If the brain only had a couple of neurons, then it might make sense to pump up the transmitters between them to get a more intense response because you don't need very refined control. However there are 100 billion neurons in the brain, and are connected in very complicated ways.

Since there's evidence that cranking up the seratonin for _every_ neuron or suppressing the dopamine for _every_ neuron isn't healthy, it stands to reason that treating "mental illness" is more complicated than turning a water faucet on or off. Who knows exactly what causes it? It could be related to patterns of connection, i.e. neuron A is better off connected to neuron B at site C, but is connected strongly to neuron D at site E. And then comes the question of whether the neuron is sending weird patterns of action potentials. Or it could be the timing, speed or synchronization compared to other neurons is off. Or it could be poor overall health of each individual neuron. Or subtle brain damage. And the problem might be localized to just some neurons in one part of the brain, and not another, making treatment of the whole brain with drugs a messy approach. You see, there are no obvious answers right now, but there are plenty of alternate explanations besides global chemical imbalance. :)

EDIT: Also, you could think of mental illness as being psychological, and something that the brain can heal on its own, given the right environment.

I think a good analogy is when you leave a paused DVD on a plasma TV for too long so it starts ghosting. Your brain, being a neural net, works suspiciously like a neural net. If you 'practice' being depressed for long enough then it's going to get easier and easier to be depressed, and being not depressed gets harder and harder.

That's not to say what actually causes mental illness or what the solution is, but if you want a metaphor that operates on the hardware layer of abstraction that you can frame other theories and/or best practices in terms of then I think that's probably as good as any.