[Freeciv-Dev] Re: (PR#15249) Aggressive AI Could Fight Itself to Obsolence

["Per I. Mathisen" <per@xxxxxxxxxxx>]

<URL: http://bugs.freeciv.org/Ticket/Display.html?id=15249 >

The current tax code follows a simple heuristic. While we can, maximally
prioritize science, putting only as much stress on taxes as we need to pay
upkeep. If, however, we are under the minimum reserve fund size, we
increase taxes by an additional 10% to gradually increase this fund.

I think this part of the code is correct, and, by virtue of its
simplicity, good.

Benedict commented: "There are circumstances in which the AI does not need
much tax income, but the AI will attempt to maximise its tax income." This
must be when preparing for war. I agree with Jason's comment that we
should not mindlessly be reversing tax decisions, here. I think making the
war preparation part of the tax code smarter would be a good idea.
However, this probably depends on the AI acquiring a better grasp of its
military needs first. Maybe.

We also need some way to figure out whether to increase luxuries to offset
unhappiness in cities. This is a notoriously difficult problem to solve,
which easily hogs the CPU. I think some kind of risk-willingness could
solve this to some approximation. We see the possibility of cities falling
going into unrest and government falling as a random chance based on ratio
of entertainers and unhappy citizens to other citizens in all cities, and
try to remedy this chance by increasing the luxury rate depending on how
far we are leading or falling behind the tech race. An AI which is far
ahead in science can afford to lag a bit in order to avoid spending lots
of resources in cities to quell unhappiness, while an AI that is falling
behind can risk having to spend more resources on keeping happiness.

I think the patch in this ticket is not obviously more correct than the
existing code, and I do not see any significant improvements while running
some simple AI measurements. So I am unconvinced that the patch is a good
idea.

  - Per