Academic Jobs - Home of Higher Ed Logo

AI Grading with Human Review Transforms Singapore Universities

240views
Submit News
a tall building with a sign on top of it
Photo by Chunjiang on Unsplash

Singapore's leading universities are at the forefront of integrating artificial intelligence into academic assessments, marking a significant shift in how student work is evaluated. Lecturers at the National University of Singapore (NUS), Nanyang Technological University (NTU), Singapore University of Technology and Design (SUTD), and Singapore Institute of Technology (SIT) are employing AI tools to grade assignments and exams that contribute to final grades. This hybrid approach combines machine efficiency with human judgment, ensuring fairness while addressing the growing demands of large class sizes and complex workloads.

The adoption reflects Singapore's broader push towards AI literacy and innovation in education, aligning with national strategies to prepare students for an AI-driven future. By streamlining grading processes, these institutions free up lecturers to focus on mentoring and curriculum development, ultimately enhancing the learning experience.

🚀 The Emergence of AI-Assisted Grading in Singapore

AI grading began gaining traction in Singapore universities around mid-2024, driven by the need for scalable assessment solutions. NTU led the way in August 2024 with pilots in physics and mathematics modules, followed by SUTD in April 2025 for computer science and design tests, SIT's October 2025 trial in food technology, and NUS's July 2025 rollout for English proficiency tests. This timeline coincides with rapid advancements in AI capabilities and post-pandemic pressures on educators to handle increased workloads efficiently.

What sets this apart is the emphasis on human-AI collaboration. Unlike fully automated systems elsewhere, Singapore's model mandates lecturer review of all AI outputs, preserving academic integrity. Students are transparently informed when AI is involved, fostering trust and encouraging critical discussions about technology in education.

Collage of NUS, NTU, SUTD, SIT campuses with AI grading interfaces

How AI Tools Function in Practice

Different tools cater to specific assessment types, showcasing tailored innovation. Gradescope, used by NTU and SUTD, scans handwritten exam scripts, groups similar answers using pattern recognition, and applies rubrics for batch grading. This is ideal for technical subjects where step-by-step workings are key, reducing manual sorting time dramatically.

At SIT, the internally developed AI-Orate platform acts as an interactive chatbot. It quizzes students on topics like food manufacturing processes—asking about code for industrial machines or microbial reaction rates—and generates transcripts with follow-up probes on reasoning. Prof Wong Shin Yee from SIT notes that it transforms assessments into dynamic conversations, allowing students to clarify incomplete initial responses for a second chance at marks.

NUS employs a validated, unnamed tool for argumentative essays in post-admission English tests, evaluating content, organization, and language. The system runs grading twice for reliability, with human auditors checking borderline cases. Associate Provost Melvin Yap highlights how this hybrid method surpasses pure human grading by mitigating fatigue-induced inconsistencies.

These tools process submissions via optical character recognition (OCR) for handwriting, natural language processing (NLP) for essays, and machine learning algorithms trained on expert-graded samples to mimic human standards.

Implementation Timeline and Scope at Each University

NUS: Focused initially on English competency for students without standard qualifications, grading over 3,000 scripts annually with over 95% assured accuracy, as mentioned by Minister Desmond Lee. Expansion to computing pilots evaluates code submissions.

NTU: Physics and math midterms/finals since 2024. Deputy President Christian Wolfrum emphasizes, “This improves consistency and efficiency while keeping decisions with instructors.”

SUTD: Short-answer and explanation questions in CS/design since 2025. Associate Provost Ashraf Kassim views AI as a “supporting partner.”

SIT: Trial with 50 food tech undergrads halved assessment time from a week to two days, enabling personalized quizzing in large groups.

Scope is limited to approved, validated tools, with department/deanery sign-off required at NUS.

Safeguards Ensuring Quality and Fairness

Mandatory human oversight is non-negotiable. Lecturers review AI suggestions, adjust for context, and finalize marks. At NUS, dual AI runs and audits catch discrepancies. Students can appeal grades, prompting re-evaluation.

Transparency rules require notification for summative assessments. Policies align with university-wide GenAI guidelines: NTU's two-lane model (assured vs. collaborative), NUS interim policy promoting responsible use without bans.

This framework addresses biases in AI training data through rigorous validation against human benchmarks.

Key Benefits for Educators and Learners

  • Time Efficiency: SIT's tool cuts weeks-long processes to days; Gradescope enables batch handling for hundreds of scripts.
  • Consistency: AI applies rubrics uniformly, free from human variability or fatigue.
  • Faster Feedback: Instant preliminary insights allow quicker iterations, vital for large cohorts.
  • Scalability: Supports personalized, adaptive questioning impossible manually.
  • Student Gains: Deeper engagement via interactive probes; second chances build resilience.

Lecturers redirect saved time to high-value activities like research or one-on-one guidance.

SUTD's Gradescope implementation details illustrate these gains in practice.

Challenges and Ongoing Concerns

Despite promise, hurdles remain. SMU and SUSS abstain, citing reliability gaps for nuanced work—Prof Venky Shankararaman stresses human judgment for transparency. Students like NUS's Leslie De Souza question AI's grasp of subtleties: “How would AI tell me what to improve on?”

Potential biases, misinterpretation of handwriting, or over-reliance risk eroding critical thinking. Ethical issues around data privacy and tool validation persist. Universities counter with pilots, audits, and faculty training.

Voices from the Ground: Student and Lecturer Perspectives

Students show mixed reactions. NTU's Ryanna Lee welcomes controls: “It’s fair if watched.” SIT's Lief Chng shifted from skepticism to appreciation for follow-ups. Lecturers praise workload relief but caution expansion needs more validation.

Broader discourse, like NTU's 2025 AI misuse penalties, underscores balanced integration.

SIT AI-Orate interactive grading chatbot demonstration

Implications for Singapore's Higher Education Landscape

This positions Singapore as an AI education pioneer, aligning with national goals like the AI Strategy 2.0. It models ethical adoption: hybrid systems enhance equity in diverse classrooms.

Stakeholders gain: lecturers reclaim time, students receive timely insights, institutions scale amid enrollment growth.

Straits Times in-depth coverage on early pilots reveals maturing ecosystem.

Future Directions and Global Lessons

Plans include wider rollout post-validation, AI literacy mandates, and committee oversight per MOE. By 2027, expect multimodal tools for essays/codes. Singapore's cautious optimism offers a blueprint: prioritize humans, transparency, continuous refinement.

For aspiring lecturers, this evolution demands AI proficiency alongside pedagogy—skills blending tech and teaching define tomorrow's educators.

a young man walking past a building with a sign

Photo by Chunjiang on Unsplash

Browse by Faculty

Browse by Subject

Frequently Asked Questions

🎓What universities in Singapore use AI for grading?

NUS, NTU, SUTD, and SIT lecturers use AI tools for assessments contributing to final grades, with SMU and SUSS holding off pending further evaluation.

🤖Which AI tools are employed?

Gradescope for math/physics at NTU/SUTD, AI-Orate chatbot at SIT, and a validated essay grader at NUS. All require human review.

Is human review mandatory?

Yes, all four universities mandate lecturer review of AI outputs, with NUS using dual AI runs and audits for borderline cases.

📢How are students informed about AI use?

Students must be notified when AI grades work affecting final results, promoting transparency and trust.

What benefits does AI grading offer?

Time savings (e.g., SIT: week to 2 days), consistent rubrics, faster feedback, scalability for large classes, and adaptive questioning.

⚠️What are the main challenges?

Nuance detection, potential biases, student trust issues, and ensuring alignment with learning goals. Validation and training mitigate these.

📅When did each university start AI grading?

NTU: Aug 2024; SUTD: Apr 2025; SIT: Oct 2025 trial; NUS: Jul 2025 for English tests.

📊How accurate is NUS's AI English grader?

Over 95% assured accuracy after validation, grading 3,000+ students yearly with human audits.

⚖️Can students appeal AI-influenced grades?

Yes, standard review processes apply, allowing re-evaluation by humans.

🔮What's next for AI grading in Singapore?

Expansion post-validation, AI literacy integration, national committee oversight, aiming for ethical, widespread use by 2027.

🇸🇬How does this fit Singapore's AI strategy?

Aligns with AI Strategy 2.0, building talent pipeline through innovative education tools.