Chatbots and academic writing for doctoral students

Abstract This exploratory case study examines how AI technologies, specifically a GPT-4-based synopsis chatbot, can serve as a sparring partner for doctoral students in Norway. Despite favourable conditions, only two-thirds of Norwegian PhD candidates complete their doctorates, partly due to challenges with article-based dissertations that require a comprehensive synopsis (60–90 pages). Ambiguities and unintended double standards exist across disciplines regarding this format, despite national guidelines. To address this, we developed a synopsis chatbot by training GPT-4 on Norwegian doctoral rubrics and literature about article-based dissertations, making it more domain-specific, context-specific, and multilingual. The goal was to determine if and how the chatbot could support PhD candidates in writing their synopses. Preliminary results indicate varying levels of AI acceptance among PhD supervisors, with 60% expressing scepticism about using AI for academic writing. However, the chatbot performed well in providing formative assessment and handling multimodal illustrations, proving to be a valuable sparring partner for doctoral students. This suggests the need to update conventional theories of formative assessment to include AI and chatbots as complementary "digital supervisors" in doctoral education. While the chatbot shows promise in mitigating issues related to unwritten rules and vague genre requirements, this early-phase exploratory study acknowledges several limitations and emphasizes the need to address ethical considerations regarding AI in academia.