## AP Chemistry Evaluation Rules

These rules apply to AP Chemistry items of any type (discrete MCQ, stimulus-based MCQ
set, and long/short free-response). They add AP-specific guidance to the metric
definitions already in effect; they introduce no new metrics. Apply each concern to the
metric it naturally belongs to. Default stance: a well-formed, AP-faithful item passes.

### Course framework alignment → `curriculum_alignment`
A valid AP Chemistry item assesses a specific piece of course content together with a
course skill. The framework's content (units, topics, learning objectives) and skills
are supplied at evaluation time through the **provided curriculum context** — do not
assume or rely on a hard-coded list of units or skills here.

- When curriculum context is provided (unit/topic, learning objective, and/or the
  skill/practice being assessed), verify the item genuinely assesses that content and
  that the named skill is the one the item actually exercises.
- The **task verb should match the assessed skill**: a quantitative/mathematical skill
  ↔ calculate/determine; an argumentation skill ↔ justify/explain/evaluate-a-claim; a
  model- or data-representation skill ↔ draw/represent/complete a diagram or graph; an
  experimental-design skill ↔ design or analyze a procedure. A task verb that
  contradicts the skill the item claims to assess is a defect.
- **Out-of-scope content (Exclusion Statements).** The AP Chemistry CED marks specific
  content as **not assessed on the exam** ("Exclusion Statements"). When the provided
  curriculum context includes **Assessment Boundaries**, an item that requires the
  excluded content, computation, or terminology is out of scope for AP — flag it under
  `curriculum_alignment` **even if it is chemically correct**. Examples of CED exclusions
  include computing the pH change when acid/base is added to a buffer, converting between
  Kc and Kp, Arrhenius-equation calculations, Lewis acid-base problems, colligative-
  property calculations, d-orbital hybridization, and phase-diagram interpretation. Treat
  the provided Assessment Boundaries as authoritative; do not invent exclusions beyond them.
- Science Practice 3 (Representing Data and Phenomena) is assessed only on the free-
  response section, not on multiple-choice; a discrete MCQ whose task is purely to
  *produce* a graph or representation is mis-aligned to its section.

### Provided reference materials → `educational_accuracy`
- A periodic table and the standard equations/constants sheet are available throughout
  the AP exam, so a valid item must **not require a memorized constant** (specific
  equilibrium/acid-dissociation values, molar masses, standard potentials, etc.). Any
  value the solution needs must be derivable, on the reference materials, or supplied
  in the stem. An item that hinges on a recalled constant is broken.
- When the provided curriculum context lists reference materials/equations for the
  topic, treat those — together with the periodic table and the standard equations/
  constants sheet — as the available toolkit: the item should be solvable from that
  toolkit plus the stem.

### Single defensible answer (MCQ and MCQ sets) → `factual_accuracy` / `distractor_quality`
<!-- This overlay is loaded for BOTH the question evaluator (discrete MCQ + FRQ) and the
MCQ-set evaluator, which appends it via the shared subject/curriculum-subject resolver in
question.py. The MCQ-set metric set differs from the question one, so the MCQ-set evaluator
frames this overlay's concerns onto its own metrics. The rules below are written to apply
to both a single MCQ and the items within a stimulus-based set. -->
- A discrete MCQ has exactly four options with exactly one defensibly correct answer;
  flag any item where a second option is also defensible under a reasonable reading, or
  where the "best" answer depends on an unstated assumption.
- In a stimulus-based set, every item must be answerable **from the shared stimulus
  alone**, independently of any other item's answer; uniform (not escalating)
  difficulty across the set is expected.
- Distractors should each correspond to a **specific, predictable chemistry error**
  (e.g. an inverted ratio, a dropped stoichiometric coefficient, confusing a quantity
  with its logarithm, using mass where moles are required, selecting a conjugate
  instead of the named species) rather than arbitrary wrong values.

### AP free-response scoring conventions (FRQ items only) → `educational_accuracy` / scoring-criteria clarity
This section applies only to free-response items (`frq_long` / `frq_short`); ignore it
for discrete MCQ and MCQ-set items, which are scored under the closed-ended rules.
- AP FRQ rubrics award points for discrete scorable elements and credit work that is
  **consistent with a student's earlier answer** (error-carried-forward). For
  justify/explain/claim-evaluation parts, AP awards the point only for a correct
  response **and** a valid justification tied to the relevant principle — a correct
  answer with missing or wrong reasoning earns nothing. Items and their scoring
  criteria should reflect these conventions.
- **Units and significant figures** are part of AP FRQ scoring: a calculate/determine
  answer is expected to carry correct units and an appropriate number of significant
  figures (AP rubrics commonly include a discrete point or penalty for these). An item
  whose given data cannot support a defensible significant-figure answer, or whose
  scoring ignores units where they are required, is weak.
- Symbolic answers follow AP conventions: balanced and net-ionic equations show correct
  states and electrolyte forms (strong electrolytes dissociated; weak acids/bases,
  solids, gases, and water intact); **redox half-reactions balance both atoms and charge,
  with electrons canceling in the combined equation**; particulate representations
  conserve atoms, show ionic species as separated ions in solution, and reflect the
  correct stoichiometric ratio.

### AP representation types → `educational_accuracy` / `stimulus_quality`
AP Chemistry uses a recurring set of representations (in stimuli and in draw/represent
tasks) that have characteristic, checkable features. A representation that violates them
is defective:
- **Particulate/particle diagrams:** conserve atoms and charge, correct stoichiometric
  ratio, ionic species shown separated in solution.
- **Photoelectron (PES) spectra:** peak heights proportional to electron counts; peak
  positions ordered by subshell binding energy.
- **Mass spectra:** peaks placed by isotopic mass with heights reflecting relative
  abundance (single element, singly charged ions only).
- **Titration curves:** equivalence-point pH and buffer region consistent with the
  acid/base strength.
- **Reaction-energy profiles:** relative reactant/product energies, activation energy,
  and catalyzed-vs-uncatalyzed paths drawn coherently.
- **Maxwell–Boltzmann distributions:** shape/shift consistent with the stated temperature.

### Physical realism (AP-specific traps) → `factual_accuracy`
- Reject setups that imply impossible chemistry: equilibrium constants, reaction
  quotients, concentrations, pH for dilute aqueous systems, and percent yields must be
  internally consistent and within defensible ranges; thermodynamic signs
  (ΔH, ΔS, ΔG) and their relationships must be coherent; rate expressions and their
  units must be consistent with the stated order.
- Redox/electrochemistry must be self-consistent: half-reactions conserve both mass and
  charge, electrons cancel when combined, and a standard cell potential must agree with
  the half-cell potentials it is built from (sign and magnitude).
