> older writing > SMthesis

A Conversational Computer Character to Help Children Write Stories

Nicholas Anthony Montfort

Bachelor of Arts
University of Texas at Austin, 1995

Bachelor of Science in Computer Science
University of Texas at Austin, 1995

Submitted to the Program in Media Arts and Sciences
School of Architecture and Planning
in partial fulfillment of the requirements for the degree of
Master of Science in Media Arts and Sciences
at the Massachusetts Institute of Technology
June 1998

By Nick Montfort
Program in Media Arts and Sciences
May 5, 1998

Certified by Justine Cassell
AT&T Career Development Assistant Professor of Media Arts and Sciences
Massachusetts Institute of Technology

Accepted by Stephen A. Benton
Chair, Departmental Committee on Graduate Students
Program in Media Arts and Sciences
Massachusetts Institute of Technology


A conversational computer character may improve students' story writing processes in ways existing software cannot. Writing stories is of great benefit to elementary students, but several factors make story writing difficult. Children comfortable with the interchange of spoken dialog cannot write as fluently without a tangible audience and the usual responses that support conversation. Software that allows a child to converse with a character could be an amusing and engaging way to help stimulate the production of stories. The conversational framework can provide a way for children to begin writing on the computer in a comfortable mode that is familiar from oral discourse and offers the additional support another speaker provides. A computer character with motivation and personality can also provide an example audience during the writing process.

To test whether story assistance software with a conversational computer character can be more educationally effective than software lacking such a character, two Macintosh programs were developed: EddieEdit, employing a conversational character who talks about planning and revision; and StoryStages, which offers identical planning and revision tips but without a conversational character. A two-week study tested both the usability of these programs and whether their educational interventions were effective. The story writing of three groups, one using EddieEdit, one StoryStages, and one a word processor, was compared. During the short time of the study there was little discernible improvement in writing ability and no statistically significant difference in improvement between the three groups, based on what they had written. Thus, examination of stories written at the beginning and end of the study did not provide support for the hypothesis. Answers on a final questionnaire did indicate that EddieEdit users had greater awareness of the writing process than those who used StoryStages, supporting the hypothesis. After the study both pieces of software were improved based on how the software was used by children in the study, and a Web version of Eddie was developed.

Thesis Committee

Justine Cassell, supervisor
AT&T Career Development Assistant Professor of Media Arts and Sciences
Massachusetts Institute of Technology

Glorianna Davenport, reader
Principal Research Associate
Massachusetts Institute of Technology Media Laboratory

Janet Murray, reader
Senior Research Scientist
Massachusetts Institute of Technology Center for Educational Computing Initiatives


I am grateful to many people who helped me as I worked on my thesis. The support of my family, friends, and Janis was particularly important. There were also many people at MIT who made very direct contributions to my thesis work and who I wish to thank. I learned a great deal from the members of my thesis committee, Justine Cassell, Glorianna Davenport, and Janet Murray. They each provided help and guidance even as I was formulating my thesis topic, and continued to help me with critiques and suggestions on my work as I progressed. The comments they provided on drafts of this thesis were incisive and extremely valuable. The classes I took classes from these three also challenged me and improved my thinking about many issues, including some I dealt with in this thesis.

My study could not have been completed without the participation of Karma Paoletti and her class at the Agassiz School, and the help of Pedy Rivera at the school's computer lab. The study also required evaluations of the writing students did. Two of the Media Lab's most capable scholars of story, Marina Umaschi Bers and Kevin Brooks, did these evaluations with great care. This was my first Macintosh programming experience, one I would not have gotten through without the help of Sunil Vemuri. Deepa Iyengar taught me just-in-time statistics that helped me analyze my data. I was helped in other innumerable ways by faculty, support staff, and students in the Gesture and Narrative Language research group, in my work area the Pond, and in the rest of the Media Lab. There are many others I should mention. I would no doubt omit some of those who assisted or supported me no matter how long a list I compiled. Therefore, I will stop here, and use the brevity of this section as my excuse for such omissions.

NEXT (Introduction)