Score Calculation

From Expertiza_Wiki
Revision as of 17:03, 3 July 2026 by Blalu (talk | contribs) (Created page with "= Score Calculation = Peer review grades in Expertiza are computed in three layers, each handled by a separate model. == Layer 1: Response (Item Weights) == <code>Response#aggregate_questionnaire_score</code> sums <code>answer * item.weight</code> for all answered, scored items in a single submitted review. Section header items are excluded (they carry no weight). <code>Response#maximum_score</code> computes the maximum possible score for the same set of answered ite...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Score Calculation

Peer review grades in Expertiza are computed in three layers, each handled by a separate model.

Layer 1: Response (Item Weights)

Response#aggregate_questionnaire_score sums answer * item.weight for all answered, scored items in a single submitted review. Section header items are excluded (they carry no weight).

Response#maximum_score computes the maximum possible score for the same set of answered items: sum(item.weight) * questionnaire.max_question_score.

Layer 2: ResponseMap (Round Weights)

ResponseMap#review_grade normalises each round's score (aggregate_questionnaire_score / maximum_score) and computes a weighted average across rounds using AssignmentQuestionnaire#questionnaire_weight.

Only the latest submitted response per round is used. Rounds with no submitted response are excluded from both the numerator and denominator.

Layer 3: AssignmentTeam (Reviewer Reputation)

AssignmentTeam#aggregate_reviewer_score averages ResponseMap#review_grade across all reviewers of the team, weighted by each reviewer's reputation score (currently defaulting to 1.0 — placeholder for future Uchswas integration).

The same computation is used by AssignmentParticipant#aggregate_teammate_review_grade for teammate reviews.

Validation

AssignmentQuestionnaire validates that questionnaire_weight is 0 when the linked rubric contains no scored items (only SectionHeader items), since a non-zero weight on an unscored rubric would produce meaningless grades.