First, I think we should probably address the concept of volume. The level of a signal has nothing to do with volume. You don't have volume until it is actually put through speakers and you are a hearing it a certain distance from those speakers. Instead you have a "level".
The level is a measure of signal strength and it isn't subjective, but rather based on what the minimum and maximum signal level are for the system you are working with. All the way off is always negative infinity and it works it's way up to 0 which is the ideal maximum signal level for the device to maintain the best quality. However, nobody talks at exactly the same level all the time, so the signal is going to vary some. On most good equipment, it can peak a little above 0db(signal, not volume), but ideally you want to keep the gain set to be at 0 and a little below it most of the time.
Eventually, if the signal gets too powerful, it will exceed what the device can handle and you end up with the waveform getting cut off at the top and bottom. This results in a partially square wave rather than a rounded one and is called clipping.
When setting the gain for recording, your primary goal is to get the majority of the signal as close to 0 as possible (the best possible quality), but you also have to make sure that the signal does not clip, which will render that part of the recording unusable. You also don't want it too much lower than 0db because then the amount of noise that is recorded will increase in relation to the amount of actual signal.
After recording, when you get in to Audacity, it is simply scaling the signal when you adjust the levels. It makes the signal behave as if it is more or less powerful than it originally was so that you can mix it with other inputs. It is all just math being done on the signal and doesn't actually alter the waveform's basic shape, just makes it have a lower amplitude.
The difference between the highest and lowest volumes on a recording is called the dynamic range. Generally, recordings will have a larger dynamic range than will typically make an ideal listening at home. For this reason, a concept called dynamic compression is frequently used. Dynamic compression is a process by which the highest levels can be automatically gained down and the lower levels can be brought up. This results in less of a difference between the highest and lowest signal levels and decreases the dynamic range. This makes a more level volume when someone plays back the audio later.
Ideally the signal will be compressed and then the level raised to be close to 0 to minimize the noise that will be generated in the playback device and maximize the signal level. Then when the device plays it back, it can accurately process the signal to produce output to the speakers at a level that is consistent with other recordings and will have a similar volume to other things that the user was listening to.
That's a pretty quick overview of a lot of what goes in to a recording, but I hope it helps give a deeper understanding of the answer to your question.