Penilaian autentik dan integriti akademik dalam era AI

Summary
How do educators assess learning honestly when artificial intelligence can draft essays and solve complex problems in seconds? At Bett Asia 2026, education technology leaders Sonia Wadhwa and Dewi Yulianti addressed this defining challenge. Moving past outdated debates about blocking AI behind network firewalls, the presenters demonstrated how K-12 institutions can combine human-centred classroom pedagogy with clear policy guardrails.

The session established that authentic assessment in the AI era requires shifting focus from the final written product to the student's underlying thinking process. By integrating AI into coursework through guided mentoring, rubric self-assessment, real-world data collection, and age-appropriate literacy frameworks, schools can elevate critical thinking while upholding academic integrity.

1. Middle School Biology & Computer Science: Food Science Machine Learning
Instead of writing a standard report on nutrition, middle school students collaborated with computer science staff to train an AI classification model. To achieve over 95 per cent predictive accuracy in identifying carbohydrates, fats, and proteins, students sourced and curated extensive real-world food samples. Deeper learning occurred through dataset preparation and understanding model training accuracy.

2. Grade 11 English: Rubric Analysis and Independent Writing
For a unit on newspaper advertisements, students were permitted to use generative AI tools (such as ChatGPT, Gemini, or Meta AI) for homework tasks. Upon returning to class, students evaluated their AI-assisted outputs against official examination rubrics to identify grading criteria and loss of marks. Students then drafted a fresh advertisement in class under teacher supervision, demonstrating authentic mastery of the rubric.

3. Grade 6 Geography & Biology: Geotagged Fieldwork Verification
Students explored a local park to capture geotagged photographs while documenting microclimate factors, including soil type, sunlight, and water access. In class, students uploaded their images to AI tools to identify plant species and compare AI-generated growth conditions against their real-time observational notes. This exercise trained students to spot AI hallucinations and evaluate environmental anomalies critically.

4. Middle School Mathematics: Guided AI Mentorship
When solving linear equations, students used AI exclusively in a 'guided learning mode' rather than requesting direct answers. Students documented where they encountered learning gaps and how the AI co-pilot helped them overcome sticking points. This approach reduced teacher doubt-clearing queries from 28 students down to three or four, while building student self-regulation and problem-ownership.

5. AI Literacy and Cross-Phase Collaboration
Middle school students developed AI-generated mathematical word problems and real-world consumer scenarios—such as evaluating fruit vendor delivery charges or comparing cab fare structures. Middle schoolers were assessed on prompt engineering and real-world application, while primary school pupils solved the resulting numeracy challenges. Middle schoolers also created weekly story podcasts for younger learners by cloning teacher voice models, defending their prompt iterations through oral viva examinations.

6. Grade 4 Art & Geometry: Visual Inspiration for Physical Design
To teach shape, symmetry, and pattern, an art teacher used generative AI to produce visual concepts projected in the classroom. Rather than copying the images, pupils used these visuals as creative springboards to craft physical floor designs, architectural layouts, and monument patterns using traditional art mediums.

7. Early Years Storytelling: Multimodal Sequencing Worksheets
Teachers used generative AI to simplify narrative texts into structured storyboards and jumbled activity sheets. In class, teachers read the stories aloud while parents accessed QR-coded audio files at home. Young learners re-sequenced the story scenes, reinforcing seven foundational cognitive and literacy skills simultaneously.

8. Assessment Creation for High-Stakes Examinations
To align classroom testing with evolving global standards (such as PISA, SAT, CBSE, ICSE, and Cambridge), teachers uploaded chapter content to AI platforms to construct higher-order assertion-reasoning and critical-thinking questions aligned with specific learning outcomes.

Institutional Guardrails: Policy, Data Safety, and Frameworks
Dewi Yulianti addressed the governance required to support innovative pedagogy. Citing a Ministry of Education survey from Brunei Darussalam, Yulianti highlighted a critical gap: while 91.4 per cent of surveyed educators used generative AI for daily tasks and 86.3 per cent sought further exploration, 81 per cent expressed uncertainty regarding data privacy and safety. This indicates that AI adoption routinely outpaces formal AI literacy.

To bridge this gap, institutions must align with national directives—such as Brunei's Generative AI Guidance in Education and Personal Data Protection Order (PDPO)—as well as international benchmarks from UNESCO and ASEAN.
 

Tier	Guidance for Students and Teachers
Red Tier	No AI usage permitted. Assessment relies purely on traditional, unassisted student output.
Amber Tier	AI usage permitted with active teacher intervention, guidance, and scaffolded support.
Green Tier	AI usage fully permitted, contingent on student disclosure, explicit tool transparency, and honest attribution.

Applied Policy Examples

Early Years Reading Coach: Year 1 pupils use AI reading tools for immediate pronunciation feedback (Green Tier). When pupils encounter difficulties, teachers step in directly without AI assistance (Amber Tier), before evaluating literacy progress through traditional unassisted reading (Red Tier).

Year 5 & 6 STEAM Robotics: Students program micro-controller robotics equipped with AI vision lenses to categorise waste items. AI is permitted during initial brainstorming and code exploration, but students must present, defend, and manually refine their physical builds.

Strategic Takeaways for Education Leaders

Prioritise Process Over Product: Evaluate the student's cognitive journey, prompt iterations, and reflection rather than solely grading final submissions.

Pair Pedagogy with Governance: Effective AI integration requires combining human-centred classroom activities with clear institutional frameworks.

Implement Scaffolded Transparency: Adopt clear operational boundaries, such as the Traffic Light framework, to establish explicit rules for AI usage and disclosure.

Protect Institutional and Student Data: Ensure all AI deployment complies with data protection regulations and personal data privacy orders.

Upskill Staff and Students Simultaneously: Professional development for educators must run parallel to student AI literacy programmes to ensure consistent standards across the curriculum.

Audio and written summaries powered by Gemini Notebook, a free generative AI tool grounded in your provided sources. Try Gemini Notebook for free. You can listen to the audio overview of the session here.