A Conversational Learning

The conversational audio guide is an exploratory system design project about conversational learning between humans and AI. In the project, we designed a conversational agent that impersonates the virtual artist to offer a pedagogic dialogue. The idea initially started from a question "Can the AI promote critical thinking and reflection to deepen the understanding of the subject as a human mentor?" The museum was perceived as an optimal ground for testing the hypothesis, as it naturally invites open interpretation [1].

Specifically, we challenged impersonating the agent as an artist to (1) increase the perceived social characteristics of the agent and (2) bring a theater-like experience to the museum environment. Based on the constructivist framework [2], we dimmed both points to be critical for the learning to take place within the dialogue. Through iterative design procedures, we increased the narrative details of the response to implement the impersonation.

System Evaluation

Through the demonstration of the conversational audio guide, our team was able to identify varying user expectations and evaluation of the system. Generally, the idea is perceived as interesting and the participants evaluated the interaction to be beneficial for appreciating the art. Meanwhile, the richness of the dialogue and the subsequent gain largely differed between the participants. The participants with more interest in the artwork and the field of art showed a greater tendency to connect with the agent.

Conversational Agent as a Learning Partner

The conversational agent is being integrated into the classroom at a fast pace [3]. Its previous role as an assistant handing out information is transforming to a more collaborative role, giving directions and offering evaluation. An increasing presence of knowledgeable agents that are replicable at scale suggests a crucial chance for expanding educational accessibility. Precisely, an agent that distributes knowledge through dialogue suggests lightweight learning for groups previously suffering from short attention spans both mentally and physically (e.g., low-vision, attention deficit). Based on this idea, the study suggests an exploratory design of the conversational audio guide, an agent impersonating an artist to provide constructivist learning. Through study, we were able to identify learning effects and design directions for agent-mediated art-pedagogic dialogue.

Reference

• [1] Roberts, Jessica, and Leilah Lyons. "The value of learning talk: applying a novel dialogue scoring method to inform interaction design in an open-ended, embodied museum exhibit." International Journal of Computer-Supported Collaborative Learning 12 (2017): 343-376.
• [2] Abtahi, Yasmine, Mellony Graven, and Stephen Lerman. "Conceptualising the more knowledgeable other within a multi-directional ZPD." Educational Studies in Mathematics 96 (2017): 275-287.
• [3] Kasneci, Enkelejda, et al. "ChatGPT for good? On opportunities and challenges of large language models for education." Learning and individual differences 103 (2023): 102274.