This lecture provides an initiation into the theory of imprecise probabilities: an extension of probability theory that is able to cope with partially specified probabilities such as probability intervals. I will focus in particular on how this theory can be applied to Markov chains. As we will see, it allows one to deal with partially specified parameters, a lack of time-homogeneity, and even with the failure of the Markov assumption, while still allowing for efficient computations. I will also explain how imprecise probabilities can provide a solution for the scaling problem of Markov chains, and will illustrate this with an application in telecommunication.