Hirschman's *Exit, Voice, and Loyalty* (1970) is usually summarised as a taxonomy.
It is better read as a warning about information. When an organisation declines,
its members have two responses: leave, or complain. These are not equivalent —
they carry different information and they have different effects on the
organisation's ability to recover.
Exit destroys the feedback the institution needed in order to fix itself.
The person who leaves quietly takes their diagnosis with them. They knew what was
wrong. They were willing to tell you. Instead they voted with their feet, and you
recorded a churn event: a number, going down, which looks like a metric to optimise
rather than evidence of a failure you can no longer see.
Loyalty matters because it buys time for voice to work. Without loyalty there is
only exit, and exit is silent.
Bridging away from X1 is exit. Selling is exit. Going quiet in the group is exit.
Every one of these is recorded somewhere as a neutral fact — a bridge transaction,
a wallet balance, a silent member — and none of them carries the reason.
The protocol sees the exit and not the complaint. That is precisely the structure
Hirschman warned about: the feedback channel is destroyed by the act of leaving,
so the organisation optimises against a metric that has already lost its meaning.
Before celebrating users "voting with their feet" and bridging to X1, ask what
feedback they took with them — their complaints were free market research you just
lost.
Build a validator post-mortem requirement for slashing events, not just automatic
penalties, because penalties are *exit* events and post-mortems are *voice* events.
When someone bridges off X1, does anything capture why — or do we only see the balance change after the reason is gone?