Page 272 - CyberSecurityEssentialsEnglish

Staged assessment helps make the development of the work visible. It begins with a question
or a plan, then a source or data, then a draft, then feedback, then a final version and a
reflection. This does not mean increasing the marking burden at every stage; short

checkpoints or samples can be chosen. The benefit is that the process becomes part of the
evidence, and it is hard to replace with a ready-made product at the last minute.

Authentic assessment asks for a performance close to how knowledge is used outside the

classroom: analyzing a local case, briefing an audience, running an experiment, designing a
solution under real constraints, or interpreting a decision. AI may be permitted as it would
be used in reality, but the learner is assessed on the quality of judgment, verification, and

justification, not on the final appearance alone.

Disclosure of AI use should be simple and proportionate. A learner might state the tool, the
purpose, the parts it helped with, and what they changed or verified. In a larger task, a short
log of usage decisions or examples of the interaction can be attached. Long transcripts

containing data or noise are not always needed; what is required is a trace that helps in
understanding the contribution.

Transparency must not become an unintended penalty. If a learner who declares a permitted
use is treated with more suspicion than someone who hides it, the culture of disclosure will

weaken. The standards should focus on the quality of learning and adherence to the rules,
and permitted use should be a natural part of the design.

Tools that detect automatically generated text do not offer enough certainty to serve as sole

evidence. Regulatory guidance notes that relying on them exclusively is unwise because of
errors, false positives, and the possibility of circumvention [11]. Where there is doubt, a fair
process based on several kinds of evidence must be followed, such as a conversation about

the work, the history of drafts, the consistency of performance, and institutional rules, rather
than an automated score alone.

Fairness matters too, because access differs. Some learners may have paid subscriptions,
faster devices, or greater command of English. If a tool is required, reasonable access or an

equivalent alternative must be provided, and the skill itself must be taught. And if it is
permitted, financial capacity must not become an unacknowledged hidden advantage.

Penalties and investigations must be designed with care. Suspicion is not proof, and a
difference in style may result from language support or from many other circumstances.

Explanation and appeal must be available, investigators must be trained, and early learning
support must be separated from deliberate misconduct. The aim is to protect both the value

of the qualification and the learner’s rights.

271