That's the thing, though.
In EU there are two very different kind of players - more roleplayers, and play-to-conquer players. In EU3 the AI is basically programed to act like the later (well, to try to). Which is immensely frustrating for the people more into the roleplaying end of things who want a world with fewer wars, more diplomacy, and generally not a world war every generation.
But if it were the reverse, competitive players would feel the AI cannot challenge them.
He gets it. I am firmly in the RP camp.
And I disagree again.

There is no contradiction. That there should be two distinct AI programs is just a horrible gameplay concept. The AI in EU3 is not programmed to act like a conquering player, it is merely programmed to look for advantages. It will not do long term planning to take you down - it just looks at its surroundings, calculates how much they fear you (threat perception), how much they like you (relations) and how the balance of forces is (military strength) and then makes a decision. Often times this is a decision to not attack you, even though a conquering human player would attack you because he would ignore relations and would only take the balance of military forces into account. When the AI decides there is no one to attack, it stays at peace and spends money on peaceful things. The conquering player would instead keep spending on military because he KNOWS he is going to attack someone soon. The AI does not really think about that, it just holds its military at whatever level is thinks is okay for peacetime.
What I think would make sense is to have difficulty system, which mainly influences the AI's opinion and threat perception values versus the player by means of static modifiers. I know EU3 had already a sort of "threat perception" because each country had a list of perceived threats.
Easy difficulty -> All AI nations get +20 to opinion vs human player, and -30 on threat perception. (Or perhaps: a cap on how threatening they perceive you, and a cap on how much they can hate you.) You can still get attacked, but only if they really hate you. AI vs AI is unchanged.
Normal difficulty -> no modifiers. AI vs AI is unchanged, and AI vs human is "default".
Tough difficulty -> -20 to opinion vs human player, +30 to threat perception. They always seek for opportunities to cut you down. AI empires ally against you more easily. Lots of warfare against you. AI vs AI is unchanged.
The reason there shouldn't be different AI scripts is that it would (I think) make the game harder to develop. You'd have not one but two components in the AI system that need to be developed, tuned, tested and so on. So IMHO it does not make sense, to ask for two AI modes, one where they are all peaceful vs one another and no one blobs, and one where they conspire to attack one another and lots of blobs form. Better to have it all be the same, and only exempt the player in some ways by giving him advantages in how they perceive him. If the game still ends up in a way where all the small nations are eaten up by 1550 then there is something wrong with the game rules, that make conquest too easy, not with the AI.
The main problem I see is, eventually you will get into a situation where even on easy mode you are seen by the AI as a REALLY weak, and REALLY tempting target. Even with caps and modifiers you will get into their crosshairs eventually. You can't rule that out even if you play on easy.