An experimental analysis of assessor specific bias in a programming assessment in multi-assessor scenarios utilizing an eye tracker

It has been experienced and reported by academic institutions around the globe that marking of most subject's assessment scripts vary when different assessors are utilized for a given subject. To understand the difference, we capture and analysis cognitive response of assessors through the visual pattern while they are marking the scripts. For this, a Java programming assessment from a real life university examination is marked by independent assessors. The assessors marked the scanned assessment scripts on a computer screen in front of an Eye tracker machine and their eye gaze data were recorded real time. Data indicate that different assessors marked the same answer script differently and their visual pattern are also varied although they were given the exact same instructions which demonstrates bias to a degree. For quality marking, several findings including the number of assessors needed are also presented in this manuscript.