I personally have a fairly complex set of "rules" bosed in large part on several things, including the point in the game, what science I'm researching, where my science is (how many beakers I have and how many to go) for critical advances, being pressured by the AI in the science dept, the number and size of existing cites, how much gold I have (can I affort to buy a settler to kill the city before size 2, if it's in a bad spot), is an AI unit nearby (and thus might pop the hut if I don't), the availability of supporting units (es. dips) to kill/bribe barbs that come out, and of course, can I affort to lose the unit doing the popping.
Usually, I can affort to lose the unit, unless it's a dip on teh way to do a specific mission.... those cases, esp. in early game, tend to let a later unit get the glory of the pop.
In the first few game turns, I will pop every hut, since the odds of units seem higher (I have not done a full parametric test).
The most devastating results in early game are usually trash advances (like HBR and Warrior Code), as sell as Advanced Tribes (usually in bad spots, high corruption, and happiness killers). The "best" result IMHO is a Nomad. Smash has posted some very interesting ideas about this and the use of NONE settlers vis-a-vis huts. Gold is a kind of Neutral outcome.
In many cases, the 2nd most desirable outcome is Barbs... this is esp. true when popping huts closer to AI cities, in teams of 2 or 3 units, or when you have Trireme/Caravel/Galleon/Transport support (e.g., pop, get barbs, bribe what you can, hop on ship, let the other barbs stay behind).
Usually, it seems to cost me about 2 units per game to pop huts... and usually because of John Wayne-ing it without waiting for more Dip support in the midgame.
Another midgame consideration is support. At midgame, I don't like Units that get attached in a remote area under Republic/Democracy. On rare occasions, I get a unit and have to disband it because of happiness and it's remote location.
My method is between Conservative and pacman, so I picked Conservative in the poll.
). What I mean is that assuming a full and massive game, I view the terrain not as it is, but for how I will transform it and make the cities fit. If a city is out of place, I like to kill it eraly and move it to where I want it. In GOTM 016, I built one city on a hill and two on swamps and several on Forests, and fixed the terrain later. My own paramount concern is that the city grow to fulfill the trade and production tasks I envision in late game.
). I set the build orders to phalanx and end the turn. The city will produce ten shields towards the phalanx. Then I buy the remaining 10 shields to complete the phalanx. In order that the phalanx is not completed next turn, I change the build order from the phalanx to an archer (requiring 30 shields) and buy the remaining 10 shields to complete the archer. I then change this to a unit requiring 40 shields (crusader, catapult or elephant spring to mind) and buy that before changing the production to a caravan and allowing it to be completed by the shields on the next turn. This way you can have a caravan every 2 turns which will go a long way to building all those wonders you have been lusting after.
This can of course be done with less than ten shields but since they finish the first row of production then this makes the whole process lots cheaper and you get to produce more caravans for your money. 
