In building effective interfaces or computer human interaction devices, a main limitation of traditional presentation design is the inability to meet individual user expectation at run-time. On-line design of individualised presentations that surpass the limits of the ”one size fits all” approaches can be made possible by user modeling techniques. Building models of users can be done throughMachine Learning, but this requires techniques that are specific to the task.One particular issue is that the user models cannot remain static, in the sense that during the use of the intended interaction more knowledge can be gathered which should in turn be used to improve the user models. The knowledge from which the construction of the models can be made is many-fold: web logs, speech, images...