Metagaming for Constitutional Design

In game theory a metagame is a game about another game, generally with the objective of finding the best rules or design for that other game. For games like chess or go it is clear that the games did not begin with all the rules and design they has today, but as simpler games, the rules and design for which evolved over the centuries. What the players were doing over the centuries was playing a metagame of finding improvements in the game rules and designs to make play more satisfactory.

Metagaming is everywhere in the world of strategic decision-making, in society, economics, politics, and engineering. In particular it is involved in the evolving designs of constitutions of government and legal institutions. Normative politics is largely a matter of metagaming.

We can carry the process to another level, playing a metagame in which the objective are better rules and designs of the metagame of finding better designs for political constitutions, which are themselves metagames for designing laws and legal institutions. Metagames can even loop back and apply to higher level metagames in a system of them. That is what provisions for amendments in political constitutions do.

Like any games, metagames can be played well or badly, but they can also be analyzed scientifically, or even solved mathematically, perhaps with the help of computer simulation models. Just as chess-playing programs can now regularly beat human opponents, so we can anticipate a day when constitution-writing programs may generate better constitutions than conventions of human beings can design. Constitution-writing software may not be within reach today, but with a concerted, well-funded effort, we can expect to achieve such software in the not too distant future that will outperform human beings.

We can speculate about what such software is likely to yield. During the course of centuries of constitutional design by humans, and the testing of those designs in real-world conditions, certain patterns can be discerned that do not seem to be subject to the vagaries of history or culture. As in the design of buildings, there is some room for taste or even whimsy, but ultimately there do seem to be recurrent and stable principles of design that we might expect to emerge from the evolution and adaptation of such designs for not only human beings of every culture, but even perhaps for other species of similarly capable and semi-autonomous social beings, anywhere in the Universe. The long-held dream of a science of politics may be within reach with the automation of constitutional design. This is one of the results that can be expected from the science of pynthantics.

The most promising approach to developing constitution-writer software is likely to use some form of genetic algorithms, that split and recombine specification components, which are not necessarily words in a natural language, and then tests each combination with simulated societies in which members use it to try to optimize their purposes, and protect their rights. Part of any such simulation is likely to include "clever lawyers" who try to use arguments to get decisions that deviate from "original understanding". An objective of the software would be to design components that are highly resistant to such usurptive efforts, without producing an excessively large document that specifies too many details. If the product is not written in a natural language, then there would need to be a translator function that would do that, so that it could be used by human beings.

So the programmer's ideological preferences do not necessarily affect the design, at least not in a predictable way. It is not likely to be practical or safe to test designs using real societies, so we have to find ways to simulate societies and their members, in which the abstract actors behave enough like human beings to test the design in the way a human society would, but much faster, allowing the selection of better designs to proceed to completion in a reasonable amount of time and at an affordable cost.

See also:

•  Constitutional Design, Jon Roland