Abstract:

The rapid adoption of generative AI has introduced a new type of systemic pressure on education, extending beyond technological integration to fundamentally challenge pedagogical practices, assessment validity, teacher roles, and institutional governance. While AI chatbots are increasingly used by students, their deployment often remains unstructured, raising concerns about superficial learning, academic integrity, and loss of pedagogical control.

This paper conceptualizes generative AI as a source of structural pressure on education systems and proposes a framework that identifies key pressure dimensions:
a) pedagogical,
b) assessment-related,
c) cognitive,
d) institutional, and
e) ethical/legal.

The framework is grounded in empirical insights collected through a series of structured focus groups with students and higher education professionals (e.g., instructors, instructional designers). The focus groups were conducted as part of a collaborative design process. Participants were asked to reflect on their experiences with generative AI tools, perceived risks, and expectations for responsible integration in education.

The analysis revealed behavior and perception patterns across stakeholder groups. Students reported a tendency to use AI for efficiency gains, often at the expense of deeper understanding, indicating cognitive and pedagogical pressures. Educators emphasized concerns regarding assessment validity, loss of insight into student learning processes, and misalignment between AI-generated outputs and course objectives. At the institutional level, participants highlighted uncertainty regarding regulatory compliance, particularly in relation to data protection and emerging AI regulations.

Building on these findings, the paper presents the design rationale of a pedagogically aligned AI tutor (ART: Augmented Reality Tutor) as a response to these pressures. ART has been piloted across four countries and nine universities, with over one thousand students, providing real-world usage data to inform its development. In contrast to general-purpose chatbots, ART is structured around teacher-controlled content, transparent reasoning processes, and interaction designs that promote active learning (e.g., guided questioning and feedback). Preliminary observations from these deployments indicate that such systems can redirect student interaction patterns from answer-seeking toward learning-oriented engagement, while simultaneously providing teachers with visibility into student learning processes through interaction data.

The paper contributes:
(1) a stakeholder-informed framework for understanding the systemic impact of generative AI in education, grounded in empirical qualitative data, and
(2) a set of design principles for AI tutors that mitigate identified pressures while preserving educational quality and teacher agency.

These findings suggest that the effective integration of generative AI in education requires a shift from tool-centric adoption toward pedagogically aligned system design, where AI supports the core structures of teaching and learning, and does not aim at replacing them.

Keywords:

AI tutors, AI, Technology, Chatbots.