Noob question about sample rates and dithering...

+1 to LSD. You technically risk a higher chance of artifacts being introduced to a single when down sampling from one sample rate to another. Dither is bit-rate specific - but you should ALWAYS track closest to the project medium specs as possible (CD 44.1, DVD 48khz, etc)
 
I record at 44.1/24-bit, I let iTunes convert the wavs into 44.1/16-bit for burning to a CD, but anything I put on the net is either 44.1/24 if it's short enough to keep it a wav or 192kbps mp3 if it's a long file. I agree with Lasse that unless you are going to be doing audio for DVD there is no reason to work in 48. Moonlapse, I'm with you man, I've heard from different circles so many opinions about the whole 44.1vs.48vs.88.2vs.96vs.etc. regarding which is honestly better. Some say the higher the better, some say it won't matter if it's going to be 44.1 in the end anyway - there's really not a solid authority about what is going on with different rates unfortunately. I just know that until I hear something for sure that there is a definite advantage over 44.1/24, I won't be worrying about it.

~006
 
I record at 44.1/24 and perform bauncing (when effects will be applied destructively)/mixdown at 32 bits.
So no dithering involved. After that do some mastering and use dithering while render to 44.1/16.
With 24 bit mixdown better to turn dithering off, until mastering... I think.
Maybe I wrong, but I think that dithering performed on any bit depth lower than 32...
 
When you bounce without dither in 24bit to 16bit you are truncating your audio. The last 8 bits are just thrown out and your are losing lots of detail. I'd let the mastering house dither your final project but if your sending a mix to a label/artist your should dither down to 16bit. A good example of the detail lost is like the reverb tail from the last transient. It might crap out and you'll be left with a very bitter taste in your mouth (ears).