Evaluation Systems and Computer-Based Testing

These three words are used loosely in daily speech but are distinct in exams. Measurement is the process of assigning a number to a characteristic according to a rule (a student scored 62 out of 100); it is purely quantitative and says nothing about worth. Assessment is the broader process of gathering and interpreting evidence about learning, through tests, observation, projects and portfolios, usually to improve learning. Evaluation is the judgment of value or worth made by comparing the evidence with criteria or standards; it combines quantitative and qualitative information and leads to decisions (promotion, grading, programme change). The relationship is a ladder: measurement is the number, assessment is the evidence and interpretation, and evaluation is the value judgment. Evaluation is integral to teaching, not separate from it: it tells the teacher what to reteach, and tells the student where he or she stands. It is not a post-teaching activity only.

• Placement evaluation (before instruction): determines the appropriate level, group or course for a learner (the entry point).

• Diagnostic evaluation: identifies the specific causes of persistent learning difficulties; it goes deeper than formative testing. A student who keeps failing fractions should be diagnosed to find the exact sub-skill gap, not merely given a lower grade.

• Formative evaluation (term coined by Michael Scriven, 1967; applied to learning by Bloom): conducted during instruction to give feedback and improve teaching and learning; it is low-stakes and ongoing. Examples: quizzes, questioning, homework, observation. It is assessment for learning.

• Summative evaluation: conducted at the end of a unit, term or course to grade, certify or promote; it is high-stakes and gives a final judgment. It is assessment of learning. Annual and semester-end examinations are summative.

• Prognostic evaluation predicts future success (an aptitude test).

Norm-referenced evaluation compares a learner’s performance with that of a peer group (norm); the result is rank, percentile or standard score, and it spreads students along a normal distribution; it tells who is better, not whether a standard was achieved. Criterion-referenced evaluation (term from Robert Glaser, 1963) compares performance with a pre-set standard or criterion; the result is mastery or non-mastery of defined outcomes, and in principle every student can achieve the standard. Self-referenced (ipsative) evaluation compares a learner with his or her own earlier performance and is useful for motivation and progress tracking. Use norm-referenced for selection and ranking (entrance tests), criterion-referenced for certification of competence (driving test, professional licensing) and self-referenced for personal growth.

Tests can be objective (MCQ, true-false, matching: high objectivity, wide coverage, quick scoring, but poor at testing creativity), short answer and essay type (good for expression and reasoning, but lower objectivity and lower coverage). Non-test techniques: observation, checklist (presence or absence of a behaviour), rating scale (degree of a behaviour), anecdotal record (a short factual account of a significant incident), cumulative record card (complete developmental record over the years), sociometry (Moreno: mapping social relations in a group), interview and questionnaire, projects and assignments. A rubric is a scoring guide listing criteria with performance descriptors for different levels (it is not a grading curve or an answer sheet); analytic rubrics score each criterion separately and holistic rubrics give one overall score. A portfolio is a purposeful collection of a learner’s work over time that shows effort, progress and reflection, so it is not a single high-stakes exam and not a peer-assessment tool. Authentic assessment tests learners in real or realistic tasks (for example, a medical student examining a real patient), and it is considered highly valid for professional skills. 360-degree evaluation gathers feedback from multiple sources (teacher, peers, self, parents, sometimes administrators). Open-book examinations test application and reasoning more than memory.

• Validity: the test measures what it is intended to measure. Types: content validity (adequate sampling of the syllabus), construct validity (measures the intended trait), criterion-related validity (concurrent when compared with an existing test, predictive when it predicts future performance). Validity is the most important quality. A reliable test need not be valid, but a valid test must be reliable.

• Reliability: consistency of scores. Methods: test-retest, parallel (equivalent) forms, split-half (corrected by the Spearman-Brown formula), Kuder-Richardson (KR-20, KR-21) and Cronbach’s alpha (internal consistency), and inter-rater reliability for subjective scoring.

• Objectivity: scoring does not depend on the scorer’s opinion. Usability (practicability): easy to administer, score and interpret; economical. Discriminating power and appropriate difficulty level. Norms for interpretation.

• Item analysis: the difficulty index (p) is the proportion of examinees who answer the item correctly (higher p = easier item; a good item typically lies between about 0.3 and 0.7). The discrimination index (D) compares the top and bottom groups (usually the upper and lower 27 percent): D = (number correct in upper group minus number correct in lower group) ÷ number in one group. It shows whether the item separates strong from weak students.

CCE assesses all aspects of a child’s development continuously (frequent formative checks through the year) and comprehensively (both scholastic, meaning academic, and co-scholastic, meaning life skills, attitudes, sports, arts and values). It was promoted by the Right to Education Act (2009) and adopted by CBSE in 2009, and it embodies the principle that evaluation should reduce the stress of a single final exam. In higher education the equivalent is a combination of internal assessment and semester-end examination, giving students grades on the basis of both. A modern assessment programme emphasises formative assessment, rubrics, portfolios and outcome-based evaluation.

CBT delivers a test on a computer; it permits instant scoring, randomised questions and option order, multimedia items, secure digital records and flexible scheduling. UGC NET has been a computer-based test conducted by the NTA since December 2018. Important terms:

• Item bank (question bank): a large pool of pre-tested, calibrated test items stored with statistics on difficulty and discrimination, from which tests are assembled. It is not a bank for exam fees, answer scripts or a storage room for paper.

• Computerised Adaptive Testing (CAT): the difficulty of each question is adjusted dynamically according to the candidate’s previous answers (a correct answer leads to a harder item, a wrong one to an easier item). It uses Item Response Theory (IRT) and estimates ability with fewer questions and greater precision. CBT in general merely delivers the test by computer; CAT adapts.

• Percentile: a percentile of 95 means the candidate scored better than (or equal to) 95 percent of the candidates; it is not 95 percent marks, not rank 95 and not 95 questions answered. Percentile scores are used when a test is held in multiple shifts to normalise results, because the question papers differ slightly in difficulty.

• Remote proctoring: online examination with live invigilation through webcam and software, plus AI flagging of suspicious behaviour. Secure browsers and biometric verification reduce cheating. Mark for review (flagging) is a useful CBT feature: candidates can flag unsure questions and return to them.

• Equity issues: all candidates should use the same digital tools provided by the examination centre; digital divide and familiarity with computers can be concerns.