The Role of Source Monitoring in Resolving Cognitive Disequilibrium on Texts with Controversial Topics
Data is provided by the student.
This dissertation investigated the effects of text position (i.e. the position that the author of the text who argues in support of a controversial issue), agent roles (tutor versus student), and agent speech acts on source monitoring and how these factors impact memory for information in a learning environment with conversational agents. Previous research on learning from text with conversational agents has supported the claim that agent disagreements stimulate a state of cognitive disequilibrium and confusion in the participants, which leads to deep learning when resolved. Source monitoring is one plausible strategy to resolve conflict engendered from cognitive disequilibrium. This dissertation combines two lines of research (i.e., cognitive disequilibrium and source monitoring) to investigate the interacting effects of agent disagreements with each other, with the texts on controversial topics, and with the participants personal positions on topics. Participants read texts that had either a pro or con position regarding a controversial topic while listening to a concurrent disagreeing or agreeing discussion between a teacher and student agent regarding the text. The results showed that participants prior belief was the biggest predictor of memory for text overall. In addition, when participants reported a strong stance regarding supporting or not supporting a topic, the agent disagreement did not affect their discriminatory memory for atypical information. On the other hand, when participants reported a neutral stance regarding supporting or not supporting a topic, the agent disagreements assisted the participants in discriminating the surface structure of the text (i.e., the wording and syntax) as well as their summarization. Source memory was not shown to be a mediating factor for learning during cognitive disequilibrium, whereas the nature of the topics had a large mediating factor for all memory measures. The results were discussed in the context of memory models that make predictions on how incongruence with prior beliefs and messages from texts and agents affects recognition and recall differently.