Assessment, Evaluation, and Learning Measurement
What to know
Assessment splits into formative (during learning, low stakes, feedback-oriented) and summative (end of learning, judgmental). Diagnostic assessment precedes instruction. Quality criteria: validity (measures the intended construct), reliability (consistency across raters/occasions), fairness, and practicality. Criterion-referenced assessment compares to a standard; norm-referenced compares learners to each other.
Item writing matters: stems should be clear, one correct answer, plausible distractors, no "all of the above" ambiguity, and no clues in grammar. Rubrics (analytic vs. holistic) improve reliability for performance tasks. Kirkpatrick's four levels-reaction, learning, behavior, results-structure evaluation; Phillips' ROI adds a fifth level. Transfer (behavior on the job) requires follow-up measurement. Use pre/post designs cautiously; gains can reflect testing effects. Reliability is necessary but not sufficient for validity.
Traps
Classic confusions: calling a quiz "valid" because it's reliable; labeling a mid-course quiz summative; or treating high satisfaction scores as proof of learning (Level 1 ≠ Level 2). Another trap is criterion contamination-grading on unrelated factors like attendance. Watch for ceiling/floor effects and for rubrics that are too vague to score consistently.
Rule of thumb
Ask "What decision will this result drive?" If it's feedback to improve learning → formative. If it's a certification decision → summative, and it must be valid and reliable. Worked example: a course with 95% satisfaction but poor on-the-job compliance → evaluate at Kirkpatrick Level 3 (behavior) with observation or audit data, not more satisfaction surveys.