AI doesn't understand city razing

I was watching an AI autoplay for playtesting, and I noticed that AI's hardly ever raze cities, even when it would be a very very sensible thing to do (such as in the ancient/classical period when a far-away city will be ungovernable). But instead the AI insists on trying to cling onto such cities, and as a result, it is beset with revolution after revolution.

For this reason, you can pretty much predict, I find, that the successful AI's are going to be the ones who manage to snag some wonders in the early game. Because the conquering AI's, although possibly initially successful and initially helping their empire stability with capturing cities, end up making their empires more unstable because they insist on trying to cling to cities instead of just razing them, taking the gold and stability bonus, and moving along.

For this reason, optimum AI play is a little 1-dimensional for the early game: build wonders. It is fairly easy to predict that the civs who build wonders (or who happen to get placed near stone or marble) are going to lead in the game. I think it would be nice if every game weren't as predictable and pre-determined, and if the warmongers could have a decent shot by being taught how to not bog themselves down in endless revolutions. (As as a result, unless Mansa Musa or some wonder-spammer is in the game, the AI global tech rate slows to a craaaaaaawl and wonders start going really late. In the game that I'm AI-autoplay testing (on noble), the leader, Suryavarman, is just now researching military science...in 1960).

So, what if, when the AI conquers a city, if it checks the RevIndex to see what the expected RevIndex rate might be like. If it's anything over +50 or so, the AI should seriously consider just razing the city (perhaps taking other things into account, like shrines/buildings/wonders, etc.) Would this be possible?

You make a good point. The AI does consider razing cities it captures based on distance, but there are some AIs which will almost never raze. Placing more emphasis on Revolution effects could help keep early conquerors from having so many issues with revolts.

Another major issue with IDW is that the AI thinks it's one of their cities due to the presence of their culture on the Citie's tile from IDW. I know from extensive testing this is seriously messing up the decision to raze by the AI.

Very true ... the AI doesn't raze a city where it has culture. That will need to be adjusted.

I don't know if it's feasible, but maybe add a "sack" option when capturing cities? Something which would reduce the population to size and provide substantial benefits to the civ doing the sacking (having a bunch of attacking civ's culture added to the city, or getting gold, workers, tech, intel, settlers, units, ?). Something that would allow "nice" leaders to attack a city but not burn it to the ground (sacked cities would still belong to the original civ after sacking). In order to make sacking a worthwhile option to the attacker, sacked cities could either have the rev index raised a lot, go into riots for a while, or even switch to a barbarian city until recaptured. While I'm thinking about it, it would also be cool to have the "puppet state" option from Dom Pedro's modcomps instated here.

So basically, in taking the city there'd be 3 options - take the city, and be forced to rebuild/defend, destroy the city (denying it to yourself and your enemy), or sack it to gain some immediate and substantial benefits, but leaving it to the original civ to clean up and restore order.

Not sure if this fits in well with the current discussion, but I thought I'd bring it up anyway. But, sacking could provide an option for peace-loving AIs who don't really want to destroy the city, but don't really want to keep it.

Sacking would be cool, but it's also complicated. What happens if you attack a city again after sacking it?

We could set it up so that you get very little or no gold the second time, but it'd still be useful as a free way to decrease the population of a city you're capturing. Capturing a recently sacked city would have to transfer a lot of RevIndex to the conqueror to fix that exploit.

Yeah, true. I was thinking that you would be unable to raze or take (or sack) a city for x amount of turns after a sack, but I guess that doesn't really make sense. Sacking a city twice would bring almost no benefit the second time, so it would probably not even be an available option for quite some time. As for taking a previously sacked city, it would make sense to have a much higher rev index and a much longer period of riots - the citizens should hate you for what you did, so taking a city soon after a sack wouldn't be as feasible as just taking it in the first place.

Maybe this can be expanded a little bit: you need x number of military units to actually take and hold the city, and if you don't have the amount, you're forced to sack or raze. That's kind of a huge gameplay change and beyond the scope of this thread, but it's an interesting option to think about.