Exploring the Validity of Comparative Judgement: Do Judges Attend to Construct-Irrelevant Features?

Abstract

When completing a comparative judgment (CJ) exercise, judges are asked to make holistic decisions about the quality of the work they are comparing. A key consideration is the validity of expert judgements. This article details a study where an aspect of validity, whether or not judges are attending to construct-irrelevant features, was investigated. There are a number of potentially construct-irrelevant features indicated in the assessment literature, and we focused on four features: appearance; handwriting; spelling, punctuation, and grammar (SPaG); and missing response vs. incorrect answer. This study explored this through an empirical experiment supplemented by judge observation and survey. The study was conducted within an awarding organisation. The particular context was within a programme of work trialling, a new method of maintaining examination standards involving the comparative judgement of candidates’ examination responses from the same subject from two different years. Judgements in this context are cognitively demanding, and there is a possibility that judges may attend to superficial features of the responses they are comparing. It is, therefore, important to understand how CJ decisions are made and what they are or are not based on so that we can have confidence in judgements and know that any use of them is valid.