Modelling proactive behaviour of conversational interfaces

Conversational Interfaces are software systems interacting with their users by means of a natural language based conversation. Their knowledge base consists of a collection of rules, whose triggering depends on text patterns recognized within the user input. The output is sometimes presented to the user in combination with a graphical visualisation, e.g. a cartoon, of the related emotion. Even if they are an interesting and promising idea, recent statistics detected poor usage and a relatively short lifetime. In this approach, a conversational system is designed which is not simply reactive to a user's request, but capable of engaging in a collaboration with its users, by taking over the initiative in problematic and unclear situations. By comparing conversational interfaces to finite state machines a formal definition of proactivity is reached and a standard implementation methodology is suggested. Finally, a case study discussing an example implementation of a virtual risk management advisor is provided.