Working paper: reversibility as a design criterion for public infrastructure
Efficiency and reversibility trade against each other, and public infrastructure decisions almost always resolve the trade in favour of efficiency without recording that a choice was made.
This working paper proposes reversibility as an explicit criterion in the authorisation of public digital infrastructure, alongside cost, capability and security. It is circulated for comment and will be revised.
The argument
Infrastructure outlives the reasoning that produced it. A standard adopted now will govern systems still running when the people who wrote it have retired, and a default set once is inherited by everyone who follows, usually without knowing that a choice was ever available.
This gives infrastructure decisions an unusual property: their most affected constituency cannot participate in them. Future users cannot lobby, litigate or object. They will simply live in what is built. Standard consultation cannot represent them, because they are not available to consult.
Reversibility as a proxy for their interest
We argue that the most tractable way to represent an absent constituency is not to guess at its preferences but to preserve its ability to choose. That converts an unanswerable question about future values into an answerable question about present design: what would it cost, and how long would it take, to undo this?
- Can the system be discontinued, and at what stated cost?
- Can the data be exported in a documented format by a party other than the vendor?
- Can a component be replaced without replacing the whole?
- Is the specification readable by someone who did not write it?
- Are the decisions recorded, including the options rejected and why?
The trade-off, stated plainly
Reversibility costs efficiency. A system designed to be replaceable is more expensive than one designed only to work. We are not arguing that reversibility should always win. We are arguing that it should be a recorded decision rather than an unexamined default, and that the party accepting an irreversible commitment should have to say so.
Comments are invited. This paper will be revised in public, with substantive changes recorded rather than quietly applied.
