← Back to Posts

Rules in a PDF invite argument: Mark Maxham on digital policy

Sean Roberts•November 13, 2020•3 min read

Last month the OECD published a paper called Cracking the Code: Rulemaking for Humans and Machines. It gives a name, Rules as Code, to the idea that a government should publish a machine-readable version of a rule alongside the human one, and it calls the idea early-stage with much still to test. I read it the week it came out and thought of Los Angeles, which has been testing it on scooters for two years without calling it anything. So when Mark Maxham and I sat down last week I asked him to describe what the city actually does.

Digital policy is this idea that instead of expressing rules in English via PDFs that allow for human interpretation, you publish them in an agreed-upon digital format. And it's meant to be human readable. The descriptors are English and the numbers are ten and twenty. It's not binary, opaque, only the machines can understand it. But the idea is that if a city publishes these rules, then the mobility providers download them and they can either interpret them with their own code or they can use the open source reference implementations that LA has built and contributed to the Open Mobility Foundation.

The first benefit is the obvious one. A rule written in English and read by two companies produces two readings and, eventually, a dispute. A rule published as data with a reference implementation produces one reading. Mark's phrase was that it prevents the misinterpretation that leads to conflict, and anyone who has watched a permit condition get litigated knows what that is worth.

The second benefit is the one that made me sit up. If the rules are data, an operator can run its own plans against them before acting. Mark's example: a company has five thousand scooters out and is about to roll a truck and drop five hundred more. It can ask, before the truck leaves, what fifty-five hundred would look like against the current rules, and learn that it is allowed two hundred. Nobody gets fined, nobody argues, and the city never has to notice. Compliance moves from enforcement after the fact to a check before it.

The third is that policy published this way can change quickly, which is where it starts to feel like a different kind of government. Caps can be dynamic: more vehicles around the stadium on game day without flooding the neighborhood for the following month. Exclusion zones can be published for a fire, a festival or road work. The city expresses an intention in data and the fleet responds, instead of a memo going out and an inspector going round.

I pushed on whether a rule language built for scooters survives contact with anything else, and this is the part Mark called the medium-term work. Taxis are different from scooters; LA is integrating taxi data now and intends to bring the definitions to the foundation once they have proven out, rather than arriving with a maybe. Docked bikes differ from free-floating ones. A Zipcar at the curb is in the public right of way and a Zipcar at its home space is not, which matters if you are counting. Buses have routes and stops. San Jose wants delivery robots in the specification; Nashville, with Vanderbilt, is looking at emergency vehicles. You go mode by mode, he said, and work out what you can recycle and what is genuinely different, rather than cramming everything into the scooter frame.

He was candid that all of it is in its infancy, governance by API included, and that even within it there are degrees of how well the implications are understood. That matches the OECD's caution. The difference is that LA's version is running, in the open, with code anyone can read, and the foundation it set up has the other cities at the table. The paper describes the idea. The city has the reference implementation.

From Lincoln Shorts, the interview series I hosted for Lincoln Network, recorded 13 November 2020.

Share this post