## Inline Choice (Dropdown / Cloze) Question Evaluation

This question is an INLINE CHOICE item — a passage with one or more blanks, each presenting a
small dropdown of 2–4 options. The student selects the best option for each blank directly within
the passage context.

It differs from a free-text fill-in (where the student writes their own answer) and from a
standalone MCQ (where the options are presented separately). The key property: the correct choice
must be determinable from the **passage context surrounding the blank**, not from grammar or
general knowledge alone.

Apply the following type-specific evaluation rules in addition to the general evaluation procedure above.

---

## STEP 2.9: INLINE CHOICE STRUCTURAL AUDITS (MANDATORY — run before Step 3 scoring)

### AUDIT IC-1 — Passage-Context Dependency → educational_accuracy

**Trigger:** Always fires.

**Emit `ic1_context_dependency` in `educational_accuracy.internal_reasoning`:**
```
ic1_context_dependency:
  blank_id: [blank number or position, e.g. "Blank 1"]
  blank_sentence: [copy the full sentence containing the blank verbatim]
  correct_option: [copy the correct option text]
  passage_context_used: [quote the specific surrounding passage text that makes the correct option determinable]
  eliminable_without_passage: [YES / NO — can the student pick the correct option using grammar/logic/general knowledge alone, without reading the passage?]
```
**MUST-FAIL** (`educational_accuracy = 0.0`) if `eliminable_without_passage: YES` for any blank.
A correct inline choice must require the student to have read and understood the passage — it is
NOT acceptable if grammar alone (e.g., article agreement) or common knowledge determines the answer.

### AUDIT IC-2 — Distractor Plausibility → distractor_quality

**Trigger:** Always fires.

**Emit `ic2_distractor_plausibility` in `distractor_quality.internal_reasoning`:**
```
ic2_distractor_plausibility:
  for each blank:
    blank_id: [blank number or position]
    for each incorrect option:
      option_text: [copy verbatim]
      plausible_misreading: [describe the reading error or misunderstanding that would lead a student to pick this — or NONE if it is eliminable without reading]
```
**MUST-FAIL** (`distractor_quality = 0.0`) if any incorrect option has `plausible_misreading: NONE`.
Every distractor must represent a genuine misreading of the passage — not a grammatically wrong
option, an obviously off-topic word, or one no reader would seriously consider.

### AUDIT IC-3 — Blank Coherence → clarity_precision

**Trigger:** Always fires.

**Emit `ic3_blank_coherence` in `clarity_precision.internal_reasoning`:**
```
ic3_blank_coherence:
  for each blank:
    blank_id: [blank number or position]
    blank_sentence_with_each_option_inserted: [for each option, write the resulting sentence]
    any_option_produces_ungrammatical_sentence: [YES / NO]
    any_option_produces_identical_meaning: [YES / NO — do two options mean the same thing in context?]
```
**MUST-FAIL** (`clarity_precision = 0.0`) if:
- Any option produces an ungrammatical sentence when inserted (all options must fit grammatically —
  the skill being tested is comprehension, not grammar correction).
- Two or more options produce identical meaning in context (the blank has no unique correct answer).

---

## Type-Specific Metric Notes

- **passage_reference**: The blank is embedded IN the passage — by definition, a passage is always
  present. Pass when the options are drawn from the passage's vocabulary, topic, or meaning. Fail
  if the options are unrelated to the passage's content.

- **distractor_quality**: Distractors should be words/phrases that are contextually plausible but
  incorrect — e.g., a near-synonym that changes the meaning, a word from a different part of the
  passage, or a word that fits the topic but not the specific sentence. NOT: random unrelated words,
  obviously wrong grammatical forms.

- **curriculum_alignment**: Inline choice items in Reading/ELA should test vocabulary-in-context,
  author's word choice, or reading comprehension — not spelling or grammar conventions. Verify
  the TEKS/standard being assessed is genuinely about selecting the best word/phrase in context.

- **mastery_learning_alignment**: Passes if the correct choice requires understanding the passage's
  meaning, tone, or argument. Fails if any test-taker could identify it from the sentence structure
  alone without reading the surrounding passage.

- **educational_accuracy**: Apply the standard answer-giveaway probe — if a student who has NOT
  read the passage can reliably pick the correct option by grammar or elimination, the item is a
  giveaway. The text-evidence carve-out does NOT apply here (this is not a cite-evidence standard).
