AI and Grade 11 academic performance: evidence is more than an assisted answer

Published · Updated

Save

AI can help a student complete a task without necessarily improving what that student can do independently. A useful review of AI and Grade 11 performance separates immediate assisted work, later unaided learning, engagement and academic integrity.

One relevant experiment—and its scope

Bastani and colleagues’ 2025 PNAS paper reports a field experiment in high-school mathematics in Turkey. It compared a general GPT-4-based interface, a tutor with learning-oriented guardrails and a control condition. The work distinguishes performance with assistance from a subsequent unaided assessment. It found that unguarded assistance could harm later independent performance, while tutor design mattered.

This is evidence about particular tools, teaching conditions, students and assessments. It does not establish that all AI tools harm learning or that Ithy improves grades. Nor does it establish an effect for every Grade 11 subject or Philippine school.

What changed in this review

The earlier version supplied generic local and international study descriptions and percentages without adequately verified study identities and support. The claimed 75% awareness and 55% ethical-concern figures are not established here. Do not cite them as results from a named local school.

Build your literature matrix around the learning outcome

Record for each studyWhy it matters
Students, subject and settingA university writing study does not automatically answer a Grade 11 maths question.
Exact tool and interventionAnswer generation, hints, retrieval and adaptive tutoring are different uses.
Comparison and assignmentRandom assignment and self-selected use support different inferences.
Assisted and unaided assessmentsTask completion with a tool is distinct from retained understanding.
Follow-up, uncertainty and limitationsA short-term result may not persist or generalise.

A proposed Grade 11 question

For example: “After teacher-approved practice with hint-based AI assistance, how do students perform on a later unaided explanation task compared with the approved comparison activity?” This is a proposal, not a completed study. Define the task, marking rubric, assessment interval and permitted assistance before collecting data.

Any classroom research requires the school’s applicable review, permission and consent arrangements. Preserve ordinary learning support, protect minors’ data and do not upload identifiable student records or confidential work to an AI service. A self-report that AI “helped” is useful perception data, but it is not a measured grade improvement.

Use AI in a way you can inspect

A student can ask for a hint, attempt the next step unaided, explain the reasoning, and check a source. For a literature review, open each cited paper and verify its population, method and actual conclusion. Follow the course’s disclosure and assessment rules. No new school survey, grades analysis or measured product benefit is claimed by this article.

Sources and further checks

  1. Bastani et al. (2025): Generative AI without guardrails can harm learning
  2. Accessible paper and study details in PubMed Central
  3. Practical citation verification checklist
Discuss this answer