Basically, people working on graphics-related algorithms needed to build a library of standard test images, so that when people published their work in an academic journal, they could easily demonstrate what that algorithm does, in a manner that is fairly obvious to anyone who is familiar with the image.

So someone, when they needed to pick an image that represents a person, scanned this photograph. And it could be argued that at the time, it was probably an interesting test image for a lot of reasons: person vs background, different textures, areas with soft and sharp focus, etc etc. If you developed, say, an image compression algorithm, those things are going to be headache in all photo portraits.

It's probably not the best image by modern standards (being a low resolution scan of a photograph off of a printed magazine - not a photo print scan, not a direct film scan, and not comparable to digital photography). Also, it's gotten overused to the point of absurdity. (Oh your hot new face detection algorithm works on this image? Well whoop-de-do.)