Probabilistic Doxastic Temporal (PDT) Logic is a formalism to represent and reason about belief evolutions in multi–agent systems. In this work we develop a theory of abduction for PDT Logic. This gives means to novel reasoning capabilities by determining which epistemic actions can be taken in order to induce an evolution of probabilistic beliefs into a desired goal state. Next to providing a formal account of abduction in PDT Logic, we identify pruning strategies for the solution space, and give a sound and complete algorithm to find minimal solutions to the abduction problem.