Conference Agenda

1. Introduction

Generative artificial intelligence (GenAI) has rapidly become embedded in everyday academic practices in higher education, with a large proportion of students reporting regular use of AI-based tools to support their studies (e.g. Johnston et al. 2024). Despite increasing research attention, important gaps remain in understanding how GenAI is used across specific learning tasks and pedagogical contexts (e.g. Belkina et al. 2025).

Recent studies suggest that pedagogical approaches to GenAI should move beyond efficiency-oriented uses and instead support higher-order learning processes such as critical thinking, creativity, and learner agency (e.g. Belkina et al. 2025). At the same time, students’ use and engagement with GenAI varies in depth and quality, raising questions about how pedagogical design can either enable or constrain students’ agency in engaging with AI (e.g. McPhee & Jerowsky, 2025). While prior research has largely focused on individual AI use and personalized learning support (e.g. Chan & Hu 2023), and only a limited number of studies have examined how generative AI practices develop within collaborative or peer learning contexts (e.g. Perifanou & Economides, 2025 Ruiz-Rojas et al., 2024).

This study examines students’ self-reported uses and experiences of GenAI across two successive student cohorts within the same course design, focusing on both individual and peer learning tasks in a university of applied sciences context.

2. Methods

The study draws on survey data collected from two student cohorts (2024 and 2025) participating in the same project-based higher education course in which GenAI use was intentionally integrated into learning activities. The course included individual and peer learning assignments, with structured opportunities to use GenAI for planning, collaboration, reflection, and presentation tasks.

Data include students’ self-reported GenAI use across different learning tasks, experiences of AI-supported learning, and selected learner-related psychological factors such as self-efficacy, motivation, and perceived study ability. In addition, team identifiers enabled analysis of similarities and differences in reported AI use within project teams. Quantitative analyses focus on cohort comparisons, associations between psychological factors and AI use, and patterns of convergence within teams.

3. Results and Discussion

At the time of submission, data analyses are partially ongoing, and the conference presentation will report the final results. Preliminary results indicate that students’ engagement with GenAI varies across learning tasks. GenAI was reported as most helpful in creating the final project presentation and least helpful in preparing for reflective discussions, indicating differences in perceived usefulness across task types.

Analyses focusing on prior experience show that students with little or no previous exposure to GenAI reported the greatest learning-related gains: approximately one third reported learning something new, and over half indicated intentions to use AI in the future. These findings suggest that structured, collaborative AI-supported tasks may support more equitable development of AI readiness.

With respect to learner-related psychological factors, analyses are ongoing. Preliminary results indicate a weak trend between study ability and reported GenAI use. Team-level analyses are expected to provide insights into AI-use practices in collaborative learning contexts.

4. Educational Significance of the Research

This study contributes to research on GenAI in higher education by highlighting the role of pedagogical design, psychological factors, and peer-learning contexts in shaping students’ AI-supported learning practices. By examining successive cohorts within the same learning design, the study provides insights into recent developments in AI use as AI is increasingly becoming a normalized element of education. The results offer practical implications for designing equitable, learner-centered AI-supported learning activities, particularly in peer learning settings that prepare students for professional practice.

Introduction

Educational systems today face complex and systemic challenges that require the ability to reformulate problems, generate innovative solutions, and design changes tailored to specific contexts. In this scenario, creativity emerges as a systemic professional skill, particularly relevant for professionals called upon to support innovation processes in educational institutions. (Ulferts, H. (ed.), 2021)

In recent years, Design Thinking (DT) has emerged as a participatory approach in educational contexts, that supports collaborative ideation processes and helps school staff reframe problems, generate innovative ideas, and test solutions (Mangione et al., 2025). Besides this, the widespread diffusion of Artificial Intelligence (AI) has highlighted its potential in scaffolding school teams' creativity: AI systems based on Large Language Models (LLMs), and in particular multi-agent systems (LLM-MAS), when designed with pedagogical intentionality, can act as pedagogical mediators. In this perspective, LLM-MAS systems support teachers' creativity by helping them reformulate problems, explore new ideas, and reduce the perceived risk in ideation processes (Zampolini et al., 2025).

Building on these premises, this study explores the use of an LLM-MAS based on DT, with the aim of supporting school futures.

Methods

To answer the research questions (RQ1) How does participation in LLM-MAS-supported DT workshops affect school teams creative competence? (RQ2) What types of solutions emerge from LLM-MAS-supported collaborative ideation processes? the study adopts a mixed-methods pilot design with an embedded and exploratory logic (Creswell & Plano Clark, 2018), aimed at examining both the impact of LLM-MAS-supported DT workshops on participants’ creative competence and the characteristics of solutions emerging from collaborative ideation processes. The intervention was conducted in five small schools located in rural areas of central Italy and involved teachers, school leaders, and other system figures, with at least three participants per school. Participants took part in collaborative DT workshops (Björgvinsson et al., 2010; Mangione et al., 2025), focusing on context-specific challenges related to school innovation and professional practice.

DT activities were supported by a LLM-MAS intentionally designed as a pedagogical mediator for ideation. The system integrates structured dialogic roles each associated with specific cognitive and creative functions (Zampolini et al., 2025). Data collection combined quantitative and qualitative sources. Quantitative data were gathered through pre- and post-intervention measures of creative competence and perceived self-efficacy, grounded in the conceptualization of creativity as a developable professional competence rather than a fixed trait (Runco, 2004; 2014). Qualitative data included interaction logs between participants and artificial agents during the workshops, artefacts produced during ideation sessions—such as reformulated problem statements and proposed solutions—and collect

Results and discussion

Preliminary observations suggest that the LLM-MAS system promotes a more balanced participation and supports divergent and convergent thinking processes. Participants show a greater propensity to reformulate problems and explore alternative perspectives, particularly in interactions with agents characterized by critical or generative roles. Indeed, teachers and system figures with greater experience in collaborative ideation processes tend to recognize the pedagogical value of the MAS system more clearly. The MAS system is interpreted as a scaffolding for ideation and not as a simple tool for automatically generating answers. Its multi-agent architecture, based on differentiated dialogic roles, appears consistent with the epistemology of Design Thinking, which conceives ideation as a collective, iterative, and situated process. Participants report increased confidence in proposing innovative solutions and greater awareness of the systemic challenges of their school context. Overall, these results suggest that MAS systems can function as pedagogical scaffolds capable of transforming AI from a passive tool to a co-design partner, supporting the professional creativity and agency of school actors.

Educational relevance of research

This study contributes to research on the futures of schooling by showing how artificial intelligence can support schools intentionally working on shared visions of change. Integrated within Design Thinking processes, Multi-Agent Systems enable schools to explore alternative futures, articulate values and priorities, and translate emerging visions into coherent directions for action. In this perspective, AI functions as a pedagogical mediator that scaffolds collective vision-making rather than prescribing solutions. By supporting collaborative ideation and reflective dialogue, MAS-based environments help schools engage with uncertainty in a generative way, strengthening professional agency and enabling deliberate, future-oriented transformation.

The rapid proliferation of digital technologies and artificial intelligence (AI) is fundamentally transforming educational systems, learning processes, and developmental trajectories. In both formal and informal learning environments, learners and educators are increasingly embedded in digitally saturated, algorithmically structured spaces that reshape how knowledge is accessed, evaluated, and constructed. This contribution introduces the concept of digital chaos as an interdisciplinary analytical framework for examining the impact of pervasive digitalization and AI on education, learning, and teaching, with particular relevance for envisioning the education of the future.

Digital chaos is conceptualized as a systemic condition arising from the continuous and often unstructured presence of digital media, AI-driven tools, and platform-based learning environments. In educational contexts, this condition manifests through information overload, constant connectivity, fragmented attention, and the growing influence of opaque algorithmic systems on learning pathways and educational decision-making (Williamson, 2017; Selwyn, 2019). These dynamics can produce experiences of cognitive strain, loss of control, and disorientation among learners, while simultaneously enabling new forms of personalization, participation, and self-directed learning (OECD, 2021).

Rather than framing digital technologies and AI solely as external disruptions or risk factors, this paper adopts an interdisciplinary perspective that highlights the ambivalence of digital chaos. From this viewpoint, digital chaos reflects ongoing transformations in educational structures, pedagogical authority, and epistemological assumptions about knowledge and learning. Learners are increasingly required to navigate complex, fragmented knowledge environments, develop strategies for managing attention and uncertainty, and engage critically with algorithmically curated content (Kirschner & De Bruyckere, 2017; van Dijck, Poell, & de Waal, 2018).

The empirical basis of the study consists of qualitative materials drawn from publicly available podcasts and online posts in which educators, learners, and educational professionals reflect on their everyday experiences with digital technologies and AI in educational settings. These narrative sources provide insight into how digital overload, algorithmic influence, and technological acceleration are interpreted and negotiated in practice. Using Braun and Clarke’s (2006) thematic analysis, the study identifies dominant and counter-narratives that frame digital technologies either as obstacles to meaningful learning or as resources that support reflection, structure, and pedagogical innovation.

A central focus of the analysis is the role of AI in educational problem interpretations. AI emerges in the narratives both as a stressor—intensifying performance monitoring, standardization, comparison, and datafication—and as a potential supportive tool that enables adaptive feedback, scaffolding, and metacognitive awareness (Luckin et al., 2016; Holmes et al., 2019). AI-based learning systems, recommendation algorithms, and generative tools are described as reshaping not only how learners engage with content, but also how educators design instruction and how educational success is defined and assessed (Zawacki-Richter et al., 2019).

Integrating perspectives from educational science, learning psychology, media studies, and critical AI research, the paper argues that shaping the education of the future requires moving beyond binary debates that position technology as either inherently beneficial or harmful. Instead, educational research and practice must focus on how learners can develop competencies to navigate digital chaos productively, including critical digital literacy, self-regulation, reflexivity, and ethical judgment (Buckingham, 2015; UNESCO, 2023).

The study proposes a reflexive educational framework that reinterprets digitally related learning difficulties not merely as deficits or failures, but as meaningful responses to complex and rapidly changing learning environments. Such a framework emphasizes pedagogical approaches that support the conscious integration of digital technologies, transparent engagement with algorithmic systems, and the creation of learning spaces that foster coherence, agency, and critical understanding (Biesta, 2017; Selwyn, 2022). Rather than attempting to reduce or eliminate digital complexity, the approach encourages educators and learners to actively work with, reflect on, and reshape digital conditions.

In conclusion, this contribution seeks to stimulate interdisciplinary and societal discourse on what successful education of the future could entail in an age of AI and pervasive digitalization. By conceptualizing digital chaos as both a challenge and a potential resource, the paper highlights the importance of educational approaches that enable learners to master current and future challenges through adaptive, reflective, and ethically grounded engagement with digital technologies.