Algorithmic Decentralization

There are two ways of decentralizing control away from a single entity:

• Governance-Based Decentralization: replace the single entity by multiple entities that govern together;

• Algorithmic Decentralization: replace the single entity by an algorithm, thus having zero entities with governing power.

Algorithmic decentralization is superior for many reasons, such as the following:

• There is no guarantee that a set of multiple entities will not misbehave in the same way that a single entity could. In fact, if the incentives for misbehavior remain the same, it is quite possible that the multiples entities will collude to misbehave. An algorithm, on the other hand is bound to execute according to its rules.

• Depending on how the multiple entities depend on each other, a governance-based decentralized system may be less resilient than a centralized one, because the failure or misbehavior of just one of the multiple entities may suffice to break the system.

• There are many theoretical results showing that it is impossible to have perfect voting or preference aggregation procedures. Most procedures are, in fact, quite flawed and do not generate outcomes that correctly reflect the wishes of the whole group.

• Truth is not a matter of opinion. It is better to rely on algorithms that provably or empirically achieve the desired outcome than to be subject to the opinions of others regarding what might or might not achieve the desired outcome.

• Governance-based decentralization tends to be costly and slow.

• Governance-based decentralization remains vulnerable to rent-seeking behaviour.

For all such reasons, The Stable Order strongly prefers algorithmic decentralization, both in our protocols and in our own organization.