Greg Foley is a lecturer in the School of Biotechnology.
Feedback is one of those things in education that is always seen as a ‘Good Thing’. (An excellent paper on feedback by John Hattie can be found here.) But providing feedback also has its dangers. To explain, I need to divert. A few years ago I did a bit of research on artificial neural networks for data analysis in chemical engineering. Artificial networks are essentially a fancy form of non-linear regression and while they are very powerful, they can be misused. One of the dangers of ANNs is that that if you don’t know what you are doing you can design networks that are ‘over-trained.’ To illustrate what over-training is, considering the two curve fits shown below. The first fit, a simple linear regression, would seem to have captured the essence of the relationship between the dependent variable and the independent variable. On the other hand, the second fit which actually matches the data exactly, is probably ‘over-trained’. It is seeing patterns where it is likely that none really exist. This fit cannot ‘see the wood for the trees’.
So what has this got to do with feedback? Often, when I am giving oral feedback to students (in a lab for example) I am struck by the fact that students tend to want a sort of recipe for success. They want to be told precisely what they need to do to score a high mark. They don’t like generalities like being told that their graphs and tables should be presented in a ‘logical order’ or that they need to improve their attention to detail. Indeed, it is often the best and most ambitious students who desire this level of precision in the feedback they receive. It’s as if students want to be ‘over-trained’ so that their work matches exactly the ‘perfect’ lab report where ‘perfect’ is defined by the lecturer’s marking scheme.
Getting the balance right between providing students with useful guidance and facilitating them to jump through hoops can be tricky.