Generative Artificial Intelligence as Evaluative Scaffolding: Effects on Students’ Perceptions and Dispositions in an Authentic Learning Task

In recent years, the integration of generative artificial intelligence (GenAI) into educational contexts has raised important pedagogical questions, particularly regarding its role in authentic assessment processes and the development of evaluative judgement. This study explores the integration of GenAI within an experiential learning and authentic assessment device centered on the construction, revision, and self-assessment of rubrics, investigating how this experience is associated with changes in the perceptions and dispositions of university students who are future primary school teachers. Using a single-group pre-post quasi-experimental design, 144 participants constructed an assessment rubric and subsequently revised it using both an exemplar and GenAI tools, in a sequence consistent with Kolb's experiential learning cycle. Three dimensions were measured before and after the intervention: reflective and criterion-based approaches, attitudes towards AI, and technostress/computer anxiety. Results, analyzed through Wilcoxon signed-rank tests and non-parametric correlations, show statistically significant changes across all three dimensions: an increase in reflective and criterion-based approaches (r = −0.725), a marked increase in positive attitudes towards AI (r = −0.890), and a reduction in technostress (r = 0.421). Lower initial levels across the dimensions were associated with larger changes, while no significant differences emerged related to socio-demographic characteristics or the tool perceived as most useful. Findings suggest that GenAI can function as genuine evaluative scaffolding — an agent-to-learn-with rather than a substitute for human judgement — capable of fostering critical reflection and the formation of evaluative judgement, provided it is integrated into intentionally designed and pedagogically mediated learning environments.