Why Reliability Is a Moral Property
Most harm in the world is caused by people who meant well.
They followed the rules. They acted reasonably. They did what made sense locally, in the moment, with the information they had.
And still, something broke.
When this happens, we look for intent. We ask who wanted what, who believed what, who tried hard enough. We tell stories about motivation because they are comforting. They let us believe that outcomes are proportional to character.
They aren’t.
Systems don’t work that way.
A system does not care what you meant. It only reflects what you built, what you tolerated, what you ignored, and what you allowed to repeat. At small scales, intent can matter. At scale, it loses importance. What remains are patterns.
This is not a cynical claim. It’s a mechanical one.
As Stafford Beer put it:
“The purpose of a system is what it does.”
(POSIWID)
If a bridge collapses regularly, we don’t ask whether the engineers were sincere. If a service fails under load, we don’t praise the team for trying. We fix the design, or we accept the consequences of not fixing it. Over time, reliability becomes the only thing that speaks.
The same is true outside of machines.
Institutions that “mostly work”. Processes that fail only for some people. Decisions that are reasonable in isolation and destructive in aggregate.
None of these require bad actors. They require consistency.
Reliability is not about perfection. It is about keeping promises under stress. When a system repeatedly fails to do that, the failure becomes ethical, not technical. Not because anyone is evil, but because predictability is how trust survives scale.
This creates discomfort. It means good intentions are not always enough. It means responsibility does not end where control becomes incomplete. It means we can be implicated without being guilty, and responsible without being fully responsible.
We resist this idea because it feels unfair. And it is, in a way. But fairness is not the question systems answer. Systems answer different questions: What happens when this repeats? Who bears the cost? What breaks first? What is quietly normalized?
This is not about assigning blame. It is about refusing denial.
It is an attempt to treat reliability as what it actually is: a moral property that emerges when our actions outlive our intentions.