scale student students measurement disproportionately

The gap

Supporting evidence — 2 representative gaps

• Gender and functional differentiation in generative AI usage among Malaysian higher education student (2026) · doi

  Self-report bias and social desirability Self-report measures introduce multiple biases that warrant careful consideration, particularly given the gendered nature of technical computing tasks. Students may report ChatGPT usage patterns that align with gender- stereotypical expectations rather than actual behaviour, for instance, males may overreport coding use to conform to masculine technical identity norms, while females may underreport technical applications to avoid violating gender-typical behaviour expectations. Such response biases would artificially inflate observed gender gaps beyond true behavioural differences, potentially exaggerating the functional stratification we documented. Lavidas et al. (2022) demonstrate that social desirability effects in student self-reports vary significantly by context, with students exhibiting different response patterns in lecture halls versus laboratory settings where technical competence is more salient. In our study, survey administration through instructor referrals may have activated evaluative concerns, particularly for female STEM students in male-dominated programmes where technical identity is under constant scrutiny. This context-dependent desirability effect could contribute to the apparent "STEM amplification" we observed, where female engineering students may have systematically underreported coding-focused ChatGPT use to manage impressions in technically evaluative contexts. If social desirability operates more strongly in STEM than non-STEM contexts, our observed 33.4-point gap in Applied Sciences could partially reflect measurement artifact rather than true behavioural differences. Additionally, recall bias may differentially affect responses across task types. Coding tasks are often more salient and memorable than routine text-processing activities (proofreading, summarising), potentially leading males to more readily recall and report technical applications while females more readily recall communicative uses. The 12- task Likert scale format requiring retrospective frequency estimates over unspecified timeframes is particularly susceptible to such retrieval asymmetries. If males disproportionately remember coding instances while females disproportionately remember writing instances, our latent profiles may partially reflect differential recall rather than differential behaviour. Survey anonymity and online administration mitigate but do not eliminate desirability bias. While participants could not be identified individually, awareness that data would be aggregated by gender and field may have activated group-level identity concerns. Female engineering students, cognisant of stereotypes about women's technical capabilities, may have felt implicitly evaluated as representatives of their gender category, prompting conservative reporting of AI-assisted coding regardless of actual usage.

  Keywords: technical desirability students gender coding report stem recall self bias social particularly rather behaviour males
• Development of a 3D Animal Miniature Learning Media Based on Augmented Reality Using Web Edu Assembler to Improve Elementary School Students’ Science Learning Outcomes (2026) · doi

  The research was limited to 55 students in one elementary school context; larger-scale implementation across diverse school settings and student populations is needed to validate generalizability.

  Keywords: school limited students elementary context larger scale implementation across diverse settings student populations needed validate