Let us formally define the problem statement.
Say there are N (>1) agents who are to take M (>1) decisions over a period of time P. For simplicity, say M < inf, i.e., the number of decisions to be taken is known at the start of the period. The decisions taken should be
- fair over the entire course P,
- incorporate a measure of the agent’s preference for some decisions over others.
What follows is the procedure that I think is pretty neat. For lack of a creative name, I am calling this the Repeated Selection Consensus Protocol.
Every agent starts off with an equal number of tokens, K. There is three phases to every decision m of M.
The agent with the highest preference for a particular decision gets to win the result but at a cost. The agents with lower stakes in the decision are compensated with the tokens of the winner so they may use them in a future, more personally affable decision.
All the voting takes place in secret. Hence after the first vote, none of the agents have an equal number of tokens and it is not possible to figure out how much an agent has without a discussion among all agents.
I am of the opinion that this is fair. A random selection is unfair to the person who desperately wants a result and is willing to forgo their future decision making powers for it. This also puts the onus of bidding smartly on the agent, knowing that there is a cost to winning a decision, balancing the current want with a future need.
Consider the case where a minority representative wants to pass a bill that is extremely important to them but is unable to because of lack of a majority, one that is dependent on representatives who may not necessarily be motivated to pass it. This protocol should move away from the effects of majority rule.
What parts of the picture I am missing? Are there related material I can read up on that touches on this subject?
Send in your thoughts at email@example.com.