03 · Analyse risk · Education

Assess harm, fairness and trustworthiness

Focus on misclassification, language/accessibility disadvantage, inappropriate disciplinary use and over-reliance by staff.

EDUCATION WORKING CONTEXT

Assess educational fairness and safeguarding

The system supports learner engagement, assessment or safeguarding. Age, accessibility, educational fairness and appropriate staff judgement matter.

Evidence bar

Group performance results
Accessibility test
Safeguarding and escalation record

REVIEW CHECKPOINTS

AI Harm Assessment

Test performance by learner group and settingRecord safeguarding and accessibility unknownsDefine launch blockers for grading or disciplinary misuseDescribe the mechanism of harmScore probability and severity separatelyTest whether a harm should block launch

INTERACTIVE REVIEW MEETING

Pressure-test your reasoning

Business sponsorBenefit pressure

Accuracy is high overall. Why should a subgroup disparity delay launch?

Write at least 40 characters to reveal the review criteria.

OperationsOversight claim

A human clicks approve, so is the risk already controlled?

Write at least 40 characters to reveal the review criteria.

AI Harm Assessment

Complete the fields against the education operating context. Your work saves automatically in this browser.

Primary harm level