Hybrid E-Assessment in Higher Education: Semi-Automated Grading of Paper-Based Written Examinations

Researchers propose a hybrid e-assessment system for higher education that combines paper-based examinations with semi-automated grading using vision-capable large language models. The approach addresses limitations of fully digital assessment while maintaining pedagogical integrity and scalability through handwritten character recognition and validation protocols.

Educational institutions face mounting pressure to scale summative assessments while preserving exam integrity and pedagogical quality. Fully digital systems often impose didactic constraints through closed-question formats, while hybrid approaches struggle with technical reliability and legal compliance in large cohorts. This research tackles a genuine institutional pain point by proposing a pragmatic middle ground that leverages recent advances in multimodal AI.

The hybrid model retains paper-based problem-solving tasks—preserving open-ended assessment—while introducing semi-automated grading through structured answer capture. Students write responses in designated table fields, which are then digitized and processed by vision-capable large language models. A two-pass validation mechanism combined with solution key comparison mitigates handwriting recognition errors, the critical technical bottleneck. This approach addresses organizational constraints (paper remains familiar to institutions), technical challenges (reliable character recognition), and legal requirements (audit trails, fairness documentation).

For educational technology providers and institutions, this represents a viable pathway to assessment automation without sacrificing pedagogical value. The methodology could accelerate adoption of AI-assisted grading in universities currently resistant to fully digital transitions. However, success depends entirely on real-world validation of handwriting recognition accuracy under examination stress conditions—a variable not yet thoroughly tested at scale.

Institutional buyers should monitor implementation case studies and benchmarking data from pilot deployments. The viability of this approach hinges on whether character recognition accuracy reaches thresholds sufficient for high-stakes summative assessment, particularly in languages with complex writing systems.

• →Hybrid e-assessment combines paper-based exams with AI-assisted grading to maintain pedagogical quality while improving scalability.
• →Vision-capable language models with two-pass validation reduce handwriting misrecognition errors in high-stakes examination environments.
• →The approach addresses organizational, technical, and legal constraints that limit adoption of fully digital assessment systems.
• →Real-world validation of handwriting recognition accuracy under examination conditions remains the critical success factor.
• →Educational institutions could accelerate AI-assisted grading adoption through this intermediate model rather than full digital transitions.