Meenal: AI eval glitch hits 40k scripts; delay or adjust?

Meenal is Controller of Examinations for a state higher-secondary board responsible for the results of over eight lakh students. This year, the board introduced AI-assisted evaluation for the first time, designed to cross-check human evaluators for consistency. Three days before results are due, Meenal's technical team reports a significant anomaly: in one subject, evaluated by a partner agency in a single zone, nearly 40,000 scripts show a marking pattern statistically inconsistent with the rest of the state, skewed unusually low. A sample recheck confirms a batch of evaluators in that zone applied a stricter, non-standard scheme, likely due to a training gap. Manually re-evaluating all 40,000 scripts before the announced date isn't feasible — it would delay results for the entire state by three to four weeks, disrupting college admissions for everyone, not just the affected zone. The board's chairperson, under pressure from the state government to avoid any delay, suggests applying a uniform statistical grace-mark adjustment to the affected scripts instead, and releasing results on schedule.

(a) What are the ethical issues involved in this case?
(b) Critically examine the options available to Meenal.
(c) Which of the above would be most appropriate and why?

Solution

The case reflects the tension between administrative timelines and individual fairness when a systemic evaluation error surfaces close to a deadline.

Meenal must weigh the disruption of a delay against a statistical fix that may not fully correct the harm to specific students.

The core issue is whether efficiency can ethically justify a less-than-accurate remedy for an error that was the system's fault, not the students'.

Stakeholders

• Meenal, Controller of Examinations
• The approximately 40,000 affected students and their families
• The remaining students across the state
• The evaluator agency and its staff
• State government and higher education institutions
• Board chairperson

(a) Ethical Issues Involved

• Accountability for systemic error: the students did nothing wrong; the fault lies entirely in evaluator training.
• Precision vs expedience: a flat grace-mark adjustment cannot accurately correct for how differently each script was affected.
• Equity across zones: affected students face real disadvantage in competitive admissions if not fully corrected.
• Institutional transparency: how the error is communicated shapes whether public trust survives it.
• Time pressure vs thoroughness: political pressure for an on-schedule announcement should not override accuracy owed to individuals.

(b) Options Available to Meenal

Option 1: Apply a uniform grace-mark adjustment and release on schedule

Merits

• Preserves the timeline for all students and downstream admissions.

Demerits

• Cannot accurately correct for how differently each script was affected; risks under- or over-compensating.

Option 2: Manually re-evaluate all 40,000 scripts before releasing any results

Merits

• Ensures true accuracy for every affected student.

Demerits

• Delays results statewide by three to four weeks, harming even unaffected students' admissions.

Option 3: Release unaffected zones on schedule; re-evaluate only the flagged zone on an expedited but genuine basis, with a short, clearly communicated delay for that zone alone

Merits

• Limits disruption to only those actually affected.
• Preserves accuracy rather than trading it away for speed.

Demerits

• Requires rapid mobilisation of additional evaluators and careful communication to avoid stigmatising the affected zone.

(c) Most Appropriate Option

Meenal should adopt Option 3.

She should:

• Release results on schedule for all unaffected zones.
• Mobilise additional trained evaluators immediately for genuine, expedited re-evaluation of the flagged zone.
• Communicate transparently with affected students and parents about the reason for the delay.
• Coordinate with higher education institutions for short, targeted admission-deadline flexibility for those specific students only.
• Commission a review of the evaluator training gap to prevent recurrence.

Conclusion

An error that originates entirely within the system should not be resolved by a shortcut that leaves its cost sitting with the students.

Genuine fairness sometimes takes longer than a press-ready timeline allows — protecting it for the group that needs it, without penalising everyone else, is the ethical administrator's job.

Click here to watch video