NaNofuzz: A Usable Tool for Automatic Test Generation [pdf]
cmumatt.github.io
cmumatt.github.io
On the other hand, I'm very happy to see that testing gets some love from research, and that at least they tried to design a study. Not a very convincing study (with 28 programmers participating), but I can imagine it to be very hard to get a lot of highly paid professionals to participate.
I am of the idea that software in general is written once but read a lot. But, for tests it is a good thing to have really extensive coverage. More than maybe a human is willing to do.