.
This paper examines standard setting in health professions education through the lens of historical epistemology. Rather than treating standard setting as a neutral technical procedure, we analyse how it became thinkable as a distinct assessment practice and how its forms have shifted across different historical contexts, driven by evolving concerns and epistemic assumptions. We develop a five-part analysis of historically situated epistemic regimes of standard setting, spanning forms of threshold judgement in which standards were operative but not yet explicit, through prescriptive, predictive, and performative approaches, to the contemporary landscape of programmatic assessment and entrustment. Across these regimes, standard setting practices are informed by changes in styles of reasoning, characteristic questions, forms of authority, and modes of justification that shape what counts as minimum competence, how it is determined, and who is authorised to judge. Our central argument is that standard setting is historically contingent and epistemically constructed, while remaining indispensable to assessment practice. Recognising this does not weaken standard setting; rather, it clarifies the assumptions about evidence, expertise, authority, and acceptable risks embedded in different methods, situates those methods within shared professional commitments and historically shifting interpretations of competence, and makes their justificatory bases more explicit. We argue that standard setting practices are best understood and defended through contextually grounded validity arguments, informed by epistemological reflection and attentiveness to historicity.
Covid-19 turbocharged the adoption of digital technologies in mainstream schools, which necessarily changed the shape of private supplementary tutoring as well. This article traces these dynamics in Cambodia, outlining the concept of Do It Yourself (DIY) Tutoring. DIY Tutoring is curriculum-linked learning outside of the classroom that a student engages with autonomously through digital platforms. The digital nature of DIY Tutoring allows for a level of flexibility and personalization uncommon in other forms of self-guided tutoring. Most importantly, DIY Tutoring alters the traditional understanding of the cost of tutoring from direct user-fees to also include indirect fees common in platform capitalism. The article reports on empirical evidence in Cambodia to articulate a framework of DIY Tutoring. Although this study focuses on Cambodia, adding to the literature on shadow education in the country, the concept of DIY Tutoring can be applied more widely, opening new directions in shadow education research.
First Nations students’ transitions into university are frequently treated as linear and uniform, despite diversity across language groups, communities and life trajectories. This study examines how a dedicated First Nations student centre in a metropolitan Australian university supports students across the “university lifespan”: imagining university, application and enrolment, first year and ongoing study. The paper draws on literature, administrative completion data, and interviews with 16 First Nations students and 8 staff. Thematic analysis identified three interlocking mechanisms through which the Centre strengthened success: pathway translation, place-based belonging, and wraparound scaffolding across academic, housing/financial and wellbeing domains, especially during the first weeks of semester. Students described the Centre as their primary access point to university systems, enabling programs, tutoring, accommodation supports and health services. Findings indicate that improving First Nations participation and completion requires not only expanded entry pathways but sustained investment in culturally safe infrastructures that hold students through transition pressure points and support diverse, non-linear journeys.
Hybrid scoring systems route responses that an automated scoring engine cannot score confidently to human raters. Whether that routing protects student groups equally is, to our knowledge, unreported. We audit three engines, a fine-tuned RoBERTa (quadratic weighted kappa .841) and two zero-shot large language models (.565 and .216), on two corpora with student demographics: PERSUADE 2.0 (15,593 argumentative essays) and ELLIPSE (3,897 essays by English language learners, each scored by two independent raters). Routing uses split conformal prediction, which turns engine probabilities into score sets that contain the true score with a chosen probability, and we compare one pooled calibration threshold against one threshold per group across six groupings, two score functions, six risk levels, and 20 calibration splits. Pooled calibration met its overall guarantee everywhere while distributing it unevenly: English language learners received 5.4 percentage points less than the guaranteed coverage under one engine and 3.3 points more under another, invisibly to aggregate accuracy or confidence. The largest disparity was not demographic: the fine-tuned engine, with demographic gaps of at most 1.9 points, spread coverage across grade cohorts by 10.5 points. Per-group calibration restored the guarantee in every condition tested, at the cost of the repaired group's automation rate (2.4 to 0.2 percent for English language learners). We recommend that scoring programs state routing rules as coverage guarantees and report group-conditional coverage for every group on which results are reported.
PISA in Brief summarises the key findings from PISA 2025 for Australian students. It compares Australia’s performance with other participating countries and economies, examines trends over time, and highlights differences in achievement across Australian states and territories and key demographic groups. PISA uses 2 main measures to describe student achievement. Mean scores show overall performance, while proficiency levels (ranging from below Level 1 to Level 6) explain the types of knowledge and skills students can typically demonstrate. This report focuses on differences that are statistically significant (in other words, differences that are unlikely to have occurred by chance). The report presents Australia's performance in science, reading, mathematics, and learning in the digital world including analysis by sector, gender, geographic location, socioeconomic background, First Nations background, immigrant background, and language background.
