Interesting. Only discovered this topic last month when it briefly rose to the main page while I was browsing.
Complex characters -> require -> more time to get to know characters [...]
Complex characters -> require -> smaller roster -> require -> better justifications for not being able to give them orders than avoiding micro.
Does not seem to hold in case of e.g. Dwarf Fortress.
I like the idea of a "work shift" game object, though, which provides finer - but still indirect - control over the individuals within. As long as people are "on the clock" they would stay closer to their workplace and always prioritise work tasks (subset of all possible work tasks, as defined by the shift or mayhap profession slots within the shift?) over others.
- If the shift is too long, too distant, etc. to fulfill their needs, the workers would eventually complain about that and eventually leave the shift (sooner switch to shifts with happier participants, later accept even unemployment). Shifts might have their own intricate quality rating fed by their participant's peak happiness (in both directions), which could act as a magnet effect (in both directions) for unemployed people.
Most shifts would probably allow to define a workplace (plain old flag, mayhap with a radius as seen in Dungeon Keeper or Gene Wars?). Certain shifts may allow to define other detail orders as appropriate (pathes for soldier patrols or attack vectors, forbidden risky tasks for standby duties, etc.).
Following such an approach would allow
direct control over the shift objects, while maintaining
indirect control over the people objects.
Robot/Golem individuals might be possible for direct control of people with restricted skill sets; if their orders are complete they would just stop, mayhap start to hibernate or return to a defined depot, etc. . They could be daisy-chained into work shifts to maintain and/or organise them for regular people, to keep micro down while not needed, while the player could assume direct control over them (in groups? Those might be specialised lightweight work shifts?) as necessary.
Personally, I see no bounds to character complexity, as long as a purely-UI element exists that provides the player with high-level matchings of person<->task and as long the individual person's AI is sufficient to take care of its needs and complain sufficiently desciptively where it cannot.
Indeed, a high level of not-immediately-important complexity is what sets apart an intricate simulation from a... gamey... game. If everything has to have an immediate purpose and meaning, perfectly fine systems (as far as their performance is concerned) are quickly perceived by the player as inadequate. Players feel motivated to abandon the game long before mastering it, because the gap between understanding and utilising said understanding masterfully seems too large, the payoff too miniscule, for the player to be bothered crossing it.
See late versions of Europa Universalis or Hearts of Iron as good examples.