The philosopher Stephane Vial (2019) defines "digital ontophony" as the technologically mediated and produced aspects of our lives, as an effect of digitalization. Through the processes of platformization (Van Dijck, Poell, De Waal, 2018) and mediatization (Couldry, Hepp 2017) our daily behaviors are in fact firmly intertwined with media and technologies. Recommender algorithms and platforms affordances contribute to forms of discrimination, symbolic and physical violence. Furthermore, the recent diffusion of linguistic models (LLM) for the production of images, such as Midjourney or DALL-E, forces us to reflect on how artificial intelligence can be a tool to influence the collective imaginary. It directs our choices and our tastes in the direction of the majority, also shaping our “aesthetic self” (Manovich 2018). Midjourney, as a sociomaterial artefact, collaborates in the confirmation of gender, racial and ableist stereotypes because it draws on an imaginary that we ourselves have provided it. As Esposito (2021) writes, the problem relating to bias is more related to the machine's inability to see, and therefore create, beyond its own preconceived. From this point of view, we take the perspective of digital methods considering algorithms "epistemological machines". Meaning we can use errors and hallucinations to chart biases in the training set, hence in the imaginary produced by LLM tools..

We present the preliminary results of empirical research carried out within secondary school classes. In the broader framework of the research that this working group is conducting on LLM technologies, in fact, the construction of methodological tools to bring into school classrooms is useful for understanding the relationship between AI Aesthetic and the aesthetic and imaginary baggage of the students.

During the laboratory activity, students are asked to create the set of characters, objects and places intended for writing a fiction. We made use of Propp's cards generated with Midjourney. We asked the students to outline the characteristics of the characters starting from the image represented, following Propp's scheme. Hallucinations and stereotypes emerged through photo-elicitation (see Harper 2002). To these responses, the participants contributed with great creative investment, since the final objective was the creation of a work of fiction. On the one hand the cards generated by AI are used as a strategy to collect qualitative data, on the other they are teaching material for media literacy.

As a first result we can observe that the initial enthusiasm with which the images were welcomed gave way anxiety caused by the "excessive beauty" of the representations, considered detached, austere, devoid of empathy. These elements then played a role in defining the characters' storylines. Starting from these first instances it is possible to outline the themes on which to build the next steps of the research.

Large Language Models - LLM as Chatgpt constitute, today, a challenge involving not only the so-called hard sciences, but also social and human sciences (Adeshola & Adepoju, 2023; Caligiore, 2023; Cristianini, 2023).

The proposed research aims to analyze the impact of artificial intelligence (AI) on university education with a focus on its ethical and pedagogical applications (Baidoo-Anu & Ansah, 2023). The study intends to contribute to a better understanding of the ongoing debate on integrating AI technologies in university contexts. The main focus of the paper is to examine the interactions between professors and students when using AI, going beyond simple technology adoption and usage. When examining the interaction between AI, lecturers, and students, it becomes clear that the quality of the prompt given to the AI is crucial.

This research intends to investigate how both subjects (professors and students) interact with Chatgpt, as well as how they derive useful information for educational goals from this technology.

The analysis provides a qualitative perspective on how AI-based tools like Chatgpt 3.5 can improve learning experiences and help to develop academic skills if used appropriately. Furthermore, the study addresses the ethical dimensions accompanying the incorporation of AI in university education, and deals with issues such as algorithmic transparency, accountability in decision-making processes, and privacy protection of students and lecturers (Cristianini & Scantamburlo, 2019; Guleria et al., 2023).

The proposed investigation, with a purely exploratory direction, contributes to the ongoing debate on the responsible implementation of AI in university education, offering an articulate perspective that goes beyond the surface, inviting debate on the transformative potential of AI and acknowledging its ethical imperatives.

Background

The digital transformation of education, through artificial intelligence (AI), offers improvements in teaching and learning. The role of AI is crucial in designing interventions that meet the needs of individual learners, highlighting the importance of personalised education amidst its complexities and potential risks (EC, 2023). The emergence of generative AI technologies, as seen in platforms such as MagicSchool (https://www.magicschool.ai/) and EduAide (https://www.eduaide.ai), enables educators to customise lesson plans, teaching/learning materials and assessments. Such personalisation capabilities of AI aim not only to improve educational outcomes, but also to ensure student well-being; a promising shift towards more responsive and adaptive educational practices (Molinaar, 2021).

Research

This study is embedded in a project that aims to refine teaching methods towards personalised education that fosters individual student growth. 37 Institutes of all school grades participated in this research activity, with at least 2 teachers involved in each school, for a total of 135 teachers (Mori et al, 2024). The work examines whether the shift from content creators to evaluators improves the personalisation of educational content, including the importance of educators being aware of the ethical implications of using AI in the classroom (EC, 2022). This awareness is crucial for making informed, responsible decisions about the use of these technologies to ensure that they contribute positively to personalised learning environments. The research and training course is now underway and is scheduled to end in May 2024. The purpose of this contribution is twofold: to illustrate the course taken with the teachers and to report an initial analysis of the teachers' final evaluations.

Methods

A final survey and interviews were used to gather feedback from teachers on their experiences, challenges and perceived benefits of using these platforms. An exploratory mixed methods approach (Trinchero & Robasto, 2019, p. 14) was used, incorporating both quantitative and qualitative data collection techniques. The research design used a mixed methods approach (Creswell and Plano Clark, 2011) of an explanatory sequential type, characterised by an initial quantitative data collection that will be deepened using a qualitative approach. The qualitative analysis will follow a grounded theory approach (Kuckartz, 2014).

Results

From the analysis of the data, it will be possible to present initial findings and illustrate the perceived strengths and weaknesses of the use of the proposed tools to enhance teaching and learning activities to support the development of students' skills. The results will contribute to a better understanding of the potential of integrating these AI driven technologies in the design of teaching for the different school orders to support the unique needs and potentials of each learner.

Conclusion

The role of human oversight in generative AI-driven curriculum creation is essential to ensure educational quality, ethical standards and relevance. Generative AI has the potential to help create educational content more efficiently. However, human oversight is necessary to verify the accuracy of the generated educational material, maintain ethical standards, and ensure that it aligns with the intended learning objectives. (Hutson & Lang, 2023)