I'm not an expert on the subject, but I'm inclined to follow the advice of guys like James Johnston and Bob Ohlsson.
To paraphrase JJ, if you dither and you can hear an improvement, do it. If you dither and you're not sure if anything happened, do it. If you dither and it compromises the audio, something else in the system is broken.
One of Bob's pet peeves is guys who manufacture plugins that don't automatically dither their output. He advocates that dither is required at every step in processing and ideally we shouldn't have to think about it because it should be automatic. Unfortunately it does chew up significant processing power, so a lot of systems will have it as an option. Modern multicore processors have made this much less of an issue. Some of the best sounding digital systems available will allocate around 1/3rd of their processing to dither. "Triangular" or "TPDF" dither seems to be preferred.
Whether or not your system automatically dithers would hopefully be covered in its documentation. Ideally, it's automatic. Doesn't mean it is.
Especially in a 24 bit target format the source will play a role. Super clean acoustic music will make any low level distortion artifacts more apparent. Going to a 16 bit enviornment will make it much more apparent. Not sure if this is the case with Skrillex. Things like reverb tails, decaying notes and the spatial 3D perception of depth and width is what gets affected. Distortion from not dithering will stack around 3 times faster and is correlated to the signal. Working at a higher resolution to the target format like you're doing is a huge help on its own.
What we can percieve from it is a hugely subjective debate but if it doesn't hurt anything, there's no reason why we shouldn't.