Every option below carries its key, the reason it earns that key, and the arithmetic in full. This is the part that is normally invisible: prep sites publish a question and name an answer, and vendors publish neither the working nor what a wrong answer means.
Scenario 1 · Deduction — the invalid form almost everyone accepts
A support policy states: if a ticket is tagged Priority 1, it is escalated to the duty manager. Ticket 4471 was escalated to the duty manager.
What follows about ticket 4471?
| Option | Key | Why it earns that weight |
|---|
| It may or may not have been tagged Priority 1. | CorrectKeyed answer | Correct, and the keyed answer is a refusal. The policy says every P1 ticket gets escalated; it does not say escalation happens only for P1 tickets, so an escalation is consistent with the tag and with several other routes. |
| It was tagged Priority 1. | IncorrectDiagnostic distractor | Affirming the consequent — the single most commonly accepted invalid inference, and the one the AssessAll graduate bank explicitly builds traps for. It reads a one-way rule as if it ran both ways. |
| It was not tagged Priority 1. | IncorrectDiagnostic distractor | The mirror error. Diagnoses a candidate who has noticed the first option is a trap and over-corrected into the opposite unwarranted claim. |
| The policy was breached. | IncorrectDiagnostic distractor | Reads a rule that requires an action as one that forbids it in other cases. Diagnoses a reader importing an assumption the text does not carry — the same failure mode a Cannot Say item catches on a verbal form. |
The working. Write the rule as P1 → escalated. We are told the consequent, escalated. From "P → Q" and "Q", nothing whatever follows about P. The two valid moves are modus ponens (P → Q with P, giving Q) and modus tollens (P → Q with not-Q, giving not-P). This item offers neither, which is the point.
What the item separates. A form that never keys "may or may not" cannot separate a careful reasoner from a confident one. Ask any vendor whether their deductive items include unwarranted-conclusion keys, because a bank without them rewards decisiveness and calls it logic.
Scenario 2 · Deduction — contrapositive, converse and inverse
An access rule states: every contractor with system access has completed the security module.
Which of these must be true?
| Option | Key | Why it earns that weight |
|---|
| Anyone who has not completed the security module does not have contractor system access. | CorrectKeyed answer | Correct. This is the contrapositive, and it is the only one of the three rewrites that is guaranteed to be true whenever the original is. |
| Anyone who has completed the security module has contractor system access. | IncorrectDiagnostic distractor | The converse. True only if the rule also runs the other way, which it does not say. This is the same error as item 1 in a different grammatical dress — which is why it is worth having both on a form. |
| Anyone without contractor system access has not completed the security module. | IncorrectDiagnostic distractor | The inverse. Equivalent to the converse, and picked by candidates who negate both halves and assume that preserves the rule. |
| Contractors complete the security module before access is granted. | IncorrectDiagnostic distractor | A claim about sequence that the rule does not make. Diagnoses a candidate reasoning from how the process probably works rather than from what was stated — the most job-relevant error on the item, because that is exactly how policies get misapplied at work. |
The working. The rule is: contractor-with-access → completed-module. Contrapositive: not-completed → not-contractor-with-access. Valid, always. Converse: completed → contractor-with-access. Invalid. Inverse: not-contractor-with-access → not-completed. Invalid, and logically equivalent to the converse.
What the item separates. Three of the four options are the same rule read in a direction it does not license. That is what a well-built deductive distractor set costs to write, and it is why the option a candidate picked is more informative than the fact that they were wrong.
Scenario 3 · Induction — where "nothing follows" becomes the wrong answer
A team observes that every support ticket resolved in under an hour last month came from a customer on the Enterprise plan.
Which is the best-supported conclusion?
| Option | Key | Why it earns that weight |
|---|
| Enterprise tickets appear to be resolved faster, and the claim is worth testing against the tickets that took longer. | CorrectKeyed answer | Correct. Induction produces the best-supported generalisation plus the test that would break it — not a certainty. Naming the test is what separates a hypothesis from a hunch. |
| Being on the Enterprise plan causes faster resolution. | IncorrectDiagnostic distractor | The causal leap. Enterprise customers may raise simpler tickets, be routed to senior agents, or sit in a friendlier timezone; the observation cannot choose between those. |
| All Enterprise tickets are resolved in under an hour. | IncorrectDiagnostic distractor | The observation inverted. It said fast implies Enterprise, not Enterprise implies fast — the item-1 error reappearing inside an inductive question, which is why candidates who have only practised deduction still miss it. |
| Nothing can be concluded from a single month of data. | IncorrectDiagnostic distractor | The most interesting wrong answer on this page. It applies the deductive standard to an inductive question. Induction never yields certainty, so a rule that refuses every uncertain conclusion refuses all of them — including the ones a business is supposed to act on. |
The working. The observation has the form: resolved-under-1h → Enterprise. Option C reverses it. Option B adds a mechanism the data does not contain. Option D demands entailment from evidence that is, by construction, not entailing. Option A states the direction actually observed, marks it as provisional, and names the sample that would falsify it — the tickets that took longer, which were never examined.
What the item separates. Deduction and induction disagree about what a good answer looks like: on item 1 the key is a refusal to conclude, and here the same instinct is the distractor. A candidate applying the wrong standard fails an item they fully understand — which is the clearest practical proof that these are two scores, not one.
Scenario 4 · Abstract pattern — the rule is arithmetic, not perception
A sequence of symbol groups runs: ■□ , ■□□ , ■■□□ , ■■□□□ , ?
Which group comes next?
| Option | Key | Why it earns that weight |
|---|
| ■■■□□□ | CorrectKeyed answer | Correct. Written as (filled, open) the sequence is (1,1), (1,2), (2,2), (2,3) — each step increases exactly one count, alternating, and the last step raised the open count, so this one raises the filled count to give (3,3). |
| ■■□□□□ | IncorrectDiagnostic distractor | Repeats the step just taken instead of alternating. Diagnoses a candidate who inferred the rule from the last transition only — a local rule that fits one step and none of the others. |
| ■■■□□□□ | IncorrectDiagnostic distractor | Applies both increments at once, giving (3,4). Total length would jump by two where every previous step added one. Diagnoses averaging the two rules rather than alternating them. |
| ■□■□■□ | IncorrectDiagnostic distractor | Reads the groups as an alternating pattern of symbols rather than as counts of them. Diagnoses a candidate working from the visual surface, which is the failure mode a language-light item is specifically built to expose. |
The working. Convert to counts: (1,1), (1,2), (2,2), (2,3). Differences alternate +open, +filled, +open — so the next is +filled, giving (3,3), written ■■■□□□. Total length runs 2, 3, 4, 5, so the next group has six symbols; any option of length seven is eliminated before the rule is even applied.
What the item separates. The check that settles this item is arithmetic on the counts, and length alone eliminates one distractor. Language-light is not the same as culture-free or preparation-proof, and no honest abstract form should be sold as either.
Scenario 5 · Process logic — the format most catalogues call "diagrammatic"
A refund flow. Step 1: if the order is less than 30 days old, go to step 2; otherwise decline. Step 2: if the item is unopened, approve; otherwise go to step 3. Step 3: if the customer has had fewer than two refunds this year, approve as a goodwill credit; otherwise escalate to a supervisor. The case: an order placed 12 days ago, item opened, and this is the customer's third refund this year.
What does the flow return?
| Option | Key | Why it earns that weight |
|---|
| Escalate to a supervisor. | CorrectKeyed answer | Correct. 12 is less than 30, so step 2. The item is opened, so step 3. Three refunds is not fewer than two, so escalate. |
| Approve as a goodwill credit. | IncorrectDiagnostic distractor | The boundary error, and the most valuable wrong answer here: it reads "fewer than two" as "two or fewer", or simply stops reading at the friendliest branch. In a live process this is the error that costs money quietly. |
| Decline. | IncorrectDiagnostic distractor | Exits at step 1 despite the order being inside the window. Diagnoses a candidate who recognised that the case looks unfavourable and jumped to the outcome rather than tracing the branch. |
| Approve. | IncorrectDiagnostic distractor | Approves at step 2 despite the item being opened. Diagnoses a skipped condition rather than a misread one — a different remediation from the boundary error above. |
The working. Step 1: 12 < 30, so continue. Step 2: opened, so continue. Step 3: is 3 fewer than 2? No, so escalate. The strict inequality is the whole item: a customer on their second refund also escalates, because two is not fewer than two.
What the item separates. This format is sold across the market as diagrammatic or flowchart reasoning, and it is deduction — the rules are handed to you and nothing has to be inferred. Every error it catches is a branch or boundary error, which is precisely the competency behind SOP compliance and no-code automation, and not the same thing as spotting a pattern.