The students experience on generative AI to develop reflective and argumentative skills in the learning process

Reading, understanding and being able to reframe what one reads is one of the basic skills for being an active participant in the world. In recent decades, there has been a growing concern about undergraduates’ poor skills in reframing and articulating their own opinions on a topic. This impacts critical thinking and the ability to express an original point of view. Especially now that the increasingly post-digital reality has complicated the way information and skills are acquired. In this regard, argumentative maps (AM) can help easier visualization of the structure of a piece of information. They diagram the structure of logical relationships between different utterances and present the chain of reasoning quickly and effectively. The methodology of AM, however, was designed for static text types, but needs to be reconceptualized to deal with increasingly dynamic and data-driven information typical of the post-digital age. Currently, everyone's attention is on the growing use of generative AI and the attempt to integrate it to support learning practices and continuous improvement of individual student performance. The potentials are many, but so are the risks. Based on this theoretical framework, a two-group quasi-experiment was conducted on the use of AMs to support the understanding of static and multimodal information and the development of critical thinking. In this experiment, ChatGPT, the most widely used generative AI technology, was employed as an intelligent agent with which students could compare their argumentative and communicative interactions. Our research sample had 10 female students, attending the course of "Educational Evaluation" at the University of Padua (L-19). The course began in February 2023 and ended in June 2023. The class was invited to collaborate with the famous chatbot to compare their response to the activity: specifically, students were asked to read a text and identify the structural components of an argument (problem, thesis, supporting arguments, objections, evidence, and conclusion) and then to collaborate with ChatGPT, proposing the same activity. From there, students were to compare their answers with those of the AI and evaluate their usefulness for their own reworking of the answer. Data on text comprehension and correct AM construction were collected by semistructured tools and check-lists; data on the perceived role of ChatGPT in personal argumentative processing were collected from posts in the online forum. The students had no prior experience using the chatbot. Initially, they were enthusiastic about the interaction. However, during the other phases of the course, they discovered the AI-powered tool was not specific and required further interactions. An initial thematic analysis conducted with Nvivo software showed widespread difficulty not only in communicating with the AI, but also in following the reasoning of its responses. A general skepticism was read among the students, especially regarding the possibility of gaining benefits and support from “a person who doesn’t even exist”. In the future, it would be interesting to investigate more nuances of students’ interaction, maybe integrating other uses of ChatGPT. We will present the case with an interactive presentation (Wooclap) and ask the audience to suggest possible lines of action to put into practice, employing ChatGPT or other AIs to support argumentative skills in Higher Education.