Paper Summary
Clark, Golder, and Golder formalize Hirschman’s framework as an exit, voice, and loyalty game. The central move is to specify when voice will actually matter. A user’s complaint is not automatically powerful; it depends on whether the institution needs the user’s loyalty and whether the user has a credible exit option. When exit is not credible, voice can be ignored. When the institution is not dependent on the user, voice can also be ignored.
This game-theoretic version sharpens Hirschman’s political intuition: voice gains force from the background possibility of exit. Without that possibility, “voice” can become cheap consultation, complaint capture, or symbolic participation.
Relevance
For Machines That Let You Leave, this is directly useful. Feedback inside a personal AI system is weak if the user cannot realistically leave. Voice mechanisms only become meaningful when paired with credible exit: data portability, interoperability, usable alternatives, reduced lock-in, and the ability to degrade or disable the system without losing one’s practical life infrastructure.