19-Soft-B2 User Interface · May 2015
Nivaar worked solution (AI-drafted; not reviewed by a licensed engineer)
National Exams, May 2015 — 04-Soft-B2, User Interface (closed book, 3 hours). Part A: answer any FIVE of the NINE questions (10 marks each); Part B: answer ALL FIVE questions (10 marks each), all based on the same case study — Happy Medical Clinic, a Toronto medical clinic switching from paper-based practice to an electronic health record (EHR) system purchased from VisualEHR Inc., an off-the-shelf vendor willing to customize the product to the clinic's needs. Most questions call for essay-format answers; clarity and organisation count. This solution answers all fourteen questions as a full study resource.
Reference texts. Rogers, Sharp & Preece, Interaction Design: Beyond Human-Computer Interaction, 5th ed., Ch. 1–3 (interaction design, cognitive aspects, mental models), Ch. 9–10 (prototyping, personas), Ch. 11–12 (data gathering, requirements), Ch. 15–16 (evaluation, lab vs. field studies); Nielsen, Usability Engineering, Ch. 4–6 (usability heuristics, iterative design, usability testing); Shneiderman, Designing the User Interface, 6th ed., Ch. 2 (guidelines, principles), Ch. 12 (internationalization); Norman, The Design of Everyday Things, Ch. 1–4 (visibility, affordances, feedback, conceptual models).
Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.
Part A — how many participants. Nielsen's rule of thumb is that around 5 participants are sufficient to uncover roughly 80–85% of a design's usability problems in a single round of formative (diagnostic) testing. This follows from the mathematics of problem discovery: if each participant independently has some probability p (Nielsen estimates p ≈ 0.31 on average) of encountering any given usability problem, the proportion of problems found after n participants is 1 − (1 − p)n, a curve that rises steeply for the first few participants and then flattens — each additional participant beyond about 5 mostly re-discovers problems already found by earlier participants, so the marginal value of testing a 6th, 7th, 8th participant in the same round drops sharply. The recommended practice is therefore not "test 5 and stop forever," but "test ~5, fix what was found, then test another small batch (~5) in the next iteration" — spreading the same total testing effort across more iterative rounds finds more distinct problems than spending it all in one large round.
Part B — when to test more users. Testing with more than ~5 participants in a single round is warranted when: (1) the user population is heterogeneous — Happy Medical Clinic's EHR will be used by receptionists, nurses and physicians with materially different tasks, expertise and goals, and Nielsen's ~5-user estimate assumes a single, fairly homogeneous user group, so each distinct user category needs its own ~5 participants; (2) the study is summative (measuring quantitative performance/satisfaction scores for statistical comparison or sign-off, e.g. "is this redesign measurably faster than the old system?") rather than formative/diagnostic — a valid statistical comparison needs a larger, more representative sample than a qualitative problem-finding pass; (3) the system or task is unusually complex or high-risk, where an undiscovered problem carries a high cost (a rare-but-serious workflow, such as an emergency medication order in the EHR) and finding a higher fraction of the remaining ~15–20% of problems is worth the extra testing cost; or (4) budget and access allow it and the marginal cost per additional participant is low relative to the value of the problems still likely to be missed.
