evidoria

← Back to browse

Good practice Imported

AKM Online — AI-Enhanced Computer-Based Assessment for Literacy and Numeracy in Indonesian Secondary Schools

Indonesia · Kuningan · See the Indonesia profile

Evidence: Observational / pre–post Top 81% 33/100 · Ask Evidence Copilot about this practice

Universitas Kuningan researchers built and validated AKM Online, an AI-enhanced version of Indonesia's national competency assessment, testing 552 students across three schools with 94.35% AI-scoring accuracy on 17,049 responses.

552 students
Student participants
17,049 responses
Individual responses collected
94.35 %
AI scoring accuracy vs expert human grading
81 %
Expert-panel rating: educational-assessment design
82 %
Expert-panel rating: informatics performance
95.8-100 %
Students needing numeracy intervention
53.9-78.46 %
Students needing literacy intervention
0 %
Mathematical reasoning proficiency
AKM Online — AI-Enhanced Computer-Based Assessment for Literacy and Numeracy in Indonesian Secondary Schools

Details

Maturity
Pilot
Promoter
Universitas Kuningan
Period
2025-2026
Keywords
assessment, literacy, numeracy, secondary education, edtech research

Context

Indonesia's national Minimum Competency Assessment (AKM) is the government's standard tool for measuring foundational literacy and numeracy skills, but paper- and generic computer-based versions struggle to deliver reliable, real-time diagnostics at scale. Researchers at Universitas Kuningan developed and validated AKM Online, an AI-enhanced Computer-Based Assessment system.

Activities

AKM Online integrates Fisher-Yates Shuffle and Regular Expression algorithms for secure test randomisation, automated response validation, and real-time diagnostic feedback. It was trialled across three schools chosen for contrasting socioeconomic contexts.

Results

The trial generated 552 student participants and 17,049 individual responses. The AI scoring engine reached 94.35% accuracy against expert human grading, and independent expert panels rated system quality at 81% for educational-assessment design and 82% for informatics performance. Diagnostics found 95.8-100% of students needed intervention in numeracy, 53.9-78.46% needed intervention in literacy depending on context, and mathematical reasoning proficiency measured 0% across the sample.

Conclusions

The study is a single peer-reviewed validation by the system's own development team, not yet an independent replication or a national rollout, and it does not address student data protection or governance safeguards for the assessment data it collects.

Implementation
Indicative cost
Low (< €50k) — Not disclosed; a university research/software development effort rather than a funded national deployment.
Time to results
Short (< 1 year) — Study period 2025-2026; single validation trial across three schools, not yet a national rollout.
Staffing & skills
Universitas Kuningan research team

Conditions for success

  • Algorithmic test randomisation and automated validation enabling real-time diagnostics at scale
  • Testing across three schools with contrasting socioeconomic contexts to check cross-context validity

Common failure modes

  • Single peer-reviewed validation by the system's own development team, not an independent replication
  • No discussion of student data protection or governance safeguards

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful