As a secondary topic close to my heart, I like to touch on artificial intelligence and a concept I have been toying with for some time has been a video game in which you lead a particular group of people from the stone age to the early modern age (before industrialism comes into play) in a real-time strategy format.
The idea is to have a fairly low level game play, akin to Dwarf Fortress, where you lead people with a growing population and technological capability. As the game progresses, more of your society is automated to allow you to focus on the "bigger picture" thereby maintaining the same APM (actions per minute by the user, a metric used to measure how much a user has to be moving about clicking the mouse to play the game competently) when the player has ten people in the nomadic stone age period and even when he is leading armies of thousands of soldiers in the early modern age.
It is the automation that is of most interest to me, intellectually, versus the grunt work of developing the whole game (graphics engine, game engine etc).
So in that vein I wanted to discuss a bit about trying to automate the economy to allow the user to handle less "industries".
The Game Economy
In the game, the player controls a group of people. The technology tree is split into three categories (fantasy, steampunk and mundane) to represent different paths for the player to take. No matter the path the general idea is that technology gives the user access to produce a new type of product or to automate their economy in some way.
A society is then composed of a number of industries, a standing military and the government coffers (which includes non-money items such as a stockpile of weapons in case of war). So for instance, once you move beyond the stone age an early settlement might be represented by:
Oat Industry: 6 workers producing 50 oats a year
Horse Industry: 2 workers producing 1 horse a year
Military: no standing soldiers
A user would then think to open up a new industry/modify a current industry, assign labour to it and watch the output level change. The user could also train people into standing soldiers or reserve soldiers, trading labour time to increase the military skill of people (all the way to putting a person in the military permanently).
An Automated Economy
As an experiment, I attempted to create a "perfectly rational" person who decided on his own what to produce, whether to change his labour and at what price to sell his products.
The AI experiment had the following format:
A constant population of 100 individuals all of whom start off as farmers who produce "food".
Four types of raw materials exist.
Four types of "common wares" exist, each requiring one of the four types of raw material to produce.
Four types of "luxury wares" exist, each requiring one of the four types of raw material to produce.
The decision model of each person, using nothing more than a script-AI, was the following:
Attempt to purchase enough food to eat for the turn. If he produces food, then reserve that much food for this purpose.
Attempt to purchase raw material for my job if I have a job. (Always choose the lowest price)
Attempt to purchase common wares with left over money; I only need one of each type. (Common wares break down after awhile)
Attempt to purchase luxury wares with left over money; I only need one of each type. (Luxury goods break down after awhile)
Produce as much of my product as possible (farmers always have max production, others might run out of raw material)
At the end of the day change the price of my product if necessary (all prices are in food, mirroring ancient Egypt):
If I couldn't sell all of my daily production of product, lower the price.
If I sell exactly my daily production of product, keep price constant.
If I sell out my daily production of product, raise the price.
Depending on how profitable I am, I can choose to switch my job.
The outcome was usually that one person would become incredibly rich, a few select people selling wares would become rich and everyone else would be in abject poverty. This isn't exactly an outcome that is useful for the player, as anybody would much rather run the economy themselves and it would still be more efficient.
There are a few ideas to help mitigate the problems.
General Improvements
Individuals should have a list of products they may wish to produce. They then swap labour time, at the margin, between the most profitable output and the least profitable output. When switching into an industry where it requires materials to produce the product (such as requiring clay to create pottery), then the individual may decide to acquire the material themselves rather than purchasing it.
Structural Differences
Early and ancient economic systems were not capitalist or free market arrangements and after seeing the result of some of the issues with low number of products and constant population (both of which were characteristic of ancient societies), it might be that automation technology should more closely match ancient solutions rather than using modern free market solutions (which work substantially better in a modern society).
For instance, the first farming arrangements in which people did not simply stash items for the whole human herd, was to have family stashes with a "big man" who tried to distribute output fairly. Tribal arrangements were an evolution of the "big man" to a permanent seat of authority who arranged the redistribution (this would be represented by the player deciding how to distribute goods and how much to keep for public purposes).
Ancient Egypt, a more complex society than tribal arrangements, had state-run farming alongside estates who were tasked with running more farms who had to pay a harvest tax. The food then was used to pay for the Pharaoh's staff and then it was used to pay for public works (such as the pyramids), as a sort of "yearly stimulus".
Generally the idea would be that the automation might be closer to what a player would wish to do rather than more realistic. That is, a player assigns farmland to certain estates who then automate the process of food production, tweaking different levels of food according to imperial decree to meet your harvest tax and then they pocket the difference (with some corruption perhaps). This frees the player from having to zone out land for farming, expanding farms, collecting people to work the farms and distributing the food to the workers (picture how one might expand an industry in a real time strategy game or Dwarf Fortress, where you build workers, then build farms, then create stockpiles and so on). The player simply receives a sum of food which can then be used to pay for public works and the military.
No comments:
Post a Comment