Ghulam Nabi Shakir PhD
For decades, public examinations in Pakistan have remained wedded to manual processes—paper scripts, human examiners and time-consuming tabulation.
While this system has pro-duced generations of graduates, it has also attracted persistent criticism for delays, inconsis-tencies, human error and allegations of bias. In recent years, however, a quiet but significant shift has begun. The initiative of Intermediate and Secondary Education Boards to introduce electronic marking (e-assessment) represents an important transitional phase—one that may ultimately pave the way for AI-assisted assessment, a far more advanced, reliable and future-oriented model.
The move towards e-assessment, where examiners mark scanned answer scripts on digital platforms, is often presented as a technical upgrade. In reality, it is a structural reform with deep implications for transparency, efficiency and credibility. By digitizing scripts, boards have already reduced risks associated with physical handling, loss of papers and post-examination tampering. Time required for marking has shortened, examiner allocation has become more flexible and audit trails now exist where every mark can be traced back, re-viewed and verified.
Yet, e-assessment remains essentially human-centered. The examiner still interprets judges and awards marks. Technology merely acts as a conduit. Artificial Intelligence, on the other hand, has the potential to fundamentally transform the very logic of assessment—without replacing human expertise, but by augmenting it.
AI-assisted assessment does not imply handing over the fate of students to machines. Rather, it refers to the use of intelligent algorithms to support examiners in marking, moderation, anomaly detection and quality assurance. Internationally, AI tools are already being used to evaluate structured responses, short answers, mathematical solutions and even essays—within carefully designed pedagogical and ethical frameworks.
In Pakistan’s context, AI could initially function as a co-examiner. It can pre-score responses based on marking rubrics, flag inconsistencies, highlight outlier marks and detect potential bias or fatigue effects in human marking. Final authority would remain with trained examiners, but their decisions would be supported by data-driven insights rather than subjective judgment alone.
The public debate on AI in assessment usually revolves around fear—fear of dehumanization, fear of error, fear of exclusion. What remain largely invisible are the systemic benefits that AI assistance can bring, particularly to a high-volume examination system like ours. One of the biggest challenges in public examinations is inter-examiner variability. Two competent examiners may still award different marks to the same response. AI systems, once trained on approved rubrics and benchmark scripts, apply criteria uniformly. This consistency does not eliminate human discretion but significantly reduces unjustified variation. AI-assisted systems create detailed logs: why a response received a certain score, which ru-bric elements were met and where marks were deducted. Such explainable assessment mod-els can strengthen student and parent trust, especially in rechecking and appeal processes that currently rely on opaque mechanisms. AI can analyze patterns across thousands of scripts to identify problematic questions, ambi-guous wording, or flawed marking schemes. Instead of discovering these issues through pro-tests or litigation, boards could proactively improve paper-setting and assessment design. Manual marking, especially under tight deadlines, inevitably leads to fatigue-induced errors. AI assistance can handle repetitive evaluation tasks, allowing examiners to focus on complex responses that genuinely require professional judgment. Speed and quality are often treated as opposing goals. AI challenges this assumption. With intelligent automation, result timelines can be shortened while simultaneously improving reliability—a critical factor for admissions, scholarships and career progression.
Skepticism toward AI is neither irrational nor unwarranted. Concerns about algorithmic bias, data privacy and contextual understanding must be addressed through policy, not denial. AI systems are only as fair as the data and frameworks governing them. This is why gradual integration, pilot testing, independent audits and continuous human oversight are essential. Moreover, AI should not be imported as a black-box solution. Local curriculum alignment, language sensitivity (especially in Urdu and regional languages) and examiner training must be integral to the process. The current e-assessment infrastructure of boards provides a solid foundation upon which such contextualized AI tools can be developed. The introduction of electronic marking by Intermediate Boards should be seen not as an end, but as a bridge—a necessary preparatory stage in a longer reform journey. Globally, assess-ment systems are moving away from purely summative, high-stakes models toward more intelligent, data-informed evaluation frameworks. Pakistan cannot afford to remain an outlier. AI-assisted assessment, when thoughtfully implemented, is not a threat to academic integrity; it is an opportunity to strengthen it. It promises fairness without rigidity, efficiency without haste and innovation without recklessness.
The real question, therefore, is not whether AI should enter our examination system, but how responsibly and how soon. If e-assessment was about digitizing paper, AI-assisted assessment is about digitizing trust. And for a system often criticized for its credibility, that may be the most transformative benefit of all.
(Author is a formal Principal of a Public School, presently working as a Director
Research in a company.
gnshakir@hotmail,com).
