Dissecting the scenario stem

Lesson 1 of 4 in Exam Mastery: Scenario Technique, Pacing, and the Credential Landscape.

An AIGP scenario item has three moving parts: the exhibit (a paragraph or two describing an organisation and its AI system), the stem (the actual question sentence), and four options engineered so that at least two feel right. IAPP frequently attaches several stems to one exhibit — a single fact pattern about, say, a fine-tuned hiring model can feed a role question, a testing question, and a monitoring question in sequence.

That structure dictates the technique. Candidates who read the exhibit first, absorb everything, and then meet the stem waste their working memory on details that belong to a different question. Invert it.

The twenty-second read: five moves, in order

  1. New scenario item appears
  2. 1. Read the stem first

    Know what you are hunting before entering the exhibit. ‘What should the company do NEXT?’ reads the exhibit very differently from ‘Which law applies?’

  3. 2. Fix the role

    Is the organisation acting as developer, provider, deployer, user — or two at once? Duties follow role; half the distractors are duties that belong to the other seat.

  4. 3. Pin the lifecycle stage

    Pre-build, build/test, release, or post-deployment? The correct control matches the stage the exhibit describes — a real control at the wrong moment is still wrong.

  5. 4. Decode the Bloom verb

    Identify/define/describe → recall, answer fast. Apply/evaluate/implement/recommend → scenario reasoning: pick the best next step, not the encyclopaedia.

  6. 5. Eliminate by trap pattern

    Strike options that swap roles, shift stages, speak in absolutes, or offer a one-time act against an ongoing risk. Whatever survives role + stage + verb is your answer.

  7. Answer, flag if <70% sure, move on

Watch the technique run on a real-shaped item. Exhibit: ‘Orbis Retail, an EU-based chain, licenses a general-purpose model from a US vendor and fine-tunes it on ten years of its own hiring records to shortlist candidates. The project lead asks what testing must be completed before go-live.’

Stem first: ‘what testing before go-live’ — a Domain III.B question wearing Domain IV clothes. Role: fine-tuning on its own data pushes Orbis beyond pure deployer — development-side duties attach. Stage: pre-release. Verb: an apply question — it wants the right testing scope, not a definition. Eliminate: any option that defers testing to the US vendor (role-swap), any option that answers with post-market monitoring (stage-shift), and any option promising testing that ‘guarantees’ fairness (absolute language). What survives is disaggregated bias testing plus the rest of the seven-type pipeline, documented — performed by Orbis, because Orbis changed the model.

The accordion below breaks the same anatomy apart layer by layer — learn to see these four layers in every practice question you review.

Layer 1 — the exhibit: what is signal, what is scenery

Exhibits mix load-bearing facts (role, stage, jurisdiction, data type, who is affected) with scenery (company names, industries, headcounts). Signal words repay a highlight: fine-tunes, procures, biometric, before launch, six months after deployment, EU-based. When one exhibit feeds several stems, scenery for this stem may be signal for the next — never conclude a detail is useless, only that it is useless for the current stem.

Layer 2 — the stem: the verb is the difficulty dial

The BoK writes performance indicators with Bloom-graded verbs, and stems inherit them. Identify / define / describe → a recall item; if you know the term, answer in ten seconds and bank the time. Apply / implement / evaluate / recommend → a judgment item; expect two defensible options and a tiebreaker (role, stage, or continuing-vs-one-off). Also watch for qualifiers: BEST, FIRST, NEXT, MOST likely — they concede that other options are partially right, and they tell you the keyed answer wins on ordering or fit, not on being uniquely true.

Layer 3 — the options: engineered, not random

Distractors are built from real material: a genuine duty assigned to the wrong role, a genuine control at the wrong stage, a genuine framework term in the wrong framework. This is good news — it means elimination is a knowledge operation, not a guess. An option you cannot map to any trap pattern and cannot fault on role, stage, or verb is probably the key.

Layer 4 — the multi-select: exact set, no partial credit

Multi-select items state exactly how many answers to pick, and score all-or-nothing — IAPP is explicit that there is no partial credit. Treat each option as its own true/false question against the stem. If you are torn between two combinations, anchor on the option you are most confident is wrong and let it veto combinations that contain it.

Tool: AIGP Exam Simulator — Run the five-move read against live questions now — the exam simulator serves real scenario exhibits with per-option post-mortems.

Interactive checkpoint quiz (2 questions) — open this page in a browser to take it.