 degrees Fahrenheit. EPA staff members realized that AMC had been using the sensors to control emissions only in that temperature range.

While this discovery was problematic, much more worrying to regulators was the idea that other automakers could be doing the same thing. It was for this reason that administrators at the EPA created the phrase “defeat device,” issued an rule banning the use of such systems, and brought attention to the idea in the media. One auto executive later said he knew that the automakers had been beaten on the issue when he saw the words “defeat device” written in a newspaper. As a rhetorical weapon, the phrase was just too good.

Since the 1970s, the EPA has occasionally found automakers using systems that functioned as defeat devices. For instance, in 1995, the agency sued General Motors over an engine control system that turned off emissions controls outside of test-like conditions. Industry resistance to regulations remains a real problem and likely always will.

The fact that Ford, AMC, and GM tried to beat federal air pollution regulations and almost no one remembers undermines some arguments we see around the VW case. On Twitter and via other media, I’ve seen individuals say that VW’s reputation has been ruined forever. These people are seriously overestimating our collective memory. 

One thing that makes the VW case different than Ford or AMC ones is that it involves the use of computer software. As the AMC case shows us, however, defeat devices are not about computers but about auto engineers using technical systems to mislead regulatory authorities. They are dishonest technologies.

Still, the presence of computers raises important questions and issues. The simple reality is that the regulatory agencies focused on the automobile—both the Environmental Protection Agency and the National Highway Traffic Safety Administration (NHTSA)—have not kept pace with the changing role that computers plays in automotive systems. During the Toyota recalls of 2009-2011, some individuals hypothesized that the computers in the company’s vehicles were causing “sudden unintended acceleration.” When NHTSA administrators testified on the matter in front of Congress, they were forced to admit both that they themselves had little understanding of computer systems and that they did not have any electrical or software engineers on staff. The agency turned technical analysis of the Toyota cars over to NASA (which did not find anything wrong with the computer systems).  

This knowledge gap is part of a more general problem around government and computing. A colleague of mine, Arjan Widlak, runs the Kafka Brigade, a consulting group that helps organizations overcome the problems of red tape.  Widlak has consulted governments in many parts of the world, and he has said that he consistently finds blind spots around computing in such organizations. Part of the issue seems to be one of responsibility. Computing has entered nearly every part of society, and yet an individual who already has responsibilities within a bureaucratic organization may not necessarily see computerization as part of his or her job. It's not that things should work this way; they just do. The result is that in many areas, including auto regulation, computerization has blindsided organizations.

Well, what to do? 

Wired has one answer. In a rather unfortunate piece of writing, titled “The EPA Opposed Rules That Could’ve Exposed VW’s Cheating,” Alex Davies calls attention to the agency’s opposition to changes in intellectual property law. The software in automobiles is covered by the 1998 Digital Millennium Copyright Act (DMCA), which makes it illegal to hack through “technological protection measures” or reverse engineer software. The Library of Congress issues exemptions to the law, and a “group of proponents” lobbied the Library to make an exemption for automotive software. The EPA opposed these efforts because its staff members worried that hobbyists hacking the software in their cars would change the engine performance parameters in such ways that the cars would violate federal air pollution laws. 

(By the way, people are already hacking their cars' engines, and when they do so to increase performance, their cars DO violate federal emissions standards. I brought the moral problems with this up with a guy who digitally tinkers with his engine and who puts it back to factory-settings before taking the car in for his annual emissions inspection, and he just shrugged.) 

It’s here that Davies article enters the realm of speculation. There are a lot of words and phrases like “could” and “it’s possible" and “good chance” in the piece. It's a real stretch. He writes, “In opposing the exemption for individual car owners to examine the software, the EPA would close an important avenue for uncovering security and safety issues in vehicle software, because often these kinds of issues are uncovered by individual research while simply examining their own product or vehicle for fun or curiosity, not formal research.” In other words, Davies uses the Volkswagen debacle to push the classic cyberlibertarian ideology, wherein intellectual property is evil (“Information wants to be free, man”) and hackers and makers will save the world. 

There is something slightly off-putting, even gross, about using this tragedy to promote such an agenda, but that isn’t even the point. I am fine with the federal government making automotive software exempt from the DMCA, an extremely flawed law to be sure, but the idea that the hacker, or reverse engineering, equivalent of citizen science is even a partial answer to corporate malfeasance around regulated technologies seems slightly crazed. The suggestion certainly lacks any sense of proportionality. The US federal government did not drop some regulated emissions in automobiles by 99 percent or drastically reduce the number of people killed and injured per mile driven by relying on tinkerers. Such change requires organizational capacity, real resources, and expertise.

Which brings us to the true solution to our current plight around automobiles, computers, and regulation. Watchdogs have criticized federal auto safety and air pollution on enforcement for years, and it has been nice to see these agencies recently step up enforcement efforts. But the reality is that these organizations face a shortfall of resources, especially since the federal sequester. Regulatory aggression does not amount to much without resources. What we are witnessing is a virtual deregulation via defunding. There are other historical examples of this. For instance, the number of automobiles recalled for being unsafe dropped drastically when Ronald Reagan became president and cut federal budgets.

The only way to solve our problem is to build organizational capacity in these regulatory agencies around computers. Doing so will require real resources. Computer scientists and other technical experts don’t come cheap. But truly this is our only option. We need offices of dedicated experts going over and testing automotive computer code for the sake of safety and clean air. 

Put simply, if we value our lives and our health, we should value the organizations that we built to protect them.