Automated and AI-assisted assessment and perceived fairness in a polyethnic university: digitalization as a potential pathway to interethnic equity in Kazakhstan's higher education

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

The digital transformation of higher education in Kazakhstan unfolds against the backdrop of the 2022 constitutional reforms and the policy agenda of building a “Just Kazakhstan,” in which equal access to education for a multiethnic student population is a declared priority. This study examines how university students evaluate the fairness of automated and AI-assisted assessment, and whether such technologies are seen as reducing subjectivity, corruption risks, and ascriptive (ethnic and linguistic) bias. Respondents were not asked to compare formats within a single item, so the study reports perceptions of automated assessment rather than a measured contrast with examiner-based assessment. A cross-sectional survey was conducted among 300 undergraduate and graduate students at a large public pedagogical university in Almaty (June 2026). The instrument comprised 17 Likert-type items measuring perceived fairness of automated and AI-assisted assessment, trust, algorithmic bias awareness, and attitudes toward AI-based simulations of intercultural disputes, together with demographic and experience variables. Students rated automated assessment as providing equal conditions and reducing corruption opportunities substantially higher than the scale midpoint, whereas the linguistic equivalence of test materials in Kazakh and Russian received the lowest fairness ratings. The composite fairness index did not differ significantly by ethnolinguistic self-identification, study level, or language of instruction; it was modestly higher among students of rural origin, and Kazakh-medium students rated the Kazakh–Russian equivalence of test items significantly lower than Russian-medium students. Exposure to automated assessment was positively correlated with trust in it and unrelated to awareness of algorithmic bias; the cross-sectional design does not permit causal inference. A hybrid model combining AI scoring with human oversight was the most preferred examination format. Read through the lens of procedural justice, which serves here as an interpretive framework rather than a measured construct, the findings suggest that students value automated assessment for properties that correspond to consistency, neutrality, and bias suppression, and believe it to weaken ascriptive influences in a polyethnic academic environment, while remaining aware of algorithmic bias and translation-equity risks. The study reframes assessment digitalization in bilingual systems as a problem of item-bank equity rather than of algorithms as such, and derives implications for university assessment policy and for the digitalization strategy of higher education. Whether assessment digitalization advances the constitutional agenda of interethnic equity is posed as a hypothesis for further research, not as a finding of this survey.