Conference Program

Generative AI is rapidly entering language classrooms, including Ancient Greek, where its apparent fluency can tempt learners (and institutions) to outsource philological judgment. This paper argues that AI should be treated not as a neutral aid but as sociotechnical agency whose computational logics can reorganize educational life around efficiency, predictability, and the production of “one-best” answers. Drawing on Illich’s analysis of “two watersheds” and radical monopoly, we hypothesize that educational AI may cross a threshold beyond which tools cease to support autonomy and instead generate dependency and epistemic homogenization. In parallel, Freire’s pedagogy of autonomy frames learning as ethical self-formation through repeated, responsible decisions, precisely the kind of decisions demanded by Ancient Greek parsing and translation.

We introduce Convivial Threshold Design for Ancient Greek (CTD‑GR), a design-oriented intervention translating non-directive pedagogy into implementable didactics. CTD‑GR operationalizes three methodological aims – co-construction of knowledge, critical examination of algorithmic and cognitive bias, and dialectical argumentation as democratic competence – through a minimal activity architecture:

(1) Threshold Mapping: teacher and students identify tasks where AI assistance becomes dependency (e.g., full translations, ready-made commentaries, automated morphology) and set “convivial limits” that preserve core human competences (parsing, reasoning about ambiguity, evidence-based justification).

(2) Textual Co‑translation Workshops: small groups produce parallel translations of a shared Greek passage, documenting alternative construals (case/tense/aspect, participial scope, particles, word order). AI may be queried only to generate contestable hypotheses (“possible readings”), which are then argued for or rejected through reference to grammars, lexica, and the Greek text itself.

(3) Bias Audits on AI Glosses and Explanations: learners test AI-generated glosses, morphological parses, and cultural explanations under prompt variation, explicitly searching for (a) systemic bias (cultural frames and canon formation), (b) computational/statistical bias (patterned errors, confabulations), and (c) human-cognitive bias (over-trust in plausible outputs). Findings are recorded in an “error-and-evidence log” that becomes a shared class resource.

CTD‑GR culminates in a collectively negotiated “convivial use agreement” for Greek study –transparent, revisable, and aligned with assessment that rewards traceable reasoning over polished output. Expected democratic outcomes include autonomy in tool-use decisions, epistemic plurality in interpretation, reflective judgment (phronesis), and dialogical agency sustained by dialectical argumentation.

Emotional intelligence constitutes a key educational and civic competence in contemporary learning environments marked by complexity, uncertainty, and intensified relational interdependence. At the same time, Artificial Intelligence is increasingly embedded in educational systems through learning analytics, adaptive platforms, automated assessment, and decision-support tools, profoundly reshaping pedagogical practices, institutional governance, and conceptions of learning itself. These developments raise critical questions regarding autonomy, agency, and the democratic purposes of education.

Drawing on non-directive and critical pedagogical traditions—particularly Ivan Illich’s notion of convivial tools and Paulo Freire’s pedagogy of autonomy—this contribution examines the intersection of Artificial Intelligence, Emotional Intelligence, and education as a crucial site for rethinking democratic learning in the age of algorithmic mediation. Rather than approaching AI as a neutral educational technology, the paper conceptualizes it as an emerging form of agency whose computational logics may privilege prediction, optimization, and standardization, potentially fostering dependency, epistemic homogenization, and the erosion of ethical self-formation when uncritically adopted in educational contexts.

Through a critical review of interdisciplinary literature in AI and Education, philosophy of education, and complexity studies, complemented by illustrative examples of AI-based educational applications, the study explores how Emotional Intelligence can function as a countervailing pedagogical resource. Specifically, it argues that emotional awareness, dialogical responsibility, and relational reflexivity are essential for enabling learners and educators to critically engage with AI-mediated learning environments rather than passively adapting to them.

Furthermore, recent theoretical work calls for reconceptualizing AI in education beyond tools—toward hybrid human-AI systems that extend human cognition while foregrounding agency, motivation, and emotion. Particular attention is devoted to Illich’s concept of thresholds of mutation, highlighting the point at which educational technologies cease to support learning autonomy and begin to reorganize educational experience around control, surveillance, and behavioral normalization. From a Freirean perspective, the paper emphasizes that Emotional Intelligence in education should not be reduced to a measurable or optimizable variable within AI-driven systems, but understood as a dialogical, ethical, and political practice cultivated through participatory and critical pedagogical relationships.

A concrete and hitherto under-theorized instantiation of this surveillance threshold is examined through an original empirical investigation of eye-tracking technology integrated within Moodle-based assessments delivered via Safe Exam Browser (SEB). Drawing on the convivial and Freirean frameworks developed in this study, the paper critically interrogates how the deployment of eye-tracking in lockdown examination contexts risks crossing Illich’s threshold of mutation—transforming assessment from a site of authentic demonstration of learning into a normalized apparatus of biometric surveillance that may induce learner anxiety, self-censorship, and performative compliance rather than genuine cognitive engagement and critical sense-making.

The contribution ultimately proposes a convivial framework for AI and Education, in which Artificial Intelligence is reconfigured from an apparatus of pedagogical automation into a reflective partner supporting critical inquiry, collective sense-making, and democratic participation. By aligning human-centered AI with pedagogies of conviviality, the study offers theoretical and practical insights for designing educational environments that resist technocratic reductionism and promote autonomy, plural reasoning, and democratic agency in complex learning systems.

In the high-stakes environment of Emergency Medicine (EM), the rapid expansion of clinical knowledge threatens to create a "radical monopoly" where learners may become dependent on non-transparent digital tools. The employment of Artificial Intelligence (AI) in this field raises central ethical questions regarding patient safety, clinical accountability, and the reliability of generated information. International guidelines, such as FUTURE-AI and the WHO guidance, emphasize that AI adoption must remain anchored in the bioethical principles of beneficence, non-maleficence, autonomy, and justice. In this context, explicability plays a crucial role, as AI systems must provide outputs that are understandable, traceable, and open to critical scrutiny by human professionals.

Our study strictly adheres to the ethical principles of AI application in medicine, ensuring that technological innovation remains at the service of human responsibility and clinical oversight. This initial version of the app represents a primary validation of purely educational content, aimed at relieving educators of formatting burdens while maintaining scientific accuracy. The strategic goal of our work was to establish a robust ethical and qualitative foundation to subsequently evolve the tool into a clinical decision support system (CDSS) for real-time use in Emergency Departments.

We conducted a two-arm, randomized controlled trial involving 123 evaluations from Italian EM residents and specialists. The study compared a traditional human-driven workflow with an innovative "human-in-the-loop" AI-assisted approach using GPT-4 to generate flashcards validated by external subject matter experts. The adopted methodology directly targeted the risk of AI hallucinations by safeguarding the central role of the human educator as the final authority for clinical validation and responsibility.

The results demonstrate that human-revised, AI-assisted flashcards achieved significantly higher mean scores in relevance (3.97 vs. 3.61), correctness (3.83 vs. 3.62), and clarity (3.86 vs. 3.71). Conversely, human-only content scored higher in informativeness (3.74 vs. 3.72), highlighting that human expertise is essential for providing nuances based on real-world experience and critical alerts.

Our study highlights that the integration of AI into medical education should be understood as a collaborative and ethically grounded process rather than a substitutive one. Framing AI as an initial support tool, systematically followed by expert clinical review, aligns its use with the fundamental principles of biomedical ethics: it promotes beneficence by enhancing decision quality, safeguards non-maleficence through human validation of outputs, preserves professional and learner autonomy by preventing overreliance on opaque systems, and supports justice by encouraging transparent and accountable use of technological resources.

Within this model, AI does not function as a surrogate decision-maker but as a form of ethically supervised cognitive augmentation, strengthening reflective clinical reasoning while ensuring that responsibility, interpretive authority, and patient-centered judgment remain firmly in human hands.

This contribution introduces a pedagogical and methodological framework for the integration of LLMs into secondary mathematics education through logic-constrained prompt learning [1], a paradigm of structured interaction in which the prompt is conceived as a logical object and the reasoning generated by AI is explicitly constrained by formal proof procedures [2,3]. Rather than employing AI systems as solution generators, the proposed approach reconfigures them as inference-guided reasoning environments and, simultaneously, as agents of reasoning verification, whose outputs must conform to transparent deductive constraints. The framework is illustrated in the context of high-school-level Euclidean geometry problems, where the primary objective is not merely to obtain the correct result, but to master the procedural structure of mathematical inference. The central idea is to reinterpret the solution of elementary geometric problems as a satisfiability-driven logical task. The proof task is then operationalized by constructing a classical analytic tableau for the set composed of the premises together with the negation of the thesis. Closure of the tableau corresponds to the impossibility of a countermodel and therefore to the logical validity of the geometric conclusion. This transformation enables students to perceive geometric proofs not as narrative arguments, but as structured inferential processes governed by explicit rules. Within this setting, the LLM is queried under stringent procedural constraints: it must expand formulas exclusively via the standard tableau rules, label each inferential step, avoid introducing undeclared theorems, and explicitly indicate branch closures and contradictions. Such constraints reduce the model’s tendency to provide heuristic or intuitive explanations and instead impose the production of traceable proof objects. The resulting interaction introduces a form of dialogical agency, in which the AI participates in the inferential process as a regulated interlocutor, and shifts epistemic authority from the generative system to the formal framework: correctness does not depend on the plausibility of the explanation, but on the verifiability of the derivational structure. Students are thus encouraged to analyze, critique, and reconstruct the tableau, developing metacognitive awareness of proof structure and logical dependencies among statements. From a theoretical perspective, the framework establishes a constructive equivalence between elementary geometric demonstration and formal logical consequence, offering a unified view of mathematical reasoning across different domains. From a pedagogical perspective, it supports three complementary learning outcomes: explicit understanding of deductive chains in geometry, development of formalization skills through controlled propositional encoding of mathematical statements, and critical engagement with AI outputs as objects of validation rather than authoritative answers. The approach is compatible with collaborative classroom practices such as inquiry-based learning and thinking-classroom models, and requires only minimal formal knowledge beyond an introduction to classical propositional logic. The proposed methodology contributes to research on human-centered AI in education by offering a replicable model for integrating LLMs into proof-oriented disciplines without sacrificing rigor or transparency. Through the direct incorporation of formal proof procedures into the prompt structure, logic-constrained prompt learning enables AI to operate simultaneously as an environment and an agent of reasoning verification, thereby fostering reflective judgment, rigorous reasoning, and participatory learning contexts.

The global AI debate reveals, at the geopolitical level, what Ivan Illich (1973) would recognize as the emergence of a radical monopoly: the control of artificial intelligence is explicitly and directly tied to the struggle for world leadership, among nations with sharply divergent values, visions, and practices (Aresu 2024). When a technology reaches this threshold, it no longer merely supports human activity but begins to reorganize social and educational experience around efficiency, predictability, and control. Against this backdrop, the relationships between Artificial Intelligence (AI), democratic education, and citizenship formation become not merely relevant but urgent.

Educational technologies are never neutral. The concept of affordance reminds us that tools push toward specific uses while remaining open to different pedagogical orientations: the same instrument can serve transmission-based or constructivist approaches. In Freirean terms, this distinction maps onto the difference between education that reproduces dependency and education that cultivates critical agency (Freire 2014).

This tension becomes particularly acute with AI. Drawing on Ranieri, Biagini and Cuomo's distinction (2023) between educating with and educating about AI - and in dialogue with UNESCO (Miao & Shiohira 2024, Cukurova & Miao 2024) frameworks on AI competencies - we argue that convivial pedagogy requires both dimensions simultaneously: using AI as a dialogical partner while developing the critical literacy to examine its computational logics, biases, and epistemic effects.

A critical pedagogy of AI must begin from where learners actually are. AI is not new to users' experience: long before conscious engagement with tools like ChatGPT or Gemini, it has shaped daily life through recommendation systems operating largely below the threshold of awareness. Levels of AI literacy are highly differentiated and mapping them is not merely a diagnostic exercise - it is itself a pedagogical act. Surfacing implicit perceptions reveals what we might call personal philosophies of AI: tacit assumptions about agency, trust, and knowledge that risk fostering exactly the epistemic dependency and homogenization that convivial pedagogy seeks to counter. Empirical investigation of perceptions and practices thus provides both a necessary starting point for educational design and a means of motivating learners toward more conscious, critical engagement (Bruni & Murgia 2024).

Recent data reveal both the scale and the stakes of this challenge. According to EU Kids Online (Mascheroni, Rosichini & Cino 2026), 89% of Italian children and adolescents used generative AI in the past month - rising to 98% among 15-16-year-olds - primarily as a learning support tool, with ChatGPT described as «a teacher always available». Yet adolescents themselves identify dependency as the dominant risk: AI, they fear, will make them «lazier and less capable». This self-awareness is a pedagogical resource. Drawing on a survey conducted with high school students in Milan, the study maps AI perceptions and practices as a first step toward designing convivial educational environments that harness these tools without surrendering critical agency to them. The findings suggest that a Freirean pedagogy for the AI era must begin precisely here - transforming adolescents' tacit concerns into conscious dialectical inquiry, and their habitual use into reflective, autonomous action.