Decision making

[Metapolitical ramble follows]

What's the optimum method of decision making on a national level? Or maybe, what's the optimum number of parties? The premise is that one party tables a proposition, the other parties vote on it. The proposition has a vector of values and its compatibility with a party is the dot product of the proposition vector and the party vector. The value of the proposition could be put as the compatibility between the proposition and the population value vector at the time when the proposition is in effect. Or the long-term propagation of the population? Is there a good reduction for the goodness of a decision?

To simulate that, you'd have to generate proposition vectors, run them through the decision making filter, and add up the value of passed propositions. A one-party system would rubber stamp every proposition, ending up with the mean of the proposition values. So it would really be more controlled by the proposition generation system than the decision making filter. Multi-party systems would work like a one-party system when one party has a absolute majority, but would actually filter propositions in other cases. A two-party system is most likely to end up with one party having an absolute majority, and it's easier to deadlock one as well.

The generation of propositions is another can of worms: how do you generate good propositions and when does a party agree to table them? How many propositions do you go through in a year? Who generates them and who must agree to them?

The value of the decisions of a one-party system could be described as N*avg(V) where N is the amount of propositions and avg(V) is the mean value of the propositions. You could also state the result of a one-party system simply as sum(V). To improve the functioning of a one-party system, you'd have to raise the amount of propositions flowing through it while maintaining the mean value or the mean value of the propositions while maintaining the amount of them, or both of the mean value and the amount of propositions. This pretty much amounts to improving the quality and throughput of the proposition generation apparatus. Or in other words, a one-party system is really a proposition generation apparatus with some pomp and ceremony on rubber stamping the generated propositions.

The value formula for decision making system that acts as a proposition filter would then be sum(filter(V)), where filter(V) is the set of propositions that pass the decision making system. The idea behind the filter is that sum(filter(V)) should be higher than sum(V). Here the proposition generation apparatus is less important than in a one-party system as propositions can be discarded. The filter biases the decisions by letting through only those propositions that align to its values. If the filter is not representative of population values (the goodness vector), it'll be actively harmful. So being able to swap out the filter and replace it with a better one becomes an important operation.

The parties of a multi-party system can be thought of as swappable filters. If there's a filter that matches population values better than the current one, it should be swapped in. If all filters are out of alignment with the population values, the swap operation becomes ineffective.

If the proposition generator of a one-party system gets out of alignment with the population values, the one-party system will start making lots of bad decisions. In a filter system, a faulty generator can only cause inaction. A faulty filter with a good generator will also cause inaction, while a one-party system would make lots of good decisions during that time. To make a filter system produce bad decisions, both the filter and the generator must be faulty (and in the same alignment).

For a one-party system to work well, the proposition generator needs to be swappable. A filter system can function at a lower throughput with a faulty generator, as long as the filter is swappable. For maximum throughput in a filter system, the generator and the filter should be swappable.

In a situation where the proposition generator is generating proven solutions, a one-party system would be more efficient than a filter system. If the propositions are a mixed bag, a one-party system would do bad decisions alongside the good, whereas a filter system would start rejecting the bad decisions. This is kind of iffy though, as the values of the propositions aren't always known.

Using light as an analog: When light is the right color, a filter will only make it dimmer. When the color of the light is wrong, a filter will help.

Now I wonder if anyone has written a simulator for decision making systems and run it through an optimization algorithm to find the optimums of that simulation...