Why should the AI pick its historical area of interest rather than the best possible action?
As historical area of interest are 90% of the time the best possible action.
If it's not in the game than the game is lacking in ability to project reality.
If the game is unable to create the fundamental basis of history (which I am the first one to agree is difficult) than that need to be fixed as to avoid blandness and everything that was bad with EU3 vanilla.
The solution to this was in EU3 to add missions and historical sphere of interest (I have not myself tried the expansions as of an general outcry against EU3 and as of lack of time, though I have heard that the expansions made EU3 into one of the better Paradox games as of said historical sphere of interests).
If the game can't create the dynamics needed for said time, than you need to stuff it down into the AI's throats as to avoid the problems that otherwise arise.
At the discussion of Opium wars etc, I'd just say that the main problem there is how to make it "special". It's not an war of conquest and an war that sprung out as of trade and political basis.
How will the engine be able to with any kind of realism be able to simulate such an event in both in how the AI handles the situation and how such event fire while still avoid having it happen in many ahistorical and (if applying historical realism or historical dynamic) idiotic places.
As an programmer I know this is extremely difficult to do and such I'd consider stuffing such events down an AI's throat is better than trying to create an bland AI that going nuts all over the world.
And before commenting on earlier games, if you compare EU3 vanilla and EU2 vanilla and let the game play from the start til the end, there's more countries left as in history with more historically correct borders in EU2 than in EU3 vanilla. I can vouch for that a hundred times over.
Even if a province may end up with an strange owner during the centuries that pass in EU2 - we don't have an Burgundy taking over half of Europe.