Today’s Question: In what situations would you convert from 16-bit to 8-bit?
Tim’s Quick Answer: Only when necessary. I recommend keeping your primary image file in 16-bit per channel mode, and only converting to 8-bit per channel when creating an image for sharing, such as when saving a JPEG derivative image to share online.
More Detail: When you process a raw capture, such as with Camera Raw in Photoshop, I recommend converting to a 16-bit per channel image rather than 8-bit per channel to maximize potential detail and fidelity, and to minimize the risk of any posterization (the loss of smooth gradations of tone and color). I only recommend converting to 8-bit per channel mode when it is necessary, such as when saving a derivative image in a format such as JPEG that doesn’t support (or require) a higher bit depth.
If you’re completely finished applying adjustments to an image that had been saved in the 16-bit per channel mode, and you want to reduce the bit depth to reduce overall file size, the risks are not too great as long as you only do so for a color (not black and white) image and you make sure to flatten the image first (assuming the use of layers in Photoshop).
For black and white images, I would never convert the master image file to 8-bit per channel because doing so carries considerable risk of posterization in the image, leading to potential banding and other problematic visual artifacts.
But overall, I recommend keeping the master image in the 16-bit per channel mode, and only using 8-bit per channel mode for derivative images created for sharing. Even in those cases, it isn’t generally necessary to use the explicit step of converting to 8-bit per channel mode, such as by using the Image > Mode > 8 Bits/Channel command in Photoshop. For example in Photoshop you can simply use the “Save a Copy” command to save as a JPEG in 8-bit per channel mode because 16-bit per channel is not supported in that case.

