Engagement vs. Reflection: A Study of User Interfaces for an Intelligent Story Authoring Assistant Brian O’Neill School of Interactive Computing Georgia Institute of Technology Atlanta, Georgia, USA boneill@cc.gatech.edu Mark O. Riedl School of Interactive Computing Georgia Institute of Technology Atlanta, Georgia, USA riedl@cc.gatech.edu ABSTRACT As a desire to create media for consumption has increased, tools have been developed to aid in the creative process for several forms of media. However, creating content that will be considered valuable by the community is challenging, and there remains a need for tools that support novices. We ap- proach this problem with ReQUEST, an intelligent story au- thoring assistant that acts as a surrogate audience, provid- ing feedback as the user writes. In this paper, we describe a study comparing two interfaces, differing in the visual repre- sentation of the story and the presentation of feedback to the user. Users of one interface appeared to enter into a reflec- tion phase of the creation process more often. However, this interface may have encouraged too much reflection, possibly leading users to revise their stories despite believing that it would not improve their creation. Author Keywords Intelligent Authoring Support, Story Authoring, User Inter- faces, Mixed-Initiative Systems ACM Classification Keywords H.5.2 [Information interfaces and presentation (e.g., HCI)] User Interfaces: Evaluation/methodology. I.2.1 [Artificial Intelligence] Applications and Expert Systems. J.5 [Arts and Humanities] Fine Arts. INTRODUCTION A growing trend in the world is the desire to create media, rather than simply consume. Today, tools exist for creat- ing nearly every media imaginable. However, the mere abil- ity to create does not guarantee that a particular community will value the new work, or respect its quality. Because of this gap between simply creating and creating with value, tools are needed to support non-expert authors in their ef- forts. Tools that support and scaffold the creation of content would be more valuable than a tool that simply manipulates data. Unpublished manuscript, September 19, 2008. Story telling is an integral aspect of many forms of media, from literature to movies, and their non-expert equivalents such as fan-fiction and machinima. Consequently, in this paper we address support for story authoring. In particular, we focus on the plot of the story, for the time-being, leaving the user to address other elements of storytelling, such as character dialogue, for later in the process. This approach is influenced by McKee’s [7] description of the screenwriting process, which begins with the plot in prose, before adding elements inherent to screenwriting, such as camera angle, character placement, and dialogue. Further, this approach leads us to our long-term goal of a mixed-initiative system [3], capable of providing feedback and solving problems in conjunction with the user on a num- ber of activities related to content creation. Artificial intel- ligence support of creative content authoring differs from mixed-initiative support of problem solving in subtle ways. Lubart [6] enumerates four ways in which computer inter- faces can support creativity: • Computer as nanny: The computer provides organiza- tional and classification services and performs routine op- erations on behalf of the user. • Computer as pen-pal: The computer facilitates brain- storming with functionality that captures and transmits to collaborators the user’s thoughts. • Computer as coach: The computer is knowledgeable about the process and can offer suggestions and stimulate creativity. • Computer as colleague: The computer forms half of a human-computer team by contributing to the solution. The computer as coach metaphor is used extensively in intel- ligent tutoring systems [16]. However, it is not our intention to teach a user how to author plots. A system-as-coach acts to facilitate improved task performance by a user, but does not attempt to solve or contribute to the problem on which the user is working. In contrast, the computer as colleague metaphor introduces automation into the creation process. Typical mixed-initiative support systems implement an ex- pert system that is capable of solving some, or all, of the problem the human user is working on. The user and the system take turns, filling in details of the solution. 1