Page 270 - CyberSecurityEssentialsEnglish

adopted a three-layer answer: a summary, the steps, and an official link. This is an example
of support improvement requiring service design, not merely a language model.

⚠ Warning: the limits of an automated assistant
Do not use an automated assistant to diagnose mental health, determine eligibility for a

sensitive service, issue a disciplinary decision, or close off the route to human contact.
The limits of escalation must be explicit and tested.

6.6.2 Learning Analytics: What Do We Collect, and Why?

Learning analytics can help reveal general patterns: questions students repeat, a step where
they stop, or a type of feedback they do not understand. But its educational value depends
on controlling the purpose and the data. Do not gather everything that can be gathered;
gather only what serves the improvement of teaching or support, with the least possible

personal identification.

Before using analytics, ask: are we analyzing to improve the activity, or to judge the student?
Do learners know what is being collected? Can the result be interpreted pedagogically? And

what is the risk of labelling if a student is described early as “weak” or “at risk”? Good analysis
opens a door to early human support; it does not issue a final verdict on a learner’s future.

In sensitive cases, a person must appear in the loop. If the system shows a pattern of
absence, anxiety, or repeated difficulty, it should not send a diagnosis or an accusation. It

should lead to safe contact: a supportive message, the option of a meeting, or a referral to
a qualified body according to institutional policy.

FIGURE 6-7: WHEN SUPPORT MUST MOVE TO A PERSON

What to notice: The boundary is not the difficulty of the question. It is whether the situation
needs a qualified human being.

Source: Prepared by the author.

269