Performance variations of the Bayesian model of peer-assessment implemented in OpenAnswer response to modifications of the number of peers assessed and of the quality of the class

The paper presents a study of the performance variationsoftheBayesianmodelofpeerassessmentimplementedin OpenAnswer, in terms of the grades prediction accuracy. OpenAnswer (OA)modelsapeerassessmentsessionasaBayesian network. For each student, a subnetwork contains variables describingrelevantaspectsofboththeindividualcognitivestateand the state of the current assessment session. Subnetworks are interconnected to each other to obtain the final one. Evidence propagated through the global network is represented by all the gradesgivenbystudentstotheirpeers,togetherwithasubsetofthe teacher’scorrections.Amongthepossibleaffectingfactors,thepaper reportsabouttheinvestigationofthedependenceofgradesprediction performance on the quality of the class, i.e., the average level of proficiency of itsstudents,andon thenumberofpeersassessedby eachstudent.Theresultsshowthatbothfactorsaffecttheaccuracyof the inferred marks produced by the Bayesian network, when comparedwiththeavailablegroundtruthproducedbyteachers.