Improving Learning and Assessment with Confidence-Based Marking

Abstract accepted for the HE Academy Meeting on Science Teaching & Learning,
Warwick,  27-28 June, 2005

Tony Gardner-Medwin, Dept. Physiology, UCL, London WC1E 6BT

Confidence-based marking (CBM) has been known for many years to stimulate constructive thinking by students, and improve both the reliability and validity of exam data.  However, it is used in few places and is sometimes regarded with scepticism.  At UCL we have used a simple but carefully designed version of CBM for ten years, including 4 years as part of summative exams (see www.ucl.ac.uk ). I will discuss the rationale of CBM and the extent of success evidenced by student evaluations and analysis of our large bodies of data from formative and summative assessment. We have recently (with HEFCE FDTL4 funding) made it easy for other institutions and other disciplines to use and evaluate CBM, to adapt existing or new material for a powerful and flexible browser-based delivery system, and to integrate the grades obtained into Virtual Learning Environments, such as WebCT.

Students like CBM a lot, finding it more searching in identifying their areas of weakness or misconception, and they fully support its use in exams. An important issue often raised by staff is concern that CBM would somehow favour one or other gender, or particular personality types. Our data shows no evidence for gender differences, despite high sensitivity and a small but clearly significant tendency for both sexes to be more cautious in expression of high confidence in high-stakes exams. Where students are excessively diffident or confident about their knowledge, CBM motivates them to calibrate their judgements correctly, since they always gain by estimating correctly the reliability of their answers, whether this is high or low.  Exam data show that CBM scores are both more reliable (Cronbach alpha) as measures of student performance and better predictors even of the number of correct answers that students will obtain on different questions.