evidoria

← Back to browse

Good practice Imported

CBSE On-Screen AI-Assisted Evaluation — digitised, AI-monitored marking for India's board exams

India · New Delhi · See the India profile · See the New Delhi profile

Evidence: Descriptive / self-reported Top 59% 47/100 · Ask Evidence Copilot about this practice

India's CBSE is scaling On-Screen Marking with AI-based quality control for Class XII board exams from 2025-26: scripts are digitised, anonymised by barcode, and AI flags examiners marking unusually fast or inconsistently, while humans remain the final graders.

3-5 marks
Average score divergence between AI-supported check and expert human graders (pilot, 250+ answer sheets)
250+ answer sheets
Answer sheets covered in pilot (2024-25 pilot)
>8 hours/week
Teacher time saved per week on manual tallying/logistics (2024-25 pilot)

Details

Maturity
Scaling
Promoter
Central Board of Secondary Education (CBSE)
Period
2025–present
Keywords
examination boards, education administration, assessment, AI quality-control

Context

India's Central Board of Secondary Education (CBSE) is rolling out On-Screen Marking (OSM) with an AI quality-control layer for Class XII board exams, moving from a 2024-25 pilot to full rollout in 2025-26 across roughly 29,000 affiliated schools. Digitised, anonymised answer scripts are processed at a digital-evaluation and AI centre in Dwarka, New Delhi.

Objectives

Standardise marking consistency and improve grading efficiency across India's largest school examination board through digitised, anonymised scripts monitored by AI.

Activities

Answer scripts are scanned at regional centres and assigned unique barcodes, then displayed on-screen to human examiners with student identities anonymised. An AI layer auto-tallies marks and monitors examiner behaviour, flagging scripts marked unusually quickly or with inconsistent patterns for supervisory human review; human examiners remain the final assessors, and the system does not autonomously grade descriptive answers.

Results

In a pilot covering more than 250 answer sheets, AI-supported scoring checks diverged from expert human graders by only 3-5 marks on average, and teachers reported saving over 8 hours per week previously spent on manual tallying and script logistics.

Conclusions

Evidence to date focuses on process quality (marking consistency, examiner time saved) rather than student learning outcomes, and no independent, peer-reviewed evaluation of the system has been conducted.

Implementation
Indicative cost
High (€500k–€5M) — Not available in source documentation.
Time to results
Medium (1–3 years) — Piloted 2024-25; full rollout for Class XII exams beginning 2025-26.
Staffing & skills
CBSE digital-evaluation and AI centre staff (Dwarka, New Delhi), Regional scanning-centre staff, Human examiners as final assessors, Supervisory reviewers for AI-flagged scripts

Conditions for success

  • Reliable barcode-based anonymisation of scripts
  • Human supervisory review of all AI-flagged scripts
  • Examiner training on the new digitised workflow

Common failure modes

  • No autonomous grading of descriptive answers -- human examiners remain necessary
  • No independent evaluation to verify pilot accuracy figures
  • Examiner training described as still early-stage

Where it fits

Governance type
national examination board
Scale
national (~29,000 affiliated schools)
Income level
lower-middle income

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.

Similar practices you may find useful