Impact of Conversational Formality on the Quality and Formality of Written Summaries

This study investigated the impact of conversational agent formality on the quality of summaries and formality of written summaries during the training session and on posttest in a trialog-based intelligent tutoring system (ITS). During training, participants learned summarization strategies with the guidance of conversational agents who spoke one of the following three styles of language: (1) a formal language for both the teacher agent and the student agent, (2) an informal language for both agents, and (3) a mixed language with a formal language for the teacher agent and the informal language for the student agent. Results showed that participants wrote better quality summaries during training than pretest and/or posttest in each condition. Results also showed that agent informal language caused participants to write more informal summaries during training than on pretest. Implications are discussed for the potential application of adaptive design of conversational agents in the ITS.