Introduction: Accompanying schools in the implementation of innovative didactic models is an important factor in gaining confidence in the use of both innovative teaching and organisational practices (Capperucci et al., 2021) and in rethinking their role in the use of technological tools such as artificial intelligence in the classroom (Gillani et al, 2023). This paper aims to contribute to the debate on didactics with AI and possible training paths for its application in education. (Panciroli&Rivoltella, 2023). The paper presents the results of the pre-test used for the large-scale investigation the AI4T project impact evaluation model: a counterfactual research design to explore the AI4T professional learning experience (Guskey & Anderman 2013), considering the constructs in the Technology Acceptance Model (Davis & al., 1989; Kemp at al., 2019). The paper includes findings on key issues that emerged from the pilot phase of the project, which involved 3 headmasters and 16 teachers, which guided the revision of the final assessment instruments

Methodology The main objective of the small-scale pilot phase was to test the entire research protocol and evaluation tools for quality assurance in large-scale applications. Each partner country was to recruit between two and eight schools. In the case of Italy, was identified 8 schools, within which there were one/two volunteer mathematics and English teachers to answer the questionnaire, in addition to the school headmaster, who were divided between the experimental and control group. The choice of schools was based as far as possible on the criteria also adopted later in the large-scale management phase: Region, Type of schools (academic/professional), An indicator of the social composition of the school, The number of volunteer teachers, Personal information of the teachers (gender and teaching experience). The questionnaires to teachers were administered before and after the training they received on artificial intelligence tools through a MOOC prepared and managed by CNR:ITD in Palermo. The tool consisted of closed and open questions. Each teacher and headmaster was then assigned a code and then the questionnaire was administered via a limes survey. Following the completion of the questionnaires, a special section was provided for the evaluation of the instruments.

Results

School Leaders stated that the questionnaires clearly addressed the main aspects concerning IA. Of particular interest to them were the questions concerning the possibility of integrating these tools into the school and the ethical implications. The opportunity to respond to the questionnaire made it possible to trigger a useful reflection to highlight the readiness of school leaders to stimulate teachers to train on these aspects

As far as teachers are concerned, the level of satisfaction with the questionnaire is also high. Most stated that it was of particular interest to reflect on emotions and their own sense of self-efficacy with respect to the use of artificial intelligence in the classroom.

The reflections of the headteachers and teachers on the instruments were useful for reviewing the two questionnaires, going into greater depth in ethical aspects for the former and in areas related to attitudes for the latter.

Augmentative and Alternative Communication (AAC) techniques are designed to support individuals with communication disabilities, enabling them to effectively convey their thoughts, needs, and emotions [1]. These methods are especially used to consolidate the autonomy levels of the individuals, for example, teaching them how to behave in every context. However, to be effective, AAC requires a personalized approach based on the cognitive level of the individuals. For example, individuals can understand symbols differently, demanding a huge effort from educators, caregivers, and speech-language professionals to create the material.

To create a personalized learning environment in AAC, we propose to use foundation models [2]. Foundation models are large pre-trained machine learning models that can generate complex texts or images. When fine-tuned for AAC, they offer educators efficient tools to craft personalized learning materials, making them indispensable for supporting autonomy. Indeed, they can create material that is tailored to the specific cognitive level, use AI-based methods to obtain suggestions for personalized content and create additional material that follows the progress of the individuals, making their learning path as smooth as possible.

Our approach is called AMBRA (Pervasive and Personalized Augmentative and Alternative Communication based on Federated Learning and Generative AI) [3]. AMBRA leverages a cloud-based system to create a shared space for the learning material that each educator can further customize. AI-based methods can also help create suggestions based on previous experience. This material can be accessed from modern devices (e.g., smartphones, tablets, laptops, etc.) in every context.

While this approach has great potential, it also opens several challenges to be discussed with the community. Our approach is built on top of free solutions, lowering the barriers and the costs to access AAC methods. Educators may exchange material more easily, creating a virtuous cycle that can boost the adoption of AAC methods. Also, low deployment costs allow shops, public places, and community places to adopt such solutions and allow for a seamless transition between school and society, creating a more inclusive environment around the individuals. For example, QR codes allow users to download specific material, which is later personalized based on specific user’s information and symbols. It will become the perfect application of the autonomy concepts that have been acquired in class. An approach based on fine-tuned foundation models demands a combination of skills that are hard to find in modern educators. AMBRA advocates the creation of interdisciplinary environments where, for example, pedagogical and linguistic skills are combined with ICT skills. Finally, the use of personalized material and its flexible creation allows educators to approach sensitive topics (e.g., sexuality) in a way that is appropriate for everyone. Also, it can lower the cultural and language barriers when used with immigrants, facilitating their integration into a new community.

In conclusion, we believe that using artificial intelligence and generative AI opens the possibility of creating new learning paths that can empower all students to obtain complete autonomy and make society more inclusive.

A topic that is becoming increasingly important in the technological and social landscape concerns the relationship between critical thinking and the formulation of relevant questions in the use of generative artificial intelligence. If we understand critical thinking as a mental process that allows us to analyze facts, situations, or statements objectively and evaluate them in a logical manner to arrive at a well-considered judgement, it then appears to us as a methodical approach that implies the ability to recognize prejudices, unverified hypotheses as well as fallacious arguments (think of fake news).

Formulating well-structured questions is a skill that becomes possible when a person uses critical thinking to investigate a given topic more deeply (Graesser & Person, 1994; Brookfield, 2012). Well-worded questions are clear, direct, and constructed with a specific purpose in mind. They can direct discussion or inquiry toward deeper understanding and can delve into information that might otherwise remain unexplored.

All this highlights the importance of asking good questions, especially if we want to use AI (Wegerif, 2002). Content generated by AI is a direct reflection of the queries provided. If what we ask is vague or ambiguous, the result provided will be too.

A relevant question in the AI context is a question that leads to accurate or creative answers (Gunning, Aha & Zhu, 2019). This is why a beta version of an AI model made by a research group at INDIRE has been tested in the classroom. Knowing the structures that underlie the functioning of a generative artificial intelligence, understanding how it works and "thinks" are moments that allow you to interact with this technology. All these aspects will be presented to facilitate the use of the INDIRE AI model.

The focus of this contribution is based on two pivotal issues: fostering self-efficacy as well as enhancing teachers’ familiarity with AI tools (Passey, 2019). During the speech teachers will be given the opportunity to discuss and to interact with the INDIRE’s model. Testing (and interacting with) it, will be a formative moment to assist teachers to confidently embrace and utilize AI in the classroom (Bandura, 1994; Ventura & Shute, 2013).

In essence, this speech aims to empower teachers to effectively navigate the digital age, thereby enhancing the overall learning experience. It underscores the critical need for teacher preparedness in the evolving AI-driven educational landscape.

Since the launch of ChatGPT in November 2022, the debate and interest in using artificial intelligence (AI) in education (AIEd) technologies have increased. When talking about the integration of generative artificial intelligence (AI) in educational settings, the debate is even more lively and polarized, with at least two parties, the concerned and the enthusiasts (Berendt et al., 2020).

As the educational landscape continues to evolve, it is essential to understand teachers' sentiments regarding the use of generative AI in the classroom, as they are crucial to successfully promoting innovation with AIEd in schools (Timms, 2016). Researchers' investigations went to different aspects, such as ethical implication, positive/negative impact on learnings, and possible applications (in the Italian context, Panciroli & Rivoltella, 2023 and Ranieri, Cuomo, & Biagini, 2024). Still, to the best of our knowledge, more investigations need to be into teachers' perceptions of these technologies.

This preliminary study explores the sentiments of in-training teachers regarding the use of generative artificial intelligence (AI) in educational contexts and their perceptions of utility, efficacy and user-friendliness. Moreover, it investigates if Generative AI, such as ChatGPT, is perceived as a technology that can potentially transform or arm education. We aim to provide insights into these professionals' sentiment on opportunities, challenges, and ethical considerations associated with integrating generative AI into teaching practices.

The survey was conducted in January-February 2024, using CAWI methodology (Computer Assisted Web Interview). Students self-compiled a questionnaire generated with Google Forms and answered closed-ended questions. The participants were 91 in-training pre-school and primary school teachers at Università del Molise - a course for educational support activities for pupils with disabilities.

The first results highlighted the participants' awareness of AI and, generally speaking, a positive attitude toward its use in education and their teaching practice. Moreover, they are optimistic about adopting AI, particularly for inclusion.

Data analysis revealed a gap between the declared knowledge and interest in AI and its actual and conscious use. The prevailing application foreseen is the drafting of documentation for teachers and research activities for students. Furthermore, teachers perceive AI as beneficial for inclusion, but its main potential is related to information retrieval. In conclusion, the ambivalence that has emerged is also confirmed by the gap between the awareness of the need for AI-related skills and training courses to acquire them on the one hand and the perspective, which appears to be the majority of not using tools such as Chat GPT at school.

The findings of this research are expected to provide valuable insights for educational stakeholders, policymakers, and teacher training programs, guiding them in harnessing the potential of generative AI while addressing the associated concerns.

Artificial Intelligence is, nowadays, at the center of many research studies in the educational field that share its potential for improving the education and training sector (European Commission, 2022). Artificial Intelligence Literacy in Education, or AILE (Wilton et al., 2022) is a set of skills that educators should possess in order to «understand, create, use, apply and evaluate AI applications in an educational context» (Wilton et al., 2022, p. 181).

As in the rest of Europe, Italy has also witnessed the technological adjustment of public administration systems. In the field of school evaluation (Freddano & Pastore, 2018), digitization and “platformization” have made it possible to update evaluation activities keeping up with new tools and new digital environments in order to facilitate evaluation procedures (Freddano & Buoninconti, 2023). In addition to the opportunities, however, several critical issues have emerged from the use of these new systems, first and foremost the difficulty for users to interact with digital tools, to read and interpret data and to make strategic use of artificial intelligence. Particularly on this last issue, Long & Magerko (2020) point out how in platforms users often fail to recognize that they are interacting with an artificial intelligence, limiting their opportunities to collaborate and act effectively as AI consumers.

The new operational challenges opened up by these transformations focus, therefore, on the issue of the competences needed by those who face evaluation processes: we speak, in fact, of the so-called “triad” (Kreinsen & Schulz, 2023) Data Literacy, Digital Literacy and AI Literacy. These three competences are complementary and interconnected, just think of the need to use digital tools to interact with an AI or the use of data as a resource itself for AI, a consideration that necessarily leads to reinterpret the processes according to an integrated reading of competences and thus according to a new set of training needs for the users involved.

The research proposed here starts from the elaboration of an integrated theoretical model of process and competences (Freddano & Mariani, 2023) that stems from the overlapping of an inquiry model (Mandinach & Gummer, 2016), a New Generation Digital Learning Environment model (Brown et al., 2015) and a digital competence system (Vuorikari et al., 2022), from which a selection of specific training areas for the evaluation process is derived. In order to map the training needs of those who carry out self-evaluation, a survey was planned addressed to the members of Internal Evaluation Teams of 159 Italian schools, selected according to stratified random sampling (by geographic macro-area and, in cycle II schools, also by school address). The purpose of the survey is to collect data on the sense of self-efficacy with respect to the proposed competences and the training needs of the users. The first results of the survey will be presented at the conference, allowing a discussion on the opportunities of a competence-based approach in the self-evaluation of schools in an integrated view of both evaluation and learning.

Generative Artificial Intelligence (GAI) presents a significant opportunity to revolutionize the governance of social and educational services. Digital technologies, widely acknowledged as advantageous for the public sector (Noveck, 2015), facilitate the creation of organizational networks and foster collaborative governance (Fischer, 2006). GAI, in particular, has the potential to permeate all sectors of personal services in a short period. For instance, in education, it is already influencing the adoption of automated tools for assessment, customization of programs and educational materials, implementation of automated tutoring systems, and analysis of predictive indicators for the risk of school dropout (Guan, Mou, Jiang, 2020). Furthermore, exploration is underway regarding the use of GAI to promote inclusivity in schools, through the adoption of multisensory technologies that enhance learning and interactions among children with autism spectrum disorders, students with Special Educational Needs (SEN), and Specific Learning Disorders (SLD) (Grimaldi, 2022). However, it is essential to recognize potential negative repercussions and unintended effects for both institutional and non-institutional actors utilizing or potentially utilizing GAI systems. Specifically, transparency in data construction processes is crucial if GAI becomes a key strategy for governing social and educational welfare.

In response to these challenges, this contribution presents the experience of Govern-AI. Govern-AI is a research and innovation program developed by the University of Naples Federico II, aimed at experimenting with GAI in the governance of social and education services. It also initiates an action dedicated to adult education, involving a Provincial Center for Adult Education (CPIA) in Campania and its faculty. This initiative aims to personalize support provided to educational staff through the implementation of a generative artificial intelligence-based chatbot. The chatbot is designed to adapt to the specific needs of learning units, thereby facilitating access to highly specific and personalized educational resources, supporting teachers in the teaching process. The goal is to enhance the effectiveness and inclusivity of adult education while promoting active involvement of teachers and CPIA officials in the decision-making process and continuous system creation. This action reflects Govern-AI's commitment to promoting educational innovation through a participatory and interdisciplinary approach, placing teacher and educational staff empowerment at the center.