What type of research does duncan watts discuss in his interview, and why isn't that type of research always feasible in sociology?