Practice · Requirements Engineering II: Specifying and Managing

Q1

"The system shall be user-friendly." Which quality of a good requirement does this most clearly lack?

A Singular B Necessary C Verifiable D Traceable

Q2

"Buddy shall never give a wrong answer." What is the main problem?

A It is not feasible: an AI component cannot guarantee zero errors. B It is too short. C It is a constraint. D It uses "shall".

Q3

"Students can post questions and vote on answers and search old threads." Which quality is violated?

A Feasible B Singular C Necessary D Consistent

Q4

R5 says "Posts are anonymous to other students." R9 says "Every post displays the author's full name." These two requirements are…

A inconsistent B both unnecessary C ambiguous but consistent D fine, because they are in different sections

Q5

"The post shall be deleted after the course ends." What ambiguity does this contain?

A It is written in EARS. B It uses a number. C Passive voice hides the actor and "ends" is undefined (last lecture? final exam? grade appeal deadline?). D None — it is perfectly clear.

Q6

Which EARS pattern fits: "____ a login fails 5 times within 10 minutes, ____ the system shall lock the account for 15 minutes."

A While … (state-driven) B If … then (unwanted behaviour) C Where … (optional feature) D Ubiquitous

Q7

"____ the AI service is unavailable, the system shall show 'Buddy is offline'." Which keyword belongs in the blank?

A While B When C If … then D Where

Q8

"Where anonymous posting is enabled for a course, the system shall display 'Anonymous student'." Which EARS pattern is this?

A Optional feature B State-driven C Unwanted behaviour D Event-driven

Q9

Which is the best measurable version of "Buddy must be fast"?

A "Buddy must be very fast." B "Buddy must use a fast server." C "Buddy's average response time must be low." D "For 95% of questions, the first word of Buddy's answer appears within 5 s, with 300 concurrent users."

Q10

Why is "95% of requests within 2 s" usually better than "average response time 2 s"?

A Percentiles are easier to compute. B Averages cannot be measured. C An average can hide a large group of very slow requests; a percentile limits how many users get a bad experience. D There is no difference.

Q11

In an SRS following IEEE 830 / ISO 29148, where do the glossary and definitions belong?

A Introduction B Specific requirements C Nowhere; SRS documents have no glossary D Appendix of test results

Q12

An agile team says: "We don't write specifications; we only have stories with Given/When/Then acceptance criteria." What is the most accurate comment?

A They have no specification at all. B The acceptance criteria are a precise, testable specification written in small pieces. C Acceptance criteria are only for plan-driven teams. D Stories replace the need for any quality requirements.

Q13

In MoSCoW, what does the W mean?

A Worth doing B Weekly C Won't have (this time) D Wish list

Q14

The team has 100 points of capacity for release 1. The Must-haves already total 95 points. What is the main risk?

A None; they will finish all Musts. B They should add more Could-haves. C MoSCoW does not use capacity. D There is almost no buffer for surprises, so the release may fail to deliver even its Musts; a common guideline keeps Musts around 60% of capacity.

Q15

On a value/effort grid, a feature with high value and low effort is called…

A a big bet B something to avoid C a fill-in D a quick win

Q16

A traceability matrix shows that requirement R3 ("Buddy gives hints only for graded work") has no linked test. What does this reveal?

A A coverage gap: an important requirement is not verified. B Nothing; not every requirement needs verification. C R3 is unnecessary. D R3 is an orphan test.

Q17

A test named testEmojiReactions is linked to no requirement. This is called…

A a baseline B a coverage gap C a change request D an orphan — it may test unnecessary work or reveal a missing requirement

Q18

The instructor asks to change "hide a post after 3 reports" to "after 5 reports". What should the team do before accepting it?

A Change the number in the code immediately. B Record a change request and do an impact analysis using traceability (design, code, tests, moderation workload), then let the Product Owner decide. C Refuse all changes after the baseline. D Ask an AI agent to update everything.

Q19

What is scope creep?

A The gradual growth of scope through small, unrecorded additions that together delay the project B A tool for measuring code size C Removing features to meet a deadline D A type of security attack

Q20

A tester cannot write any test for the requirement "The system shall handle large files appropriately." This tells us that the requirement is…

A perfect, because testers should be creative B only relevant to the database C a constraint D not verifiable (and ambiguous) — it must be rewritten before building

Q21

Buddy's "graded question" detector is tested on 100 questions: TP = 18, FP = 6, FN = 2, TN = 74. What is its recall for "graded"?

A 0.75 B 0.92 C 0.90 D 0.18

Q22

For the same detector, which error is worse from an academic-integrity point of view?

A Both are equally bad B False negative: a graded question treated as ungraded (Buddy may give a full solution) C True negatives D False positive: an ungraded question treated as graded (student gets hints only)

Q23

Only 3% of questions are about graded exercises. A detector that always answers "not graded" has what accuracy, and what does this show?

A 50% B 100% C 97% — accuracy is misleading when one class is rare; recall for "graded" is 0 D 3% — accuracy punishes it correctly

Q24

Omar gives an AI coding agent the requirement "hide a post after 3 reports". It implements "3 reports, even from the same student". What is the best lesson?

A Never use numbers in requirements. B Ambiguity in a spec turns into silent assumptions; the requirement should state "3 reports from 3 different students", and the agent should be asked to list its assumptions. C Agents cannot write code. D The agent should have chosen randomly.

Q25

Which prompt is the best way to use an AI assistant for requirements review?

A "Add any requirements you think are missing and mark them as approved." B "Write the complete SRS for StudyBuddy." C "Is this requirement good? Answer yes or no." D "List every question a developer would need answered to implement this requirement, and give two different possible interpretations of it."

Q26 Short answer

Rewrite the following requirements using EARS templates, making them unambiguous and testable. State any assumption you make.

(a) "Notifications should be sent when appropriate." (b) "The system must cope with AI outages." (c) "Users who misbehave will be blocked."

Q27 Short answer

Explain traceability and describe the three practical questions a traceability matrix answers, with a StudyBuddy example for each.

Q28 Short answer

Buddy's graded-question detector gave TP = 45, FP = 15, FN = 5, TN = 135 on the instructor's test set. Compute precision, recall, F1 and accuracy (two decimals), and say whether it meets the requirement "recall ≥ 0.95 and precision ≥ 0.80".

Q29 Short answer

List five elements that a complete specification of an AI feature (such as Buddy) should contain beyond ordinary functional requirements, with one example sentence for each.

Q30 Short answer

Here are seven candidate items for StudyBuddy's first release (capacity 60 points): WKU login (8), post a question (8), privacy notice & consent (5), Buddy answers from course files (21), search (8), emoji reactions (3), study-group matching (13). Assign MoSCoW priorities and justify, respecting the guideline that Musts use at most about 60% of capacity (36 points).