Two sets of figures reached Jamaica's Ministry of Education in the same week this August. First, 4,048 registered CSEC candidates, 1,875 of them from public schools, did not turn up for the May and June 2026 sitting, after the government had paid their subsidised exam fees. Second, the Caribbean Examinations Council recorded 128 confirmed irregularity cases at that sitting, up from 80 the year before, including a category it had never tracked: artificial intelligence misuse.
AI can help with both. Attendance models can flag candidates likely to miss an exam weeks before the sitting, and properly calibrated integrity checks can catch undisclosed AI use without accusing honest students. Neither needs CXC to redesign its exams. Both use data CXC and the ministries already collect.
The two stories are usually told apart, one about wasted subsidy and absent parents, the other about students using ChatGPT. They come from the same sitting and the same system: a regional exam body serving sixteen member states with very different resources, using processes built before every candidate carried a chatbot in their pocket.
What the no-shows cost Jamaica
Jamaica subsidises CSEC fees in core subjects so no student is priced out of the qualifications that most decide further study and work. In 2026 the subsidy covered 20,459 English Language candidates and 18,309 Mathematics candidates at an average of J$4,947.50 each, about J$191.8 million in total. Candidates in parishes hit by Hurricane Melissa were covered at J$4,690 and others at J$5,205. Every subsidised subject a candidate skipped was money spent for nothing. Across the region, more than 35,000 candidates missed the May and June CSEC and CAPE sittings.
Permanent Secretary Dr Kasan Troupe told the Jamaica Gleaner on 30 August that every unjustified absence is a loss of taxpayers' money. Linvern Wright, principal of William Knibb Memorial High School, called the absences a sign that students are not taking their education seriously. Stewart Jacobs, president of the National Parent-Teacher Association, put the responsibility on parents. Educator Ralston Smith has proposed contracts requiring families to repay the fee when a student skips without a valid reason.
None of them has a way to know, before exam day, which candidates will not come. That is a prediction problem, and schools already hold the data to attempt it.
AI misuse became a new category of exam offence
CXC's Director of Operations, Dr Nicole Manning, reported that irregularities rose from 80 in 2025 to 128 in 2026. At CSEC they nearly doubled, from 57 to 106; CAPE was steady at 22 against 23. The main categories were unauthorised devices, collusion, unauthorised material and AI misuse, which she called "the new one on the block".
CXC has not published a breakdown of the AI cases. Reports in the Gleaner, the Observer and Barbados Today point to chatbots used during timed, closed-book components, with the output submitted as the candidate's own work. That is separate from the disclosed AI use CXC began allowing in some school-based coursework in 2026. The first is permitted; the second can lead to a cancelled grade or disqualification.
CXC also granted compassionate consideration to 359 candidates (288 at CSEC, 71 at CAPE), down from 377, and activated disaster recovery protocols after Hurricane Melissa disrupted about half of Jamaica's exam centres. Absence, misconduct and storm disruption all hit the same stretched system.
The gap between territories is wide
Regionally, 72.18 percent of CSEC entries earned acceptable grades in 2026, against 93.52 percent at CAPE. The averages hide large differences:
| Measure (2026) | Result |
|---|---|
| CSEC Mathematics, 14 territories | 43 to 77 percent |
| CSEC Mathematics, 7 territories | 30 to 38 percent |
| CSEC Mathematics, regional average | 42 percent (up 4 points) |
| CSEC Mathematics, Jamaica | 48.7 percent (44 percent in 2025) |
| CSEC Social Studies, lowest and highest territory | 44 and 100 percent |
Dr Manning flagged the disparity herself. A student in a territory scoring 30 percent in Mathematics has less access to qualified teachers and past-paper practice than one in a territory scoring 77 percent, yet both receive the same certificate. Subsidy reform and cheating enforcement will not close that gap; more teaching support for the students furthest behind will.
The Core Argument
Wasted subsidy, rising misconduct and the gap between territories share a cause: the system learns what happened only after the exam, once the seat is empty, the answer is submitted or the results are out. Prediction before the sitting and better checks during it can shorten that delay for all three.
Four uses of AI that fit CXC's existing data
1. Predicting absence
A model trained on attendance records, mock exam participation, past registration history and disruption events such as storm damage can score each candidate's likelihood of sitting weeks ahead. A school with a cluster of high-risk English or Mathematics candidates gets a list of names for a call home, a transport check or a guidance session. Parents stay responsible, as Stewart Jacobs says; the list gives schools time to help them.
2. Detection calibrated on Caribbean writing
Detectors built on American and British writing misread Caribbean English, and a false accusation costs a student as much as a real one. CXC needs a detection layer trained on regional scripts that flags unusual answer similarity or sudden changes in writing style for a trained examiner to review. The examiner decides, as a bank's fraud team reviews flagged transactions rather than freezing every unusual one.
3. Adaptive tutoring where scores are lowest
The seven territories scoring 30 to 38 percent in Mathematics lack consistent access to qualified teachers and exam-style practice. Adaptive platforms that adjust difficulty to each student's answers have raised test scores in under-resourced schools in parts of Africa and South Asia, India's Mindspark programme being one studied example. A pilot in those territories, in Mathematics and English, targets the subjects where both the subsidy and the gap concentrate.
4. Checking registrations before paying the subsidy
Instead of paying for every registration and learning months later who stayed away, the ministry could cross-check registrations against attendance and mock participation before paying, and direct the high-risk cases to support. The goal is to move some of the money from empty seats to transport stipends and tutoring, not to penalise students.
I am less certain about the fourth idea than the others. A model that flags students as unlikely to attend can become a reason to withhold support from them, which would defeat its purpose. Any ministry that tries it should publish how flags are used.
Why the stakes are high for students
CSEC and CAPE decide entry to regional universities, applications to UK and Canadian institutions, and many first jobs. Subsidy lost to empty seats is money that could have paid for a bus fare or after-school tutoring. A detector that confuses honest Caribbean writing with AI output could cancel the wrong grades, and one cancelled subject can cost a university place. And if the territory gap keeps widening, employers and universities may start treating the same certificate differently depending on where it was earned.
What to do next
- Jamaica's Ministry of Education with CXC: run an absence-prediction pilot before the January 2027 sitting on the 2026 attendance and registration data, and compare turnout between flagged candidates who received follow-up and those who did not.
- CXC: commission a detection model trained on CSEC and CAPE scripts, with a written rule that no grade is cancelled without review by a trained examiner.
- Ministries in the four to six territories with the lowest Mathematics and English results: apply jointly for an adaptive tutoring pilot and track results over the next two CSEC sittings.
- School principals: check mock exam attendance against CSEC registrations by March each year and contact every registered candidate who missed the mock.
- Parents: if your child will not sit a subsidised subject, tell the school before the withdrawal deadline so the place can be managed.
Frequently Asked Questions
What are CSEC and CAPE, and who runs them?
CSEC (Caribbean Secondary Education Certificate) and CAPE (Caribbean Advanced Proficiency Examination) are the region's main school-leaving qualifications, set and marked by the Caribbean Examinations Council. CSEC is usually taken at 15 or 16, roughly equivalent to UK GCSEs. CAPE is a sixth-form qualification accepted for university entry in the Caribbean, the UK, Canada and the United States.
Are AI writing detectors reliable enough to decide a cheating case?
No. Detectors give probability scores, not proof, and they produce false positives. A 2023 Stanford study found that popular detectors wrongly flagged a large share of essays written by non-native English speakers as AI-generated, which is a warning for any region whose English differs from the training data. A detector score should start a review, which then looks at drafts, notes and the student's own explanation.
Can a candidate challenge an irregularity finding?
Yes. Queries go through the candidate's school and the Local Registrar, CXC's national office in each territory, within CXC's published deadlines. Ask in writing for the evidence behind the finding, and keep any drafts, notes or school records that show the work was the candidate's own. Deadlines are short after results are released, so act quickly.
How do students apply for compassionate consideration?
Through the school, which submits the request and supporting documents, such as a medical certificate or death certificate, to the Local Registrar. Apply as soon as possible after the affected exam, because CXC sets a deadline for these requests. It is meant for illness, bereavement or other events beyond the candidate's control on or near the exam date.
Can students use AI tools for CXC coursework at all?
In some school-based assessments, yes, if the use is disclosed as CXC's guidance requires. Undisclosed use, and any use during timed exam components, is treated as misconduct. Rules can differ by subject and are updated, so students and teachers should read CXC's current AI guidance for each syllabus before starting the SBA.