- Anarchism: individuals or small groups make independent decisions with no way to enforce* their decisions
- Direct Democracy: People vote on decisions, and enforce the outcome of the election
- Indirect Democracy: People vote for individuals to make decisions, and enforce those individuals' decisions
- Privilege: Certain individuals (somehow) become privileged to make enforced decisions; with no effective mechanism to arbitrarily withdraw that privilege.
(The economic privilege of capitalism is not anarchism, it really is privilege, private law. Without the right to enforce his economic decisions, Bill Gates' $35 billion doesn't mean anything more than scoring 35 billion points in Asteroids. The "freedom" touted by Randians, Libertarians and "anarcho"-capitalists is nothing more than the freedom to acquire and exercise privilege — but of course the first thing that someone will do with privilege is arbitrarily determine who in the next generation receives their privilege, making the acquisition of privilege not free.)
The decision-making problem is not primarily about expertise, the accurate understanding of objective reality (although expertise is necessary). It's about motivation. In Prisoner's Dilemma situations, if I don't have the assurance that it's in another's self-interest to cooperate, then I cannot cooperate myself. If I cooperate then the other agent will simply exploit my own cooperation; I will end up worse off than if I defected.
Anarchism — decision making without any form of coercion — can achieve mutual cooperation only when the agents can play Tit-for-Tat. When Tit-for-Tat is inoperative, anarchism can achieve mutual cooperation only if all agents always choose to act cooperatively even when its physically possible to defect and exploit others' cooperation*. Even if a small number of agents routinely defect without punishment, they will eventually dominate the population through positive feedback.
Tit-for-Tat works only when the threat of future defection is operative, and when the number of future iterations is unknown. This limitation — not the limitation of imperfect knowledge about objective reality — places the upper bound on the time-scales of decision making: The "lag" between iterations not only cannot exceed the lifetime of the agents, but must be a small fraction of their lifetimes to preserve the uncertainty of the number of iterations.
Furthermore, Tit-for-Tat can work only when a betraying agent can be decisively identified; it does not good to punish an agent who did not betray. As endeavors become more complicated, with more participating agents, individual responsibility becomes more diffused and more difficult to identify. Again, this diffusion places an upper bound on the complexity of endeavors.
The other method to achieve mutual cooperation is to take the decision-making "out of the game", away from the participating agents, and place it in the hands of someone who has a "meta-interest" in helping the participating achieve mutual cooperation. In other words, by creating a government with individuals who do not have an interest in any individual agents' self-interest, but do have an interest in the mutual self-interest of all the agents. If the government can remain disinterested in individual agents' self-interest, they can facilitate endeavors to the limit of time and complexity where the mutually beneficial strategies can be known, which exceeds the Tit-for-Tat limitations and, with advanced information technology, has no discernible upper limit.
This task maintaining a disinterested government is difficult, and at best meta-stable. It may be impossible. But we know that to simply abandon the attempt to create such a government dooms us to restricting the scope of our endeavors to those that can be managed by anarchism.