Regulators fight the last war: Mark Maxham on software and consent decrees
After we finished talking about scooters, I asked Mark Maxham about a different tool. When regulators want to change how a big technology company behaves, they often reach for a consent decree, usually as the price of approving an acquisition. Mark builds platform software for a living at Lacuna, so I wanted a developer's view of whether that works.
His answer started with timing.
I think there's an incredible mismatch between innovation in the context of technology and good governance. Innovation is constantly on the move, right? Which means that by the time regulating bodies have recognized that there's some sort of a problem, they're fighting the last war.
He ran through examples. Tell Facebook it has to let you change how your feed is filtered, and it has probably already changed it. Order a company to let users delete their data, and it may already have done that in response to its own users. Google favoring its own shopping results, Amazon and its sellers: by the time a case arrives, the conduct may or may not still be there. He called it fighting a war on quicksand. He is put off by monopoly behavior, and said he hated to sound like a libertarian because he is not one, but too often the cure is worse than the disease.
Then he walked through the clock. Staff study a problem and draft something, committees hold hearings, and at the state or federal level you are six to twelve months in before the debate starts. Then the lobbyists arrive. Two to three years pass before anything is done, and when a law turns out to be written too narrowly or badly, it is almost never fixed. Instead it collects exceptions, or another law is passed on top of it.
His example was California's AB5. The legislature wanted to change how ride-hailing drivers were classified, but it could not write a law about two companies, so it wrote a sweeping definition of who counts as a contractor. The collateral damage started at once and the exceptions followed, until, in his words, the law was Swiss cheese. Then the companies it was aimed at spent around $200 million on Proposition 22, which passed ten days ago and writes their own rules for their own drivers. What remains of AB5, as he sees it, is mostly the unintended consequences. The lesson he drew was not about labor law. A broad law written to hit a narrow target hits everything else first.
Consent decrees, he thinks, have the same problem in a different form. The Justice Department has been fairly hands-off about big technology acquisitions for the past decade. If it got aggressive and started attaching conditions, it would be setting multi-year rules for markets nobody can see the future of. People predict price Armageddon when two companies combine, and sometimes they are right, but nobody knows in advance.
This is the gap every software engineer will recognize as a release cycle against a rule cycle. A product that ships every week and a rule that takes three years to write will never meet. The rule is always aimed at a version that no longer exists. That is the strongest argument I know for what Mark had described earlier the same afternoon, rules published as data that can be updated as fast as the thing they govern.
The Justice Department sued Google last month, and the House antitrust subcommittee has recommended changes that would take years to legislate. Mark's question is the one to keep asking about every remedy on the table: will it still be aimed at something real by the time it takes effect? What he thinks should replace that cycle is the other half of this conversation: government power, corporate power.
From Lincoln Shorts, the interview series I hosted for Lincoln Network, recorded 13 November 2020.
