Learners’ Reactions to Chatbot Communication Breakdowns: Insights Into Fostering Learning

Abstract

Chatbots can promote productive discussion and learning. However, chatbots sometimes fail to interpret users’ intent, leading to communication breakdowns and user disengagement. To alleviate these issues, researchers have experimented with designs that may increase users’ willingness to fix breakdowns. This study explores the affordances of varying chatbot personalities (as a peer versus an expert) to prompt user repairs. Data sources drew from the chat logs of 18 small groups discussions in science. Results suggest that groups more often rephrased and provided explanations during communication failures with the peer chatbot than the expert version. These patterns, particularly providing explanations, had learning implications. Findings illustrate the affordances of designs that enact certain communication strategies—ones that may enrich learning during frictions in computer-supported interactions.