The method according to the invention is characterized in that on the basis of an existing task model, which may be provided by an expert, it is enhanced with the current state of the user in his task, the events allowing a change of state of the user are described, and the interaction to be performed with the user to manage an event is described for this event occurring during a state of the user. Advantageously, before each interaction procedure, the list of constraints necessary for triggering the interaction is added, and after each interaction procedure, the values that this interaction should provide according to the result of the interaction and which should be presented to the user as feedback are added.