I'm a long time lurker here, and so sorry for barging in with a post which probably is incorrect/makes no sense
But I'm seeing Necratoid make an argument that 'to prevent the positive feedback effect, one should just increase the intelligence of the AI until it is at a level with the player' again and again, and not being given any form of response. The problem with the Civilisation games, though, as has been said on this thread, is that when one player gets ahead any distance they tend to stay ahead. This applies to both the player and the AI (which requires a lot of skill to catch up with and surpass on higher difficulties, becoming nigh on impossible if it's improved to the skill of a player). The important reason that improving the AI simply would not work, though is this: there are almost infinite numbers of variables existing, both in game and out of game.
For instance, assume that Koshling programmed a perfectly balanced AI that would, with specific starting conditions, give him a challenge through the entire game without steamrolling him. Great! Now if he gives it to someone else, with a little less skill in the game (me, for instance), they'll probably find that the AI begins to get an edge on them. As the tech difference increases, and the AI begins to conquer the world, it rapidly becomes clear that they will win and the game is not worth continuing. Conversely, if the same scenario is given to a player with slightly more skill, they can get a small lead and increase it. Increasing a difficulty level, though, would likely result in a loss as the AI gains power.
Worse still, if Koshling regenerated the map, and played on a poorer starting position, he'd find himself unable to keep up (and vice versa for better position). Even random factors (does the AI declare war when its demand is refused?) during the game could swing the balance of power and cause one civ to begin to steamroll the others.
This issue is more prevalent in C2C than vanilla cIV, I'm guessing, simply because of the longer games. The point is that while a better AI will decrease the likelihood of the player steamrolling the AIs, but not reduce the overall chance of a steamroll significantly. This is the problem with a vicious circle, which needs some external factor to prevent it from getting out of control.
As for what you'd get for being further ahead in research, you have all the benefits that being ahead in tech brings: improved infrastructure, the ability to crush spearmen against tanks, etc. Just as normally a research will be nigh on useless to you if you don't do anything with it, so in this circumstance the increased power will allow you to do things you otherwise couldn't: finish conquering the continent, get a head start on building factories, make a colony on the moon first, etc. Just bear in mind that this (metaphorical) 'golden age' where you're a massive way ahead of your opponents will not last forever. If I've understood the system, though, they'll never overtake you, or end up in a better position relative to you than had you researched less well through this system. So you are still gaining advantages from being ahead, but the game rewards you
more for what you're doing recently than what you did 4000 years ago: as should be the case, really.
I'll go back to lurking now