Hands-on practical work is widely treated as the epistemic core of school and university science. Artificial intelligence (AI) can now produce laboratory reports, answer practical questions, grade student work, and replace physical experiments with adaptive simulation, so the products of practical work can be obtained without the embodied activity that gives them value. This systematic review asks whether AI-era technologies threaten the manipulative, tacit, and procedural competencies that practical work is meant to build. We integrate cognitive offloading, the tacit dimension of learning, and the deskilling dynamic into a conceptual model, derive five propositions, and use them as an a priori coding framework. Following PRISMA 2020, we screened 3,897 unique records and synthesised 20 empirical studies (2019 to 2026) through best-fit framework synthesis, appraised with the Mixed Methods Appraisal Tool. Coding confirmed the two theorised displacement families, cognitive offloading and embodied substitution, resolved into five mechanisms led by simulation substitution (10 studies) and answer provision (9 studies). The central proposition, cumulative skill erosion, remains untested: no study measured practical competence experimentally, so the deskilling thesis rests on anecdotal and perception data. We contribute an operationalised model that separates task-level offloading from cumulative deskilling, an assessment of the evidence graded by the strength of the study designs that produced it, and a measurement agenda.
This study examines how the integration of innovative educational technologies within a Future Classroom Lab (FCL) environment, as an example of a Smart Learning Environment (SLE), is associated with changes in pre-service teachers’ science learning outcomes, perceived teaching competence, science teaching self-efficacy beliefs, and affective dimension of learning, including positive and negative emotions and attitudes toward science learning. To this end, an instructional intervention based on the principles of Smart Education (SE) was implemented in the general science course in the second year of the bachelor’s degree in Primary Education. The intervention consisted of a sequence of activities implemented across different zones of a FCL and involving educational technologies such as augmented and virtual reality, 3D modeling, micro:bit, and AI- and Scratch-based coding activities. To examine pre-post changes associated with participation in the instructional intervention, linear mixed-effects models, controlling for gender, age, place of residence (urban or rural), educational background, and pre-course grade, revealed significant improvements in science learning outcomes, perceived teaching competence, and science teaching self-efficacy beliefs. Changes in the affective domain were limited, and none of the affective outcomes showed a statistically significant pre–post change in the primary mixed-effects analyses. Overall, these findings reinforce the idea that the educational value of SLE lies not only in the technologies they incorporate, but also in the pedagogical and spatial configurations of the FCL through which those technologies are meaningfully and intentionally embedded.
Automated scoring tools have shown promise for assessing student scientific reasoning at scale, but their application to evaluating students’ transfer of learning to new contexts remains understudied. This study examined how well a BERT-based multi-task classification model (Bidirectional Encoder Representations from Transformers, BERT) reproduced expert human scoring of undergraduate responses to near- and far-transfer problems in biochemistry, with a focus on noncovalent interactions. Using responses from 792 students, we trained and evaluated the model across six rubric categories for near-transfer problems and five categories for far-transfer problems. Quantitative results indicated strong model performance in near-transfer contexts (overall accuracy = 0.91) across all categories except for the high-quality claim category. Performance declined in far-transfer contexts (overall accuracy = 0.87), particularly in categories involving evidence use and claims for new interaction predictions, but remained robust overall. Qualitative thematic analysis showed that the model aligned with human raters when student responses used discipline-specific vocabulary and explicit, unambiguous reasoning. Misalignments occurred when responses involved vague claims, alternative but valid reasoning, superficial scientific language, non-normative ideas, or missed contextual information. The model also showed limited sensitivity to graded levels of scientific accuracy due to the use of a three-class output structure. Together, these findings demonstrate the potential of BERT-based models to support automated scoring of near-transfer problems while highlighting key challenges in analyzing student thinking in far-transfer contexts.
Conceptual reconstruction remains challenging because learners often retain alternative conceptions even after encountering scientific explanations. Grounded in the broader conceptual change framework, this study explored conceptual reconstruction among pre-service elementary school teachers through refutational and AI-supported reflective learning on free-fall motion. Twenty-five pre-service elementary school teachers participated in learning activities designed to challenge and re-evaluate existing conceptions through refutational text and reflective interactions with AI. The findings showed a significant improvement (p = .007) in pre-service teachers’ conceptual understanding following the intervention. Most students demonstrated conceptual reconstruction toward scientific conceptions while engaging in activities involving cognitive conflict, reflection, and conceptual evaluation. However, conceptual reconstruction was not uniform, as some students maintained their initial conceptions or exhibited conceptual regression despite engaging in reflective activities. The Epistemic Network Analysis (ENA) results revealed different reflective structures between the refutation phase and the AI-supported reflective phase, indicating that students engaged differently with conceptual challenges and scientific explanations across stages of learning. Students used AI-supported reflection to compare, evaluate, and reconsider their existing conceptions rather than simply adopt scientific explanations. These findings suggest that conceptual reconstruction is a negotiated process shaped by cognitive conflict, reflection, and conceptual evaluation. The study contributes to understanding how reflective engagement with refutational experiences and AI-supported reflection can support conceptual reconstruction in science teacher education.
The use of individual student laptops and tablets in K-12 classrooms enables large-scale implementation of innovative learning technologies. Theories of multimedia and immersive learning suggest that immersive features can influence cognitive and affective learning processes. Therefore, 360-degree virtual reality learning environments (VRLEs) represent a promising area for further investigation. This study compares an interactive desktop-based low-immersive 360-degree VRLE with a traditional paper-and-pencil approach to investigate effects on students’ conceptual understanding of Galilean transformations of uniform motion in position-time graphs and on student engagement. Both conditions worked with multiple representations within a Predict-Observe-Explain framework to support active learning. A total of 313 students (176 female, 137 male) from the advanced track of upper secondary school participated in the pre-post classroom study. Conceptual understanding was assessed using the Inventory for Galilean Transformation of Uniform Motion in Position-Time Graphs. Students in both conditions improved their understanding, but no significant difference between conditions was observed. However, students using the VRLE reported a stronger inclination to perform well. These findings suggest that 360-degree VRLEs may support certain aspects of self-reported student engagement related to effort, but further research is needed to better understand how learning outcomes can be improved.
