Happy to share a new economics paper which should be of interest to researchers working on foundational questions in distributed consensus. The paper has implications for canonical blockchain-related impossibility results in computer science such as Bracha-Toueg, and shows the connection between them and the Revelation Principle in economics.
A one-page summary is available from the site. There are a few high-level take-aways:
- incentive compatibility is possible decentralized systems, contra Hurwicz (1972)
- the solution requires privacy on the base-layer (non-scalar reports) and forces a trade-off between incentive compatibility and ex post verifiability.
- the strategy that induces incentive compatibility cannot be implemented by revelation-equivalent (direct) mechanisms, which blocks the universal applicability of several canonical results in computer science which share the structural assumptions of the Revelation Principle.
It is worth noting that the paper does not refute or weaken any classical impossibility results in computer science of economics. The key observation is that when a base mechanism (blockchain) preserves privacy and compartmentalization, agents may rationally coordinate in ways that are unobservable to the protocol, but can be indirectly observed in decentralized systems. This creates incentives for cooperation that cannot be created in standard direct-mechanism or verification-based (staking-and-slashing) models.
The solution is counterintuitive but lovely. A good analogy would be a Parliament that optimized its own security costs by inviting players to step outside the building whenever voting on changes that would adjust overall spending on it. Even though the mechanism cannot observe players stepping outside, it can indirectly intuit and optimize its own security costs by forcing a subset of decisions which could affect it to be made in the absence of its enforcement guarantees.
Happy to clarify scope or assumptions if useful.
