Unlocking Deep Learning in Indonesia’s PISA 2025 Era: Moving Beyond Nominal Score Gains to Build Cognitive Agency

Ilustrasi by AI

In a comprehensive analysis published in September 2026, researcher Loso Judijanto from IPOSS Jakarta evaluated Indonesia’s latest Programme for International Student Assessment (PISA) 2025 performance. The evaluation synthesizes PISA 2025 evidence alongside peer-reviewed literature to examine how expanding educational access, foundational literacy deficits, and artificial intelligence interact in modern Indonesian classrooms. Despite nominal increases in reading and science scores compared to 2022, the study underscores that overall student performance remains critically low, with a vast majority of 15-year-olds failing to reach basic proficiency thresholds. The findings matter because relying solely on score increases hides severe learning deficits that require structural reform as digital technologies and generative AI transform everyday study habits.

Background and Context

International assessments like PISA serve as vital diagnostic tools to measure how effectively national education systems prepare students for real-world reasoning. Indonesia’s PISA population coverage expanded significantly, reaching approximately 90% of 15-year-olds in 2025 compared to just 46% in 2003. While this structural expansion represents major progress in educational inclusion, it coincides with widespread challenges in foundational competencies. Today’s students navigate a learning landscape saturated with digital devices and generative AI tools. Understanding how these technologies impact learning has become a critical priority for policymakers, educators, and social leaders attempting to bridge the gap between school attendance and true intellectual capability.

Research Methodology

The study utilized a qualitative literature review incorporating thematic synthesis. The researcher integrated newly released PISA 2025 country data with recent peer-reviewed academic publications from 2020 through 2026. Data triangulation focused on official OECD report figures—representing 14,168 Indonesian students across 419 schools—and evaluated five core domains: assessment performance, foundational learning, teaching quality and equity, digital distraction, and generative AI utilization. Through iterative coding, the review constructed an "Access-Learning-Cognition-Digital Ecology Framework" to examine how external inputs translate into cognitive outcomes.

Key Findings

The synthesis highlights several critical insights into Indonesia's educational landscape:

  • Nominal vs. Statistically Significant Recovery: Science scores reached 389 (up from 383 in 2022) and reading scores reached 365 (up from 359), while mathematics stood at 364 (down from 366). However, the OECD concluded that performance in all three domains remained statistically unchanged from 2022 and below 2015 levels.
  • Severe Foundational Proficiency Gap: Only 17% of Indonesian 15-year-olds attained baseline Level 2 proficiency in mathematics, compared to an OECD average of 65%. Baseline proficiency was reached by 27% of students in reading (versus 69% OECD average) and 37% in science (versus 74% OECD average).
  • The Teacher-Support Paradox: While over 80% of students reported high emotional and instructional support from teachers, this high perceived support did not automatically yield high cognitive activation or independent reasoning.
  • Digital Distraction vs. AI Adoption: 23% of students reported frequent peer distraction from digital devices during science lessons. Simultaneously, 53% used AI chatbots weekly for learning, 39% used AI to summarize texts, 32% for preliminary research, and 30% for drafting written assignments.

Real-World Impact and Policy Implications

The research emphasizes that Indonesia must shift focus from curriculum coverage and raw international rankings toward building "cognitive agency"—the ability to independently reason, evaluate evidence, and solve complex problems. To prevent generative AI from replacing essential learning processes, schools must implement "human-first" task sequencing where students complete independent thinking before utilizing digital assistance. Policymakers and educators can leverage these insights to establish diagnostic remediation in foundational mathematics, enforce structured digital-device policies in classrooms, and implement AI literacy programs that teach critical verification rather than passive consumption.

Expert Commentary

Highlighting the core challenge facing modern education, Loso Judijanto writes:

"In an age when machines can generate answers in seconds, the comparative advantage of education is no longer giving students access to answers. It is developing people capable of deciding which questions matter, evaluating which answers deserve trust, reasoning when no ready-made answer is sufficient, and learning independently throughout their lives."

Author Profile

Loso Judijanto, M.Sc., is a senior researcher affiliated with IPOSS Jakarta, Indonesia. His field of expertise encompasses educational policy, statistical modeling, public administration, and technology integration in learning environments.

Source Information

 


Posting Komentar

0 Komentar