Critical Thinking

Impact is, however, only onepossible way of making

Impact of Adaptive Educational SystemBehaviour on Student MotivationJan Papou?sek and Radek Pel ?anekFaculty of Informatics, Masaryk University Brno{jan.papousek,xpelanek}@mail.muni.czAbstract.In this work we try to connect research on student model-ing and student motivation, particularly on the relation between taskdifficulty and engagement. We perform experiments within widely usedadaptive practice system for geography learning. The results documentthe impact of the choice of a question construction algorithm and targetdifficulty on student perception of question suitability and on their will-ingness to use the system. We also propose and evaluate a mechanismfor a dynamic difficulty adjustment.1 IntroductionThe goal of adaptive educational systems is to make learning more effective andengaging by tailoring the behaviour of the system to a particular student. Theadaptive behaviour is based on student models which estimate the knowledge ofstudents (and potentially other characteristics like their affective state). Whilea lot of research has focused on development and evaluation of student models,relatively little attention has been devoted to the way the outputs of models areactually used in educational systems. The typical use of student models is formastery learning, e.g. research studies 3,9,6 have evaluated the impact of usedmodels and their thresholds on over-practice and under-practice.The use of student models only for judging mastery is, however, only onepossible way of making a system adaptive to behaviour of its users. Adaptiveeducational systems have the potential to make learning more engaging by keep-ing students in the concentrated flow state 4. One of the conditions for theflow state is the balance between skills and difficulty of presented problems. TheInverted-U Hypothesis predicts that maximum engagement occurs with moder-ate challenge 10. There is extensive research on this topic (e.g. 1); the researchis, however, based mainly on laboratory studies, the results of research are toa certain degree contradictory (see e.g. the discussion in 10), and it is notclear how to apply the hypothesis in the development of a practical educationalapplication.In this work we connect the use of student models with the research onoptimal level of challenge. We study the impact of adaptive behaviour of aneducational system on student motivation in a widely used educational systemfor learning geographical facts.In the previous works the specification of the adaptive behaviour was basedmainly on intuition of system developers and was not evaluated 8,11 or wasevaluated using only comparison to a control group without any tuning of thedifficulty 2. The most similar research is by Lomas et al. 10 who evaluated theInverted-U Hypothesis by testing many variants of an educational game (num-berline estimation). They failed to find the U-shaped relation between difficultyand motivation. For their study the relation was monotone (simpler problemswere more engaging). Explaining the result they state that maybe they “nevermade the game easy enough” 10. Our experiments are similar, the main dif-ference is that we use a more realistic educational application. Another similarresearch was done using Math Garden software 7. The authors compared threeconditions (target success rate 60%, 75%, 90%) and showed that the easiestcondition led to the best learning (mediated by a number of solved problems).For our work we use a widely used application 11 for learning geography. Wehave performed randomized online experiments (multivariate testing) to evaluatethe impact of the adaptive behaviour on student motivation. The appropriatedifficulty of questions is evaluated using proxy measure of student motivation(number of questions answered) and student self-reports (perception of questiondifficulty). The results show that the adaptive behaviour is advantageous andthat the suitable portion of correct answers per user (success rate) is around 65%(with students who used the system in school preferring easier questions). Wealso propose a dynamic difficulty adjustment of the target success rate and weshow that this mechanism improves the adaptive system b

x

Hi!
I'm Simon!

Would you like to get a custom essay? How about receiving a customized one?

Check it out