HAICA: A framework for reasoning transparency in AI-enabled higher education assessment

Abstract

Generative Artificial Intelligence (GenAI) has destabilized conventional assumptions about the evidentiary basis of higher education assessment. When students can produce polished and technically plausible outputs with AI support, final submissions alone may no longer provide sufficient grounds for judging learner reasoning, decision-making, or cognitive ownership. This paper addresses that challenge by developing HAICA (Human-AI Collaborative Assessment) as a conceptual framework for assessment redesign in AI-enabled higher education. HAICA is grounded in constructivism, self-regulated learning, assessment for learning, human-in-the-loop governance, and a socio-technical perspective. The framework is organized around three interrelated mechanisms: Human Ownership, the Human-AI Collaborative Cycle, and Governance Controls which together support reasoning transparency as the intended assessment outcome. The paper makes three contributions. First, it reframes the GenAI challenge in assessment from a problem of detection to one of evidentiary design. Second, it positions reasoning transparency as a central construct for restoring defensible academic judgement in AI-enabled contexts. Third, it translates this framework into explicit design principles and propositions for future empirical testing. An implementation-informed illustration demonstrates the framework’s practical plausibility. The paper argues that defensible assessment in the GenAI era depends on redesigning assessment so that human reasoning remains visible, assessable, and governable.