Web Desk
The New Zealand Qualifications Authority (NZQA) has announced it will start using AI to mark Year 10 writing tests beginning in May 2025.
This decision, which follows a successful experimental phase, coincides with increasing worries about the improper application of AI in educational environments.
AI: Risk and Opportunity in Education
During an appearance before the Education and Workforce Select Committee, NZQA chief executive Grant Klinkum confirmed that artificial intelligence presents both challenges and opportunities in education.
He acknowledged that while AI can help streamline marking, it also raises concerns about academic integrity, especially for internally assessed work.
Currently, 80% of NCEA credits come from internal assessments, while only 20% come from external exams.
Klinkum suggested that this imbalance could be risky in the age of generative AI.
“If AI can be used to generate internal assessments, then external assessments become more crucial to verify student performance,” he said. “That 20% external may no longer be enough.”
Automated Marking to Speed Up Results
The new automated marking system will be applied to Year 10 writing assessments. AI will handle the first round of marking, followed by human checks on key assessments.
Klinkum explained, “We will have humans re-mark around 40% of test papers near the pass/fail threshold.
That way, students who are close to passing won’t be unfairly penalized.”
The decision follows a trial involving 36,000 students, where AI’s agreement rate with human markers was 80%—on par with human-to-human agreement.
Faster Feedback for Students
One of the major benefits of automated scoring is speed.
Results will be delivered to students more quickly, giving them more time to prepare for second attempts if needed.
“This will help schools and students better plan for upcoming assessments, especially if they need to retake the test,” said Klinkum.
Concerns About Cheating Persist
Last year, RNZ reported that some schools had already begun changing assessment methods to prevent AI-generated cheating.
As AI becomes more accessible, educators are grappling with how to maintain academic fairness while still benefiting from technology.
