Abstract

My research aims to investigate to what extent Artificial Intelligence (AI), like Generative pre-trained transformers (GPTs) can be used to facilitate or enhance dialogic education (DE) on complex topics like climate science, particularly in university settings. Discussions involving climate science are often polarizing because they concern not only the sciences, but a multitude of interconnected socio-economic concerns. Consequently, the likelihood of reaching consensus or compromise is complicated by the diversity of perspectives involved. The ability to have constructive conversations on these topics is therefore essential to this process because environmental action and policy decisions are collaborative problem-solving endeavors among individuals from multiple backgrounds and perspectives. My research involved developing a custom GPT (Agora) based on dialogic principles and testing its performance while collecting data on user interactions. Agora was used as part of a focus group study that included 11 participants who engaged with Agora on topics related to climate science. The study also included discussion periods before and after the Agora interactions. Data from surveys, reflection exercises, and chat logs were analyzed to determine the extent to which Agora facilitated DE, and to better understand user experiences and preferences when conversing with Agora. The results of the study showed that Agora was able to facilitate DE by asking questions that encouraged participants to reflect on their beliefs and assumptions. Participants also expressed increased interest in using GPTs for personal applications in the future, despite their initial reservations about AI. The study also revealed that the quality of interactions with Agora depended largely on the participants' expectations, perceptions, motivations, and approaches. Additionally, reflection exercises indicated that many participants currently struggle with having conversations about climate change, including with those in their personal networks. Some of these struggles included barriers to perspective-taking and perspective-getting, emotional regulation, and conflict resolution. As such, participants expressed a strong interest in learning how to better communicate with others on climate topics, and in receiving personalized guidance on how to navigate interpersonal challenges without alienating others. These insights suggest that there is a need for skills development in climate communication, and that a DE approach in combination with AI could be a promising avenue for addressing this need.