The Proctoring Flags
A university conducts an online proctored examination for 600 postgraduate students. The proctoring software flags 14 candidates for 'tab switching' and 'multiple faces.' Of these, three have documented electricity outages in their area, two use screen-reader software that the system misreads, and one had a sibling enter the room to deliver medicine. The examination committee must decide within 48 hours whether to cancel the results of all 14, hold a re-test, or clear them. The software vendor says its algorithm has 92% accuracy in controlled trials. Student groups allege arbitrary action; faculty members are divided. The university's ordinance says 'use of unfair means shall be established by evidence,' but the controller of examinations treats the software report as presumptive proof. The committee also has logs showing that one flagged candidate's webcam disconnected for 90 seconds before reconnecting.
1. Is the software report sufficient evidence to cancel all 14 results? Apply the ordinance's standard and the base-rate issue.
Read the worked answer
No. The ordinance requires evidence that unfair means were used; a software flag is a screening signal, not proof. The vendor's 92% accuracy in controlled trials means an 8% error rate in that setting, and in a real examination the base rate of cheating is low, so a large share of flags can be false positives. The documented power outages, screen-reader misreads and the sibling's entry are concrete alternative explanations for several flags. Treating the report as presumptive proof reverses the burden of proof. The 90-second webcam disconnection is also ambiguous: it is consistent with a technical fault as well as with misconduct. Cancelling all 14 results would impose a collective sanction without individual evidence and would likely be vulnerable to challenge. The committee should separate flags with independent explanations from those without, and decide case by case.
2. Design a fair process the committee can complete within 48 hours.
Read the worked answer
The committee should: (1) triage the 14 into those with documentary explanations (outage proof, accessibility software, medical entry) and those without; (2) give each candidate written notice of the specific flag and an opportunity to respond with evidence, even if the window is short; (3) review the proctoring logs, video and technical metadata rather than the summary report alone; (4) for ambiguous cases, offer a targeted re-test under controlled conditions or an oral/viva assessment rather than a blanket re-test; (5) record reasons for each decision. A blanket re-test penalises innocent candidates and may not be feasible for all; a targeted process is fairer. The committee should also check whether the software has been validated for screen-reader users. The controller should not pre-determine the outcome.
3. What should the university learn about the role of technology in assessment?
Read the worked answer
The university should treat proctoring software as a risk-flagging tool, not a judge. It should audit the tool for accessibility bias, publish an appeals process, train invigilators to interpret flags, and validate vendor accuracy claims on its own student population. The ordinance's evidence standard should be reflected in a written protocol that distinguishes a flag, a suspicion and a finding. The episode also shows the danger of automation bias: a numerical accuracy figure can appear more objective than it is. The committee should recommend periodic review of online assessment policy, including alternative assessment designs that rely less on surveillance. The immediate lesson is that technology can prioritise cases for review but cannot replace evidence and natural justice.