You need to dither if your dynamic range is very big, and you are using the dynamic range which is near the noise floor in your 24 bit project.
If you convert to 16 bit in this case, you will get your softest parts of the music having audible distortion because the your smaller numbers are being truncated heavily, and dithering can help you solve this.
However, dithering introduces random noise, and Noise Shaping is another technique which is commonly used when dithering to shift the noise floor up to a less audible spectrum.
If your music has really soft parts, (live recording of orchestral?) Its recommended to dither and noise shape when bouncing to 16 bit.
However if its normal music using compression to push the loudness limit, dithering would have little audible effect.