Artificial intelligence is no longer a speculative concept relegated to the future or in 80s sci-fi movies. It has weaselled its way into our daily lives, embedded within the laptops and phones we use to navigate our education. The time for technological apprehension has passed. We must pivot from fear to strategic integration, establishing robust frameworks that harness AI as a catalyst for learning, without sacrificing the fundamental ability to think critically for ourselves.

This is a watershed moment for education in the Caribbean. While the latest Caribbean Examinations Council (CXC) guidelines demonstrate a commendable effort to keep pace with a rapidly shifting global landscape, they also introduce unprecedented risks that demand immediate, transparent discourse.

For decades, our educational paradigms have been hyper-fixated on the final grade, often at the expense of genuine comprehension. We strongly endorse the CXC’s newly implemented AI Scale Levels, specifically levels two and three for school based assessments (SBAs), because they rightfully transition the focus from the polished final product to the cognitive journey.

The mandate for an appendix of prompts is a wondrous play in academic integrity. By requiring students to document the exact parameters of their interactions with AI the CXC is demanding methodological transparency. Students can no longer merely present an answer; they must explain how they got to such an answer. This eradicates the façade of superficial knowledge. If a student cannot express themselves in a way that shows they would have grasped an understanding of a topic then they have not earned the academic credit. We welcome this rigour; it rewards genuine intellectual labour.

“I think artificial intelligence is used a lot by students every day, but I remain deeply concerned about its unfiltered application in the classroom. School must remain a sanctuary where AI supplements, rather than replaces, the rigorous work of true comprehension. AI is a brilliant tool for organisation, a catalyst for learning, and a springboard for ideation. However, we must draw a hard line to ensure it remains strictly a tool to aid us, and nothing more,” said Denise Bell, chairperson of the Student AI Task Force.

Despite these positive steps, a glaring vulnerability remains, and that is the immense fallibility of AI detection software. Our most pressing concern is the disenfranchisement of honest students due to machine error. Current detectors frequently generate false positives, punishing students simply because their original work is highly articulate or structurally complex. It is a gross injustice for a student’s academic future to be jeopardised, or hard work summarily dismissed on the unverified whim of an algorithm, especially for something as important as an SBA.

We firmly believe that the CXC must utilise these detectors merely as diagnostics, not as fact. A machine must never possess the unilateral authority to dictate a student’s future without rigorous human intervention.

To address the current chaos of AI validation we strongly recommend reducing the reliance on a fragmented array of “AI humanisers” and detection tools to a single, unified verification standard. Minimising this diversification will streamline the quota for verification, ensuring that all students are judged by the same, transparent metric rather than a confusing patchwork of third-party software.

When we cannot implicitly trust the equity of machines we must default to trusting ourselves. This is why we champion a new standard for CXC, which empowers students to defend their submissions orally. If an algorithmic detector flags a piece of work, the singular standard for verification must be a viva voce, an opportunity for the student to explain their methodology directly to an examiner.

We have conducted several consultations with a wide range of students, and a recurring concern was that their SBAs are being flagged for AI use, even when they insist they completed the work independently. This is deeply troubling to me, as there are even instances in which students have entered their own names into detection tools, and it would have flagged them for usage of AI.

Students also feel pressured to deliberately “dumb down” their work to avoid being flagged. This not only discourages genuine effort, but honestly undermines the overall integrity of the learning process, which contradicts the purpose of implementing AI detectors into SBA marking in the first place.

A step forward would be the implementation of a “selective scanning” protocol to address the widespread issue of entire documents being indiscriminately flagged by AI detection tools. This means that specific sections such as tables of contents, bibliographies, appendices, and raw data sets are excluded from scans. These areas naturally contain structured and repetitive language that can trigger AI detection.

In addition, we suggest “draft trailing”, wherein students are tasked to maintain a log of their document history in platforms like Google Docs or Microsoft Word. This time-stamped record of edits and revisions serves as strong, verifiable evidence of the student’s independent work and the natural development of their ideas over time.

Equally important is the introduction of a “secondary review” system to ensure that AI flags are not treated as final judgements. If a student’s IA use exceeds a reasonable threshold, the subject teacher who is familiar with the student’s writing style and academic ability should conduct a thorough review. Where necessary, students should also be allowed to orally defend their work, demonstrating their understanding of the content and the reasoning behind their responses. If a student can clearly explain their process, then the fundamental goal of learning has been achieved. Alongside this, schools must invest in improving teachers’ technical literacy by emphasising that AI detection tools are simply indicators and are not definitive proof. Training workshops and exposure to patterns in AI can help educators better understand the limitations of these systems, ultimately leading to fairer and more informed decision-making.

“It is fundamentally concerning that there is no standard for verification when it comes to AI verification, yet there are set statistics like 20 per cent and 30 per cent without actually knowing what those metrics are weighed against. If there is no verifiable standard then how is a student supposed to know what 20 per cent and 30 per cent of AI usage looks like? This creates a glaring issue of interpretation that has the potential to jeopardise an honest student’s work. And at the end of the day the education system fails if it penalises a student for their own work,” shared Asher Mundle, chairperson, Academic & Career Development Task Force.

Furthermore, we cannot discuss AI integration without confronting the glaring socio-economic realities of our region. Access to premium AI tools, subscriptions, and high-speed Internet is fiercely stratified. If the CXC and our regional governments fail to implement a comprehensive, universally accessible AI policy, we risk inadvertently engineering a tiered educational system, one where affluent students leverage technology to accelerate their success, while under-resourced students are systematically left behind.

We are demanding a policy that democratises access, providing every student with the resources and training necessary to utilise AI ethically and effectively. The integration of AI into our examinations requires utmost precision and a faithful commitment to fairness. The National Secondary Students Council proposes to CXC the immediate drafting of a Student AI Bill of Rights, a foundational document that explicitly guarantees a human review process and establishes a clear, unbiased appeals pathway for any student penalised for an AI-related discrepancy.

We stand ready to learn, adapt, and master this technology. But we refuse to be evaluated, categorised, or dismissed by a machine completely devoid of contextual understanding. Our intelligence, our struggles, and our triumphs are distinctly human. Our grading systems must remain undeniably human, too.

Brian Anderson is president of the National Secondary Students’ Council. Send comments to the Jamaica Observer or nsscja@gmail.com.

Automated AI assessors risk penalising students for independent thinking..