AI Issue

[AngleWyrm]

I proposed a treaty where we can colonize each other's systems. They countered with No Hostilities in Unclaimed Space. I accepted.

So a couple turns down the road, and they lose ships to the minefields guarding my systems--And now they're annoyed?

It was the AI's idea not to allow cooperation, but now it is upset about it's own decision. This could use a little more thinking through.

If the AI is going to play xenophobic treaties, then it should also recognize the outcome of said treaties.

[Elsemeravin]

Seems kind of logical for ones to be annoyed when making mistakes. I personally am in such case.

[AngleWyrm]

Maybe...On the one hand, it seems self-defeating, contradictory, possibly even like playing against a less skilled opponant.

But on the other hand, if the enemy were to successfully invade my territory under such a treaty, it might be perfectly happy to continue. As long as it's getting over on me, and not the other way around.

It might be working after all, it just needs some successful follow through assaults to make the thing pay off. Last time, it died on my warp point mines, and then quit the treaty.

Is it possible to have the AI check treaties, and amass territorial invasions based on treaty elements?

And can the AI work up a multi-turn plan? Like (Research Mines, Build MinSweeper, Add to Assault Fleet, Invade x,y)

[Aiken]

SE5 AI is just a set of rules "If X happened then change state to Y and do Z". You're demenading it to act like a human, though. SE5 AI is indifferent to its past decision and not evaluate complex situations with large amount of ambiguous factors. Such behaviour requires a very sophisticated AI which is difficult to mod (not se5 way).

[Slick]

It's really a shame that the AI can't have any long term ideas, goals, agreements, diplomacy, etc. They constantly evaluate everything each turn and act based on current situations, regardless of what happened the previous turn or diplompatic agreements including those which it proposed in the first place. It is also a shame that this sort of thing is not able to be modded.

The closest thing that AI has is it's change in "mood" toward other empires, it's happiness levels and the hard-coded underlying 4X strategy.

I'd so love for the AI to even simply get upset if I don't answer a request. Right now, if the AI asks me to evacuate a system, I can say YES and he'll be happier with me even if I don't do it, I can say NO and he'll be angrier with me even if I do evacuate, or I can not respond and do whatever I want with little effect other than my annoyance to the AI that I'm still in the system he wants. All this follows from the fact that the AI doesn't remember what peviously happened and just looks at time-now.

[Makinus]

The AI remembers me of an alien empire in a sci-fi novel, where the Aliens are extremely advanced, have much more systmes than humankind, huge fleets and dominate the human race but are only capable of focusing in the "now", being incapable of thinking too far (they do a little) in terms of future and past, and are incapable to imagine that some humans are thinking and planning a rebellion a few decades in the future...

[AngleWyrm]

call Sys_Set_AI_Storage_Long(sys_long_Player_ID, 1, 1, gbl_ai_state_attack_system)

To restate my question in more of a computer language than an interpersonal one:

I would like to use a couple AI_Storage registers; one to hold the date of plan activation, one to hold the step number in a masterplan sequence. Sort of a mini-state within the state machine. Then I would call an indexed set of functions based on step number.

Sys_Set_AI_Storage_Long: boolean
plr_index: long
elem_index: long
list_index: long
new_long: long

Sys_Get_AI_Storage_Long: long
plr_index: long
elem_index: long
list_index: long

What elements and lists are available/in use? Can I just make them up?