NZQA Confirms AI Marking System Does Not Disadvantage Students
Independent review validates AI-driven NCEA literacy assessment accuracy and human oversight protocols.
1AI Assessment System Validated
The New Zealand Council for Educational Research recently reviewed the NZQA’s Automated Text Scoring model. The independent audit concluded the system is fit for continued operational use in NCEA literacy writing assessments. This technology has been utilized since 2025 to streamline grading processes, ensuring students receive their academic results significantly faster than before.
2Performance and Accuracy Metrics
The AI system demonstrated strong performance, with scores exactly matching human markers in 79.6% of instances. Furthermore, 84.5% of total AI scores remained within one point of human evaluations. While minor discrepancies occurred, the NZQA maintains that these variations did not negatively impact the final grades awarded to the participating students.
3Human Oversight Remains Central
Despite the integration of machine learning, human markers retain final authority over the assessment process. Experienced professionals cross-check approximately 30% of all scripts, specifically focusing on those near the Achieved or Not Achieved boundary. If any disagreement arises between the AI and a human marker, the human evaluation always takes precedence.
4Addressing Potential Bias Concerns
The NZCER review specifically analyzed the system for potential demographic biases. Findings indicated no substantial differences in how AI and human scores aligned across various gender, ethnicity, or socioeconomic groups. However, the report recommended ongoing monitoring and consistent quality checks to ensure the model maintains its high standards of fairness and accuracy.
5Efficiency Gains for Schools
Implementing AI marking, combined with other process improvements, has yielded tangible benefits for the education sector. In 2025, results were delivered to schools and students up to four and a half weeks earlier than in 2024. This efficiency allows students more time to prepare for subsequent assessments if they do not pass.
6Scope of Current Technology
The NZQA currently restricts AI usage to the literacy writing co-requisite assessment. The authority explicitly advises schools against using AI for internal assessments, noting that individual schools generally lack the volume of scripts required to train a reliable model. Consequently, the technology remains centralized under the NZQA’s strict, controlled operational framework.
7Future Expansion and Testing
The NZQA is exploring the potential to apply AI to more complex assessment standards, such as distinguishing between Achieved, Merit, and Excellence grades. Current testing utilizes scripts from previous years rather than active student work. Any future expansion of this technology will require rigorous testing and the continued maintenance of essential human oversight.
8Confidence in Hybrid Models
Four independent reviews, including two by the NZCER and two by technical consultancy ClearPoint, have reinforced the NZQA’s confidence in the system. While the authority continues to innovate, officials emphasize that human markers will remain central to the assessment process for the foreseeable future, ensuring academic integrity remains protected during technological transitions.
Original article: Newstalk ZB — NZQA Confirms AI Marking System Does Not Disadvantage Students