Hello fellow Datafriends,
we need to talk about ai peace deals.
Let's look at this AI vs AI war (i changed nations to explore stuff, that's why there is the US flag). What kind of peace deal would you expect in this situation?
Mexico wants to transfer santo Domingo (in US control) and the USA wants to return their cores and conquer some extra stuff. The USA already controls 5/9 (or 5/10?) US war goals, has an overwhelming front advantage, but also some sizable debt.
So, obviously, you get a white peace:
But why? I made a 6min youtube video about the topic, which explains everything in detail if you want all the information: https://youtu.be/ONGm0eS6xn4
To cover the most important points in written form for reddit, since reddit is about reading stuff:
- If an AI nation wants to enforce a wargoal in a peace deal, they get acceptance equal to 3x(war goal value). The AI nation that is on the losing side of that enforced peace deal, loses acceptance equal to 8x(war goal value). Due to this imbalance, the AIs will never agree on something except white peace or capitulation.
In another test, i wanted to give the US the state of Utah and Mexico got a revoked US claim. Adding up the acceptance gains/loses of both actions, both sides acceptance decreased by 50 to 80 acceptance. Both sides hated the deal. As a test, i changed some values in the ai defines, such that gaining and losing a wargoal gets a 3x modifier. Now there was an exchange of wargoals possible.
- The AI is currently super likely to make white peaces, because the acceptance for white peace gets increased by 2 each month. That is to much in my point of view. This makes white peaces way to rewarding. I had scenarios, where the AI liked white peace more than gaining a minor war goal. An increasing acceptance for any peace, not just white peace, might be nice as well.
As proof of concept and as basis for more testing, i made a tiny mod which changes 3 values in the ai-defines: https://steamcommunity.com/sharedfiles/filedetails/?id=3717059839
With these changed values, the following happend in a different US-Mexican war. (Now that i think about it, i havent tested the results for the war shown above. My mistake.) This is the first time for me that i actually saw an exchange of wargoals/ the enforcement of only some war goals, so i think we are onto something here.
In general, the balancing for ai peace deals probably need way more thought and testing. I only tinkered with US-Mexican Wars and there wasn’t much population in these states and thus the war goals values are likely relatively low. The system might still block peace deals in Europe, since the value of war goals might still be to high compared to the average acceptance. But it is a very important step in the right direction and it’s relatively easy to do.