Many realistic usage situations of software products may be influenced by both the individual characteristics of the user, the past usage history, as well as the current dynamic environment. Can you incorporate such information into Markov chains? If yes, how? If no, why not?