Cultural and ethical responsiveness of AI-driven assessment: Gender and geographical equity, adolescent mental health, and inclusive educational outcomes in Calabar, Nigeria

The increasing integration of artificial intelligence (AI) into educational and mental-health assessment raises concerns about cultural relevance, psychometric fairness, accessibility, and ethical governance, particularly in resource-constrained contexts such as Nigeria. This paper examined the cultural, ethical, and psychometric implications of AI-driven assessment, focusing on differential item functioning (DIF), adolescent mental-health screening, community-centred assessment, and AI companionship. This integrative literature review synthesised published research on cultural and ethical assessment practices, gender- and location-based DIF in AI-generated and West African Examinations Council (WAEC) English Language items, AI-based mobile screening of adolescent mental-health conditions, community-centred assessment, and AI companionship among Nigerian university freshmen. The synthesis incorporated four directly relevant Calabar-focused studies, a Nigerian AI companionship study, international empirical literature, and child-centred AI guidance. The reviewed evidence indicated that culturally responsive and ethical assessment practices support inclusive assessment; both AI-generated and WAEC English Language items exhibited gender- and location-related DIF, although lower DIF magnitudes were reported for AI-generated items. Community-centred approaches strengthened ownership and accountability, while AI-based adolescent screening demonstrated potential for early identification alongside concerns about privacy, demographic bias, validation, and clinical follow-up. These findings informed an integrated framework connecting accessibility, cultural and linguistic validity, subgroup fairness, data governance, human oversight, and post-assessment support. AI-driven assessment can advance equitable educational and mental-health outcomes when culturally grounded, psychometrically validated, ethically governed, and supported by professional judgement and community participation. AI should complement rather than replace human expertise in assessment and intervention.