Developing Evaluative Judgment and Critical AI Literacy through Transformative Assessment

This best practice demonstrates how generative artificial intelligence (GenAI) can be purposefully integrated into assessment design to strengthen students’ evaluative judgment, critical thinking, and AI literacy. Rather than treating GenAI as a shortcut or threat to academic integrity, the approach embeds AI use within structured learning sequences that emphasise critique, reflection, and quality judgment. Students engage with AI-generated content alongside peer and lecturer feedback, using explicit rubrics and exemplars to evaluate work. Assessment is reframed as a learning opportunity, supporting disciplinary understanding while developing ethical, critical, and reflective engagement with GenAI.

This best practice was implemented at Stellenbosch University within undergraduate teaching contexts, particularly in the Faculty of Economic and Management Sciences Extended Curriculum Programme. The practice followed a structured, scaffolded learning design in which foundational academic skills were prioritised before introducing GenAI tools.   

Students initially focused on critical thinking, argumentation, and disciplinary writing without AI support. Only once these foundations were established were GenAI tools such as ChatGPT, Grammarly, and the Hemingway App introduced as refinement and feedback tools. GenAI was deliberately positioned as an additional “voice” within a broader feedback ecology that included peer feedback, lecturer feedback, rubrics, and exemplars.   

Students were required to compare feedback from peers and GenAI, explicitly justifying which suggestions they accepted or rejected. This reflective process made the criteria for quality visible and strengthened students’ ability to exercise evaluative judgment. In assessments, students critically analysed AI-generated texts, identifying weaknesses in argumentation, referencing, and contextual accuracy. By integrating GenAI into both learning activities and formal assessment tasks, the institution ensured that AI engagement supported deep learning rather than replacing it.

Impact was assessed using qualitative and performance-based indicators embedded within teaching and assessment, with a focus on learning processes rather than grades alone.

Student reflections showed increased metacognitive awareness, as students explained how they engaged with peer and GenAI feedback and justified which suggestions they accepted or rejected. These reflections demonstrated clearer understanding of academic quality, argument strength, and the limitations of AI-generated outputs.

Assessment artefacts provided additional evidence of impact. In invigilated assessments, students critiqued GenAI-generated essays, identifying fabricated references, weak arguments, missing contextual grounding, and unsupported claims. This demonstrated the development of critical AI literacy using disciplinary criteria.

Comparative analysis of student work across the semester indicated improvements in coherence, argumentation, and structural clarity. Classroom observations and lecturer reflections further highlighted increased student confidence in discussing AI, reduced anxiety around academic integrity, and more nuanced ethical reasoning.

Together, these indicators show that the practice supported evaluative judgment, critical engagement with AI, and transferable graduate attributes.

• Scaffolded assessment design
• Explicit rubrics and exemplars
• Peer and lecturer feedback processes
• Purposeful use of GenAI tools (ChatGPT, Grammarly, Hemingway App)
• Reflective and evaluative learning activities