I'm well versed in AI programming and how common it is to use cheating on the part of the AI instead of programming a better AI (more cost effective).
As a student in French Literature, I'm a total n00bette in what concerns A.I. programming. My opinions, here, are consequently only worth 2 cents.
1. Micah, it appears to me, points in the right direction : COST. It's much cheaper & faster, in terms of paying person-hours of A.I. design & coding and of overall development time, to invest in cheap cheats, instead of hiring specialized A.I.-programming human geniuses who will spend a lot of 12-hour workdays designing & coding.
2. When we play against comps & bots, we're not really playing against relatively dumb machines : we're playing against human programmers.
If we beat the A.I., it's not really because we human gamers are "too smart", it's mostly because the human programmers were not given the time (paid person-hours) to refine & test more complex A.I. strategies & tactics.
We never really play against autonomous Cybernetic Intelligence. We play against human coders.
3. There's also the limitations of the customer's computer. Even if the human developers & coders were given the time to design & program very sophisticated A.I. adversaries, it would lead to a commercial disaster :
Not only because it would have cost a lot in salaries, but also because 99.9% of the customers' computers would not be complex & fast enough (per second) to process the numerous, sophisticated instructions, which would require a lot of computations.
Games would have to be designed & optimized for full use by dual-core & quad-core machines, and even then, it would not be enough. There are no "petaflop" computers available, yet, for gamers (unless they work in some Pentagon basement).
4. Everybody who knows about the Galactic Civilization 2 series knows that it has the best non-cheating A.I. in PC games of Strategy. The series offers many levels of difficulty, and the A.I. gets bonuses only at the topmost levels. The principal A.I. designer of that series could afford to invest man-hours of his time into refining the A.I. because he owns the company (Stardock) and enjoys doing A.I. programming. Most other game developers do not have the cash and the time to do that.
Consequently, those of you humans who believe that you are "too smart" in comparison to comps & bots do not take into account the limitations of A.I.-programming investment ... and of your relatively too slow & simplistic home computers.
/my 2 ¢