I second that a "recorder" might be desirable...I also have some ideas about how this might work. I am not much into programming, so I do not know if it would work in practice or not (and I also do not know the mechanics in GalCiv and, shame on me, I actually did not test the beta of LH yet). Here is my naive ideas:
1. let the game reserve some resources for a "recorder" (say 2-5GB on your hard-drive)
2. set a number of games the game is "watching"
3. let the AI evaluate your strategy and implement a counter in the next game or even replicate your (!) strategy for a specific faction (evaluation could also be done if I am not playing the game)
4. give feedback how well the AI did in countering your efforts or using your own strategies
5. repeat feedback loop every 5 (10, 20...) games.
I guess (or hope) a very basic mechanism could be implemented by giving the AI a core set of strategic rules and then let it choose based on individual play-style with some simple Bayesian-updating heuristics (works wonders in reducing SPAM e-mails). In principle that would mean every player ended up with its own "AI" or at least an (optional) refinement that is tailored to his own style of playing.
Besides the technical issues, I see the following problems:
- you need a "reset" button, because players might delieberately "confuse" the AI in the watching phase or the AI will be locked-in if players fundamentally altered their play-style, the AI will also end up totally wrecked on some occasions (remember its a heuristic).
- updating and patching will become a mess
- the rich content and to high number of variables might prevent the use of such a system.
In essence such a system would be based on stochastics rather than a set of pre-determined rules and it gets better with every game it "records".
Best,
Kobito
PS: I got lost looking for a concept how the AI actually works in the Dev Journals. I remember having seen something where frogboy commented on his evaluation of the AI and how to improve (a video, maybe). The search function was not helpful, because it had so many hits about people's feedback for the AI. Can anyone provide a link?