Conversation as a Basis for Interactivity

not interactivity, while dynamic, in that they are moving, they are not interactive: the user simply sits passively and watches them. He does not interact with them. Interactivity occurs when the user gets to do something, and when the users action affects the behaviour of the software. In its most basic form, then, interactivity implies user control and dynamic experience. Most attempts to define interactivity involve taxonomies and continuums-interactivity going from 'none' to 'maximum'. Simple pragmatic taxonomies of interactivity focus on physical and mechanical things-what the user can do, how the program would behave in response. For example, many taxonomies of interaction identify five basic levels of interactivity: presentation; simple branching; complex branching; complex branching with feedback; and user generated content. Kristof and Satran (1995) presented a physical/mechanical taxonomy that focused more explicitly on the notion of user control-the more control, the more interactivity: It is not the case, however, that more interactivity is always better, indeed Murtaugh (1996) asks why anyone would want to exert control over a well-constructed story. Every project has an optimal degree of interactivity but taxonomies give little guidance to the degree or form of interactivity for a given task. ABSTRACT Interactivity is a widely used term. While well described in terms of mechanics, it is lacking in an underlying model. We propose conversation as a basis for considering interactivity. Key implications of this are the provision of information for inferential processing, non-sequential access to information and an approach based upon an unfolding dialogue between the user and the application. Natural language processing and artificial intelligence may not be available in the near future but this should not stop us examining further the components that make conversation work for us and attempting to apply them in our interactive applications.