Let's start with sound quality. Where are you setting the bar? Do your audio clips need to be simply clear and understandable, or do you want them to sound like a Hollywood voiceover? If it's the former, you can use lower sample rates and sacrifice some of the highest-frequency information for a reduction in file size.
In digital recording, frequency response is half the sample rate. A 32 kHz sample rate will preserve frequencies up to 16 kHz, which is above what the typical adult can even hear (and above what most mics will capture). A 22 kHz sample rate records up to 11 kHz, which still captures most of the clarity and sibilance in the voice. You can keep coming down, as a voice recorded with an 8 kHz sample rate can be clear and understandable enough for some applications.
The next main factor is file size. More compression equals more artifacts and, as you compress further, a loss of clarity. Lower sample rates create less data, hence your bit rate can come down. You're recording in mono, which allows you to further reduce your bit rate.
This is where you engage your ears with some test files. Experiment with sample rates to hear how the sound becomes less clear and "open" as you go down in frequency. Establish your minimum sound quality there, then start experimenting with compression bit rates. If you settle on 16 kHz mono for your recording, for example, start with a bit rate of 64 Kbps (nearly lossless at this sample rate). Drop your bit rate by about 10 Kbps for each recording until the gritty compression artifacts become unacceptable.
Eventually, you'll end up with a recording that represents an acceptable tradeoff between audio quality and file size. Hint: don't make your decision based on one playback system. Listen to your samples over a phone speaker, in earbuds, in a car, through a bluetooth speaker. Try to listen through every type of system your end users may use.
