AI evidence extraction

L’intelligenza della scrittura. Pratiche, percezioni e tensioni educative nell’epoca dell’IA

Angela Spinelli · Cineca Institutional Research Information System (Tor Vergata University) · 2025

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10

Relevance

0/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

Methodology & findings

Study design

Qualitative exploratory research involving approximately 250 university students from a General Didactics course (2024/25) at Tor Vergata University in Rome.

Sample

N = 250, 3 groups

Primary method

Qualitative analysis using critical discourse analysis (CDA) protocol. Data were coded at multiple levels using: (1) bottom-up approach through recursive and comparative reading of textual data, and (2) top-down approach informed by theoretical and pedagogical reflection. Co-occurrence analysis (Wicks, 2017) was employed to identify units of meaning (Tesch, 1990). Analysis was conducted using NVivo software. Triangulation was ensured through phases of coding, intra-textual and inter-textual analysis.

Main result

The study found that students describe AI as "una 'seconda mente', un facilitatore nei passaggi iniziali e di rifinitura della scrittura" (a 'second mind', a facilitator in the initial and refinement stages of writing). However, alongside functional opportunities, "una costellazione di paure che si condensano nella categoria del rischio percepito" (a constellation of fears that condense into the category of perceived risk) emerges, with students describing AI as "comoda ma pericolosa" (convenient but dangerous). The analysis identified three prevalent discursive axes: pragmatic-technical, ethical-critical, and educational-institutional, revealing that "il soggetto scrivente è ancora attivo, ma spesso in modo solitario, incerto, frammentato" (the writing subject remains active, but often in a solitary, uncertain, fragmented manner).

Reports effect sizes.

Research paradigm

qualitative-interpretivist

Author conclusions

The authors conclude that "educare alla scrittura con l'IA significa abitare l'epoca del post-plagio" (educating in writing with AI means inhabiting the post-plagiarism era). They argue that the pedagogical task is not to introduce AI but "portare alla soglia della consapevolezza e della dicibilità i modi in cui essa è già coautrice dei testi degli studenti/esse e come concorre al loro processo produttivo e creativo" (to bring to the threshold of awareness and speakability the ways in which it is already co-author of students' texts and how it contributes to their productive and creative process). They emphasize that fears should not translate into repressive policies but should "aprano ad uno spazio ampio di confronto, dialogo e costruzione partecipata di orizzonti nuovi" (open a wide space for dialogue and participatory construction of new horizons).

Risk of bias

Selection bias: Forum participation was mandatory but focus groups were voluntary; Social desirability bias: Students may have self-censored due to institutional distrust regarding AI use; Teacher influence: Forum posts were mediated and annotated by the instructor, potentially shaping responses; Contextual differences: Different discursive environments (written Forum vs. oral focus groups) may have influenced expression patterns; Selection bias: focus group participation was voluntary, potentially excluding students with negative views or no AI experience; Context bias: Forum discussions occurred within formal pedagogical setting with teacher mediation, while focus groups were informal; Temporal bias: data collected at course completion may reflect retrospective rationalization; Researcher positionality: teacher-researcher conducted study in own course, potentially influencing student responses; Selection bias: Focus group participants were volunteers, not representative of all course students; Social desirability bias: Institutional distrust may have inhibited candid discussion of AI use; Moderator effect: Forum posts were mediated by instructor commentary; focus groups were unmediated but informal setting differences between corpora affect comparability; Self-selection bias in forum participation

Open questions raised

The authors identify the need for: (1) making AI use more explicit and integrated into institutional educational processes rather than remaining tacit and informal; (2) clarifying ethical and epistemological limits of AI use, including the presence of bias and algorithmic opacity (black box condition); (3) developing institutional dialogue to thematize multiple tensions as part of learning and training processes; (4) constructing shared ethics for AI use through intergenerational dialogue; (5) advancing data literacy to support digital citizenship.
Need for explicit integration of AI as an educational tool within institutional settings
Lack of clear ethical and epistemological boundaries regarding AI use in academic writing
Need for institutional dialogue to thematize tensions and concerns about AI-mediated writing
Absence of guidance on intergenerational learning in the context of AI-enhanced education
Need for transition from surveillance-based control to trust-based relationships in managing AI use

Extracted from: pdfAgreement 58%

Explore related topics

AI Manuscript Writing Academic Writing Qualitative