- Possible justification
- Assessment can shape a student’s opportunities and relationship with an institution. A defensible human role should involve understanding the work, giving reasons and being able to change an assessment. The case rests on accountable judgment and an effective response to error, rather than a presumption that a person always performs better than a tool.
- Strongest objection considered
- A teacher can be biased, rushed or mistaken. Requiring a person to sign off may add delay without improving the decision, or conceal routine acceptance of a tool’s answer. Compare substantive judgment with feasible assistance and challenge routes; do not treat human presence as proof of fairness.
- Who might benefit or be excluded
- Consider students, families, teachers and institutions separately. A student needs an understandable assessment and a workable route to question it; a teacher needs room to disagree without unreasonable workload. Less-resourced schools and pupils who struggle to complain may fare differently. These are questions to test, not observed effects.
- Resources and costs to investigate
- Estimate the time needed to inspect work, check a tool’s contribution, explain a result and respond to disagreement. Examine training, accessible materials, procurement and who bears extra work. Proposed time savings should be assessed across the whole workflow, including corrections and challenges; no saving or funded entitlement is established here.
- An alternative to compare
- Compare teacher-led marking without AI with teacher-led marking using limited AI assistance. Identify each package’s time, training, funding and support needs. Either could include checks by another assessor or a second assessment that students can access. Specify who can reject a tool’s recommendation and what happens when a teacher cannot evaluate it. Compare learning, fairness, access, workload and cost.
- What to check before using this elsewhere
- A proposal should identify the education authority or institution able to adopt it, its legal basis, the accountable assessor, permitted assistance, resources, reason-giving and a usable challenge route. Distinguish advisory expectations from enforceable duties. Test student and teacher interests separately before borrowing a design across jurisdictions.
- Evidence that could change the assessment
- Change the design if sign-off merely conceals automation, extra checks worsen access without useful gains, or a feasible alternative improves learning, fairness and challenge at sustainable cost. Conversely, evidence that independent judgment corrects consequential errors would strengthen a resourced role. State how workload or unequal outcomes would trigger revision.
These are HBC’s judgments and questions for further research. They do not show actual benefits or costs, establish workers’ preferences or authorise us to speak for affected people.