Alpha Centauri 2

Sid Meier's Alpha Centauri & Alien Crossfire => Modding => Bug/Patch Discussion => Topic started by: Dio on July 14, 2015, 12:05:24 AM

Title: Information on Social Model AI
Post by: Dio on July 14, 2015, 12:05:24 AM

After confirming the location of the procedure that controls AI social model selection, I plan to determine how this operates. What changes would you like to see in how the AI controls Social Model selection?
My changes so far include:
1. Figure out why the Hive does not run Police State, Planned, Wealth, and Eudaimonia while not at war.
2. Figure out why the Gaians do not ever run Cybernetic and Green together.

Title: Re: Information on Social Model AI
Post by: Dio on July 14, 2015, 06:13:43 PM

It appears as though the AI for Yang will run Planned while they have two or more bases but not more than four bases. Once they have four bases, and are not at vendetta, they will run free market. If they are at vendetta with another faction, then they tend to start running Police State, Planned, Power and Thought Control. This combonation occured on a standard map with Transcend Difficulty.

Title: Re: Information on Social Model AI
Post by: Dio on July 28, 2015, 08:54:25 PM

One of the few instances in which I can get the unmodified AI to run Cybernetic is by having the faction possess The Cybernetic Backbone Secret Project. I have been unsucessful in getting any AI faction to run the Green Social Model. Has anyone ever seen the AI run the Green Social Model?

Title: Re: Information on Social Model AI
Post by: Nexii on August 04, 2015, 08:29:15 AM

I believe I've seen the AI run Green when I modified the benefits/penalties.  Very rare though...I think not that much is known on how the AI picks SEs. Does the AI select based on the slot (i.e. "Green") or on the bonuses/penalties (i.e. the PLANET/EFFIC that it gives)?  Or it could even be some combination, as aversions and preferences are explicitly in the faction files. You could try switching all the bonuses of Planned and Green for example to test this out.  From what I saw with custom SEs the AI does tend to avoid SEs with really bad penalties.  But I don't think there was much code to put an emphasis on PLANET, which perhaps was because the AI doesn't pollute

Title: Re: Information on Social Model AI
Post by: Dio on August 16, 2015, 12:41:23 AM

It appears having an increased Population limit from a faction bonus affects AI Social Model Selection.

Title: Re: Information on Social Model AI
Post by: Nexii on August 21, 2015, 01:28:46 AM

I was doing some ecodamage testing and was able to get the AI to select Green of their own volition.  Thought you might find that interesting.  (My Green SE is+3 planet, +1 effic, -2 ind this probably doesn't matter though).  And I also mod ecodamage rates.  I think if you do something similar though it might pick Green - give it maxed tech and a large city with all forest

Title: Re: Information on Social Model AI
Post by: Nexii on August 28, 2015, 04:57:20 AM

I can confirm through playtesting (modding ecodamage rate) that the AI factors in whether cities have ecodamage when deciding whether to switch to Green SE.  I changed the ecodivisor from 2000 down to 1 (extreme case) and most factions went Green.  Interestingly a few (Morgan, Believers, Spartans) did not even in the face of extreme ecodamage.  I'm guessing the AI priority flags also play a role.

Title: Re: Information on Social Model AI
Post by: Dio on August 28, 2015, 04:57:57 AM

I now know with certainty that the ecodamage at each base factors into the AI social model choice. I also know that diplomatic relationships have an impact on the AI Social Model choice.

Title: Re: Information on Social Model AI
Post by: Nexii on August 28, 2015, 05:00:02 AM

Yea I've seen the AI take Democracy (generally GROWTH SE) more when at Peace/Pact and Fundamentalism (generally any +MORALE SE) when at Vendetta.  I think it doesn't factor distance to enemies all that well though.  Or for that matter the relative strength of enemies with those relations.

Similarly when having massive riots the AI seems to do a halfway decent job of raising luxuries or switching to Police State.  I think in past versions sometimes riots would linger more...now I don't see it as often.

Title: Re: Information on Social Model AI
Post by: Dio on August 28, 2015, 05:03:53 AM

Quote from: Nexii on August 28, 2015, 05:00:02 AM

Yea I've seen the AI take Democracy (generally GROWTH SE) more when at Peace/Pact and Fundamentalism (generally any +MORALE SE) when at Vendetta.  I think it doesn't factor distance to enemies all that well though.  Or for that matter the relative strength of enemies with those relations.

Similarly when having massive riots the AI seems to do a halfway decent job of raising luxuries or switching to Police State.  I think in past versions sometimes riots would linger more...now I don't see it as often.

The AI increases the percentage of energy allocated to psych by 10% while one or more of its bases has a drone riot.

Title: Re: Information on Social Model AI
Post by: Nexii on September 07, 2015, 03:28:31 PM

Actually I haven't seen that the AI consistently raises PSYCH or goes to Police State, even when there is a lot of riots.  Might be related to superdrones, that the AI doesn't think a base should riot when it does.

Title: Re: Information on Social Model AI
Post by: vonbach on September 08, 2015, 12:01:33 AM

Heres a question how does the AI behave when you alter the Government and economic systems?

Title: Re: Information on Social Model AI
Post by: Nexii on September 08, 2015, 10:04:30 PM

The AI considers the benefits of custom SE choices.  Though if a faction has specific preferences that will still take precedence