Can Conversational Agents Promote Our Comprehension? : Effects of Knowledge Level and Representational Form

摘要

POC caster, a conversational information-providing agent generates understandable conversational representation from text-based documents, and it introduces opinions of community members by using a conversational representation in Public Opinion Channel (POC), an interactive broadcasting system on the Internet for supporting community knowledge creation. We evaluated its transformation rules in an experiment with a factorial design. In the experiment, we investigated the relationship between knowledge level and representational form. Forty-two participants (21 with a high knowledge level and 21 with a low knowledge level) were selected for a knowledge task. They listened to two types of sentences (the sentences about a well known topic or about an unknown one) and judged which sentences were easy-to-understand as quickly as possible. The results indicated that the conversational representation affected comprehension, and was especially aided those without knowledge about the topic. Moreover, the speed of selection was faster in the conversational representation than in the text representation. We explored possible explanations and implications of these results in relation to human cognition and text comprehension.