AR Gardner-Medwin

  • The skill of knowing how reliable your knowledge is

    AR Gardner-Medwin, University College London

    NA Curtin, Imperial College

    Research seminar

    Themes: assessment, skills development and lifelong learning

    In both formative and summative assessment we should take account of the students' ability to justify and express reservations about answers, as well as the correctness of the answers. A lucky guess does not deserve the credit due for confident and well justified knowledge, and a firm misconception is much worse than acknowledged ignorance.

    This paper will provoke discussion about how this skill relates to assessment of knowledge, and how it can be developed and assessed with a well designed scheme for Confidence-Based Marking (CBM: see www.ucl.ac.uk/lapt ). Though the concept has been well researched over some decades, mostly before large-scale IT implementation was practical, it has seldom been widely applied. A simple scheme has been in use with medical classes for many years now at UCL, initially set up to help stimulate critical reflection and good study habits. In recent years it has been used also, partly at the students' instigation, for exams (with OMR technology). We use a 3-point confidence scale: 1,2,3, entered for each answer, with 1,2 or 3 marks awarded when the answer is right and 0,-2,-6 when it is wrong. Recently a link has been set up between the CBM software and student VLE systems, using WebCT at UCL and Imperial College and a locally developed VLE at Winchester College. This provides simple access to formative tests with automated grade recording and management. A wealth of data, much but not all of it published so far (see the website), addresses the important conceptual as well as practical issues that arise through use of CBM.

    We hope to provoke discussion about these issues and address them where possible with relevant data and experience. Most people recognise that the skill of knowing and expressing the reliability of knowledge is crucial to human interaction and highly valued, though it is seldom taught or subject to practice or assessment. Relevant data include student evaluations and comments, comparison of performance by relatively naive and well practised subjects and comparisons by gender, ethnic grouping and formative vs summative assessments. Staff new to the concept are sometimes concerned that it might merely reward self-confidence rather than knowledge; but this can be shown to be not at all true. There is much statistical data about reliability and validity in exam assessment, each significantly improved with CBM.

    For those interested in evaluating CBM in their own contexts or in relation to any of a wide range of disciplines and question styles, the web site should be the main source; but practical issues may also be raised and discussed in the session.

    Acknowledgements: DA Bender, P Tatham, J Levy (UCL), M De Iorio, M Burrow (Imperial), C Ryan, H Bassett (Winchester College), HEFCE FDTL4 for funding.