Loading...
In 2006, Volkswagen launched an ambitious strategy called 'Mach 18' — become the world's largest automaker by 2018. The plan hinged on cracking the American market with 'clean diesel' cars that promised both performance and low emissions. There was just one problem: the nitrogen oxide (NOx) standards set by the U.S. EPA were nearly impossible to meet without sacrificing the driving experience German engineers prized. Rather than accept the tradeoff, VW engineers developed a 'defeat device' — software that detected when a car was undergoing an emissions test by monitoring steering wheel position, vehicle speed, engine operation duration, and barometric pressure. During testing, the engine ran in a clean but sluggish mode, producing NOx levels within legal limits. On the road, the softwar...
Popular framing: Corrupt VW executives chose to cheat; jail the bad apples and the lesson is learned.
Structural analysis: A measurable goal (NOx in lab tests) diverged from the real-world goal (actual emissions), so Goodhart's Law converted the test into the work — once the metric and the underlying property could be decoupled, the cheapest way to hit the target was to game the metric. Bonus structures tied to volume and market share aligned every internal incentive with passing the test; principal-agent asymmetry meant engineers who raised concerns were sidelined while compliant engineers were promoted. Normalization of deviance hardened the defeat device into routine across nine years; replace the executives and the same metric-plus-incentive architecture re-creates the fraud.
The popular framing focuses on punishment of bad actors, which satisfies moral intuitions but leaves the underlying system unchanged. If the metric remains gameable, compensation remains tied to metric-dependent targets, and regulatory monitoring remains slow, the next Dieselgate is a matter of when, not whether — just in a different industry or with a different technology.
