Programmers are familiar with navigating interfaces and protocols. Whether it’s ASCII, a function signature, an OOP interface, HTTP, or gRPC, these kinds of interfaces are ubiquitous, exist at all levels of abstraction, and are necessary for connecting distinct modules to form a coherent system.
I've been musing about the ways that culture, laws, and authority act as dynamic social interfaces. These structures seem to emerge as solutions to social coordination problems. Different solutions have different advantages and disadvantages.
Large groups have coordination problems
Many families, communes, and kibbutzes demonstrate a successful kind of anarcho-communism - there is a high degree of trust, and individuals get along through mutual aid and without formal codes. One explanation is to consider that the ad-hoc coordination exhibited in small groups works because of a kind of empathic simulation. Each individual has a highly accurate predictive model of social interaction within the group; given virtually any scenario and a subgroup, an individual can simulate how the subgroup will react to the situation. I speculate that the ability to accurately anticipate behavior is the essence of “trust”. (This sentiment seems core to expressions like “I don’t like him, but I trust him”).
However, we see empirically that the social cohesion of unstructured groups tends to dissolve as the number of members exceeds about 150. Robin Dunbar’s research on primate brain size and social groups suggests that these problems are due to fundamental cognitive limits on Homo sapiens ability to process information. This makes sense given the model of trust presented above. As the number of members in the group increases, the size of the set of possible interactions undergoes a combinatorial explosion. At some point, the cognitive load to grok it is too much. At this point, group trust sharply decreases. As simple experiments in game theory show, trustless coordination is expensive. More is different.
The solution to trust and coordination problems is the introduction of external structures to mediate the interaction. For example, the prisoners dilemma is easily solved in the real world by two key mediators:
- The mob boss, who decrees that finks will sleep with the fishes
- The mobster’s code, which suggests that “snitches get stitches”
Each of these structures creates a kind of transitive trust for the prisoners. The mob boss has a monopoly on use of force; since his decrees have guaranteed enforcement, he can trust anyone to do what he says. Since Prisoner A trusts the outcome of the mob boss’s decree, and the mob boss trusts that Prisoner B must do as he says, Prisoner A can trust Prisoner B not to rat. Likewise with the mobster’s code. More generally, these two structures can be thought of as analogous to government fiat (centralized enforcement) and culture (distributed enforcement). Each structure takes a complex web of interactions and flattens it into a simple set of rules - hence the term, “social interface”.
Different interfaces, different advantages
The level of formality and mechanism of enforcement of these social interfaces create tradeoffs. A very formal social interface (a written law or code of conduct) is clearly understood and enforced. However, a less formal social interface (a culture) permits more nuance and adaptation to changing circumstances. Central enforcement creates common understanding and prevents vigilantism. Distributed enforcement may be more specialized to local circumstances.
It is also interesting to consider that some styles of social interfaces may be more stable at different group sizes. What kinds of interfaces scale? Which ones don’t?
Distance and varied systems introduce friction and reduce trust. Local communities typically have high-trust relationships because of common culture and higher accountability for local externalities (the “don’t shit where you eat” principle). Trust is lower between more distant polities. Culture and legal structures differ more. Individuals have a less predictive mental model for behavior. Federated models are necessary.
Impersonal and Unequal
It seems clear that for humans to collaborate at the massive scales that we do today, we need social interfaces that simplify the ever-more-complex trust issues that have emerged. However, there are also clear problems with the interfaces we have today. Creating interfaces that are consistent means the circumstances of individuals have to be abstracted away. Enforcement of interfaces traditionally requires monopoly on force, which means that hierarchies must be created and some individuals are given more power. As our societies become larger, they seem to create more inequality and social alienation.
What comes next?
Past innovations have let us coordinate beyond our individual cognitive limits. Some argue that the written word built ancient empires and that the printing press empowered nation-states. We now have a global internet, instant video call, and trustless distributed ledgers. Some LWers are thinking about prediction markets. Scott Alexander has posted a number of interesting thought experiments in alternative governance.
What failed ideas were just too early? What new and better social modes can we create?