Being colorblind and a photographer makes setting white balance in changing light extremely difficult. If we have one of the color deficiencies present in 4 or 5% of the population, we can't simply slide the temperature and tint sliders until people "look right." Faces that look right to us often look
tinted green or pink to those with normal color vision.
Instead, we hunt for something white or neutral in the photo and use the white balance tool when editing. If were lucky, many photos were taken under the same light and we can blindly apply those settings to all of them. If photos were not taken under the same light, there is little we can do other than guess or hunt for something we suspect is neutrally colored, short of leaving a gray card in every photo or setting WB manually each time. Even in daylight, a subject moving in and out of shade makes a very significant difference in proper WB.
While a series of photos may occur under many different light sources, it often has significant overlap in subjects. If the WB can be set correctly for a single photo, software could then allow another point to be placed on something else, such as a yellow corsage, since its proper color is now known. As this subject moves in and out of the shade, or indoors under incandescent or fluorescent light, WB can be done against this yellow corsage and the color for all future photos can be corrected without finding something neutrally colored in each photo.
What is really being described above is a system allowing something that can recognize objects but not color to properly WB photographs. In the above case, that system is a colorblind photographer. It could just as easily be a computer algorithm. This algorithm can do pattern matching against a set of photos using this same technique. Spotting a bright yellow flower against a red dress and normalizing the colors across many photos given a reference photo with known proper white balance would not be difficult for a computer algorithm.
A much simpler alternative to this method would be software overlaying a mask showing all neutrally-colored objects in a photo when a WB spot is selected. Choose the known white edge of an Uno card in a photo and the software would highlight all other items that are now known to be neutrally-colored, such as a gray suit worn by a subject and a black book spine in the photo. Now that other items are known to be neutral, the colorblind photographer can WB against those items in subsequent photos.