Back to search

University of Ioannina

Generative artificial intelligence and the transformation of science education: an activity theory analysis

Abstract

dc:description

The rapid integration of Generative Artificial Intelligence (GenAI) and Large Language Models (LLMs) represents a profound paradigm shift in education, transitioning the pedagogical landscape from an era of information retrieval to one of probabilistic information generation. This transition presents an acute crisis of mediation for early childhood preservice teachers, who are frequently tasked with teaching counter-intuitive scientific threshold concepts, such as light and optics, Earth and space systems, floating and sinking, and energy transformation, often while battling their own low scientific self-efficacy. This thesis investigates how AI-based educational tools impact future teachers' conceptual understanding of natural sciences, their shifting professional identities, and their learning practices. To systematically analyze this disruption and account for the stochastic, agentic capabilities of LLMs, the research employs third-generation Cultural-Historical Activity Theory (CHAT) to map the systemic contradictions introduced by artificial intelligence. By viewing technology as a dynamic mediating artifact, the study operationalizes structural tensions, including the primary contradiction between the tool’s linguistic fluency (exchange value) and its propensity for scientific hallucinations (use value), as well as secondary contradictions clashing with traditional university rules and the teacher's changing division of labor. Methodologically, the study adopts a quantitative, quasi-experimental research design with repeated measures to assess the cognitive and attitudinal impacts of an AI-mediated instructional intervention. The sample comprised 249 undergraduate preservice teachers enrolled in the Department of Early Childhood Education at the University of Ioannina, Greece. Data was collected using standardized pre-tests and post-tests to measure conceptual knowledge in the natural sciences, alongside a multi-dimensional Likert-scale questionnaire capturing attitudes toward technology, artificial intelligence, and chatbots in both general and educational contexts. The empirical findings reveal a profound cognitive and affective shift. Following the intervention, participants demonstrated a highly significant improvement in their natural science knowledge, with mean scores rising from a baseline of 59.98% to 68.73% (p < .001, partial η^2= .939). Simultaneously, overall attitudes toward artificial intelligence and technology became significantly more positive. However, the data also exposed rigid, structural anxieties. Despite widespread adoption and increased enthusiasm for technology in science education, participants' fears regarding pedagogical dependency and ethical concerns (such as privacy and academic integrity) remained persistently high. Crucially, a regression-based mediation analysis with bootstrapping (5,000 resamples) confirmed that the cognitive knowledge gained during the intervention acted as a significant partial mediator between baseline and post-intervention attitudes (indirect effect ab = 0.0614). This indicates that professional motivation and technological affinity did not develop through passive exposure but were driven directly by the subjects successfully utilizing the tool to resolve their own cognitive frictions. Ultimately, this thesis argues that the integration of generative algorithms in science education is not merely a technical upgrade, but an ecological paradigm shift. It demands that teacher education programs formally embed critical artificial intelligence literacy within their curricula, shifting the professional identity of the future educator from a traditional transmitter of facts to an “epistemic orchestrator” and critical auditor of synthetic pedagogy. Furthermore, by contextualizing these localized pedagogical actions within broader quaternary contradictions, such as the massive environmental footprint of LLM computations, the research underscores the necessity of aligning technological integration with critical environmental consciousness and global sustainable development goals.

Degree

thesis:*
Grantor dc:publisher
University of Ioannina
Year dc:date
2026

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Efthimiou, Georgios
  • Ευθυμίου, Γεώργιος

Subjects

dc:subject × 18

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/61976

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Efthimiou, Georgios; Ευθυμίου, Γεώργιος. Generative artificial intelligence and the transformation of science education: an activity theory analysis. University of Ioannina, 2026. http://hdl.handle.net/10442/hedi/61976